Every source behind the read

63 sources in Engineering leaders and teams, with product documentation Showing all 94 sources

1. No opinion stated

   Stack Overflow, 2026 Developer Survey

   AI 2026 | Stack Overflow Developer SurveyDeveloper survey, Stack Overflow

   Stack Overflow's 2026 Developer Survey reports Claude Code (66%) and GitHub Copilot (59%) as the leaders for coding agents among respondents.

   "What are they using? Claude Code (66%) and GitHub Copilot (59%) are the leaders for coding agents."

   Published Oct 6, 2026self-described: 30,903 valid responses collected June 23 to August 5, 2026

   Practitioner account, self-described; second-hand; not counted, takes no position How GILD checked it

   Origin

   Published practitioner account

   Checked

   Self-described; Second-hand

   Role

   Name self-described; employer unknown; role unknown

   Relationship

   Checked only from the record itself; an outside check is open

   How they know

   Reports others' experience

   Support

   No artifact or independent corroboration recorded

   Date and version

   2026-10-06

   Not counted

   no aspect: excluded, takes no position on any assessed aspect

   Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

   Published on

   Engineering Blogs, Studies & Reviews

   Who

   Organization-wide

   Evidence

   Measured

   Applies to

   Current

2. Mixed

   Cem Dilmegani, with Şevval Alper, AIMultiple

   AI Code Review Tools BenchmarkIndependent benchmark, AIMultiple

   Replaying 309 pull requests, AIMultiple ranked Copilot code review third behind CodeRabbit and Greptile; its suggestions were accurate and good at catching typos, with shorter analysis than the two leaders.

   "GitHub Copilot is very successful at finding typos in code and makes spot-on suggestions; its analysis is shorter than those of CodeRabbit and Greptile."

   Published Sep 22, 2026self-described: AIMultiple RevEval: 309 pull requests replayed through CodeRabbit, Greptile, GitHub Copilot, and Cursor Bugbot, with developers grading a subset

   Practitioner account, self-described; structured test; counts toward Delivered outcomes, Adoption and upkeep How GILD checked it

   Origin

   Published practitioner account

   Checked

   Self-described; Ran a structured test; Specific details; Observed on November 2025 versions of all products

   Role

   Name self-described; employer unknown; role unknown

   Relationship

   Checked only from the record itself; an outside check is open

   How they know

   Ran a structured test

   Support

   No artifact or independent corroboration recorded

   Date and version

   2026-01-14, November 2025 versions of all products

   Counts toward

   Delivered outcomes, measured outcomes: mixed, G2 (2 units before bounds), capped: C-relationship

   Counts toward

   Adoption and upkeep, setup: favorable, G2 (2 units before bounds), capped: C-relationship

   Not counted

   Delivered outcomes, correctness: repeats a stronger account in its group

   Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

   Counts toward

   • Overall signal
   • GitHub Integration: Positive
   • Quality Against Alternatives: Raised as a concern

   Published on

   Engineering Blogs, Studies & Reviews

   Who

   Professional team

   Evidence

   Measured

   Applies to

   Current

   Observed on

   Copilot code review on default settings

3. Mixed

   Yaroslav (Yarik), tech lead for the coding-agent rollout, Superhuman

   How Superhuman Scaled AI Coding Agents Across Their Entire Eng OrgConference talk video

   Superhuman's engineers started with Copilot's autocomplete, which they found useful, then moved to Cursor and Claude Code for agent work.

   "…a year and a half ago, we first experienced Copilot. It was very cool auto complete."

   Published Aug 18, 2026self-described: Superhuman, formerly Grammarly, engineering organization

   Counts toward

   • Overall signal
   • Inline Completions: Positive
   • Is Copilot the main AI coding tool, or the one engineers keep beside another?: Moved to another tool

   Published on

   Podcasts, Talks & Videos

   Who

   Organization-wide

   Evidence

   Reported

   Applies to

   Current

4. Mixed

   Kon5ole, Hacker News commenter

   Comment on: Kimi K2.7 Code is generally available in GitHub CopilotHacker News comment

   A fan of Copilot CLI and its model switching whose whole department moved to Claude Code after the June 2026 billing change burned the monthly budget every two to three days.

   Published Jul 2, 2026self-described: Engineering department at work, Copilot CLI

   Practitioner account, pseudonymous; direct use; counts toward Operational control, Cost predictability and limits How GILD checked it

   Origin

   Published practitioner account

   Checked

   Pseudonymous; Customer of GitHub Copilot; Used it directly; Specific details; Observed on Opus 4.6

   Role

   Pseudonymous account; identity not pursued

   Relationship

   customer (self disclosed)

   How they know

   Used it directly

   Support

   No artifact or independent corroboration recorded

   Date and version

   2026-07-02, Opus 4.6

   Counts toward

   Operational control, models and deployment: favorable, G2 (2 units before bounds)

   Counts toward

   Cost predictability and limits, cost predictability: unfavorable, G2 (2 units before bounds)

   Counts toward

   Cost predictability and limits, limits in practice: unfavorable, G2 (2 units before bounds)

   Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

   Counts toward

   • Overall signal
   • Agent Execution: Positive
   • Value for Money: Mixed or negative
   • Usage / Cost: Raised as a concern
   • Is Copilot the main AI coding tool, or the one engineers keep beside another?: Moved to another tool
   • Does Copilot's usage-based billing work for the people paying for it?: Unpredictable

   Published on

   Developer Communities

   Who

   Professional team

   Evidence

   Reported

   Applies to

   Current

   Observed on

   Copilot CLI, AI credits billing from June 2026

5. Negative

   kasey_junk, Hacker News commenter

   Comment on: Kimi K2.7 Code is generally available in GitHub CopilotHacker News comment

   When vendor management at work assumes Claude through Copilot equals Claude Code, their side-by-side runs show Copilot underperforming three or four times out of five.

   Published Jul 2, 2026self-described: Engineer asked to explain tool differences to vendor management at work

   Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account How GILD checked it

   Origin

   Published practitioner account

   Checked

   Pseudonymous; No tie to GitHub Copilot found in GILD's check; Used it directly

   Role

   Pseudonymous account; identity not pursued

   Relationship

   No tie to GitHub Copilot found in GILD's check (byline, content disclosure, platform api profile); that is not proof none exists

   How they know

   Used it directly

   Support

   No artifact or independent corroboration recorded

   Date and version

   2026-07-02

   Not counted

   Delivered outcomes, task completion: repeats a stronger account in its group

   Checked Oct 30, 2026 under gild-policy/1.0.0. The scorecard

   Counts toward

   • Overall signal
   • Agent Execution: Mixed or negative
   • Quality Against Alternatives: Raised as a concern
   • Is Copilot the main AI coding tool, or the one engineers keep beside another?: Moved to another tool

   Published on

   Developer Communities

   Who

   Professional team

   Evidence

   Reported

   Applies to

   Current

   Observed on

   Claude models through Copilot versus Claude Code

6. Positive

   K3UL, Hacker News commenter

   Comment on: Kimi K2.7 Code is generally available in GitHub CopilotHacker News comment

   Their company compared costs after the June change and kept Copilot: a seat includes matching usage at provider prices across several model vendors, unlike Claude Enterprise.

   Published Jul 2, 2026self-described: Company that compared Copilot and Claude Enterprise after the June 2026 change

   Practitioner account, pseudonymous; opinion; not counted, repeats a stronger account How GILD checked it

   Origin

   Published practitioner account

   Checked

   Pseudonymous; No tie to GitHub Copilot found in GILD's check; Opinion; Specific details

   Role

   Pseudonymous account; identity not pursued

   Relationship

   No tie to GitHub Copilot found in GILD's check (byline, content disclosure, platform api profile); that is not proof none exists

   How they know

   Opinion, no use stated

   Support

   No artifact or independent corroboration recorded

   Date and version

   2026-07-02

   Not counted

   Cost predictability and limits, cost transparency: repeats a stronger account in its group

   Not counted

   Operational control, models and deployment: repeats a stronger account in its group

   Checked Oct 30, 2026 under gild-policy/1.0.0. The scorecard

   Counts toward

   • Enterprise Administration: Positive
   • Value for Money: Positive
   • Usage / Cost: Discussed without concern
   • Is Copilot the main AI coding tool, or the one engineers keep beside another?: Main tool
   • Does Copilot's usage-based billing work for the people paying for it?: Workable

   Published on

   Developer Communities

   Who

   Technical leadership

   Evidence

   Reported

   Applies to

   Current

   Observed on

   AI credits billing after June 2026

7. Positive

   Emerson Murphy-Hill, Jenna Butler, and Alexandra Savelieva, Microsoft

   Adoption and Impact of Command-Line AI Coding Agents: A Study of Microsoft's Early 2026 Rollout of Claude Code and GitHub Copilot CLIResearch paper, arXiv

   Studying tens of thousands of Microsoft engineers in the early-2026 rollout of Claude Code and GitHub Copilot CLI, the authors find first use spread mainly through social networks and adopters merged roughly 24% more pull requests than they otherwise would, with merged PRs as a proxy for output.

   Published Jul 1, 2026self-described: Microsoft engineering, tens of thousands of engineers (Microsoft owns GitHub)

   Practitioner account; structured test; not counted, tied to the vendor How GILD checked it

   Origin

   Published practitioner account

   Checked

   GitHub Copilot staff: not counted as independent; Ran a structured test; Specific details

   Role

   Name unknown; employer unknown; role unknown

   Relationship

   vendor employee (suspected)

   How they know

   Ran a structured test

   Support

   No artifact or independent corroboration recorded

   Date and version

   2026-07-01

   Not counted

   Delivered outcomes, measured outcomes: excluded, praise from someone tied to the vendor, kept as an attributed position

   Checked Oct 29, 2026 under gild-policy/1.0.0. The scorecard

   Not counted

   Its authors work for Microsoft, which owns GitHub. Kept in the record; it takes part in no figure.

   Published on

   Engineering Blogs, Studies & Reviews

   Who

   Organization-wide

   Evidence

   Measured

   Applies to

   Current

   Observed on

   Copilot CLI and Claude Code, January 5 to April 29, 2026

8. Negative

   Jeff Davis, VP of Core Infrastructure at Indeed

   2x the power users: How structured AI training scaled developer productivityConference session page with transcript, DX Annual via Engineering Enablement by DX

   With Cody and Copilot, budgets and training in place, weekly AI use at Indeed sat near 25% in early 2025; Indeed later retired both once their usage fell to about 3% and moved engineers to agentic tools such as Claude Code and Cursor.

   "We had Cody, Copilot, we had unlimited budgets, we had training, we had lots of leadership talk about using AI more. We were still hovering at around 25% of developers actively using AI tools on a weekly basis."

   "And then we actually have phased out Cody and Copilot. So sort of that first generation of tools we’ve actually already fully retired. Our usage when we actually did retire them was down to about 3% of people. So our users actually abandoned them before we did."

   Published Jun 29, 2026Job site with more than 2,000 engineers; talk given with product manager Michael Redding

   Counts toward

   • Overall signal
   • Licenses alone do not produce the gain
   • Is Copilot the main AI coding tool, or the one engineers keep beside another?: Moved to another tool

   Published on

   Engineering Blogs, Studies & Reviews

   Who

   Technical leadership

   Evidence

   Reported

   Applies to

   Current

   Observed on

   Copilot and Cody as Indeed's first AI coding tools, later retired

9. Positive

   Ty Smith, Principal Engineer at Uber

   Uber’s journey of measuring AI impact on developer productivityConference session page with transcript, DX Annual via Engineering Enablement by DX

   Smith reports that when Copilot and Cursor launched, Uber lacked usable telemetry, so it ran qualitative surveys before layering in causal methods. Co-presenter Abhishek Tibrewal reports a difference-in-differences study found a 19% incremental lift in PRs per engineer for Copilot power users, about 8% of engineers.

   "The core question was, if there was no Copilot, what would have the same developer shipped? And that’s your counterfactual, and the delta between that counterfactual and what the developers actually shipped is the incremental impact that we were looking at. We got 19% incremental impact for this power user cohort."

   Published Jun 22, 2026Uber developer platform team, presented with colleague Abhishek Tibrewal

   Practitioner account, self-described; ran the rollout; counts toward Delivered outcomes How GILD checked it

   Origin

   Published practitioner account

   Checked

   Self-described; Customer of GitHub Copilot; Ran the rollout; Artifact checked (benchmark output)

   Role

   Name self-described; employer self-described; role self-described (Principal Engineer at Uber)

   Relationship

   customer (self disclosed); customer (self disclosed); customer (self disclosed)

   How they know

   Ran the rollout

   Support

   1 artifact (benchmark output)

   Date and version

   2026-06-22

   Counts toward

   Delivered outcomes, measured outcomes: mixed, G1 (1 unit before bounds), capped: C-relationship, C-unknown-control

   Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

   Counts toward

   • Overall signal
   • Measuring Impact: Discussed without concern
   • Once review and rework are counted, does Copilot reduce engineering time?: Net time saved

   Published on

   Engineering Blogs, Studies & Reviews

   Who

   Professional team

   Evidence

   Measured

   Applies to

   Current

   Observed on

   Copilot and Cursor in Uber's AI-assisted era, measured with a difference-in-differences study

10. Mixed

    Kelly Anne Pipe, Head of Developer Experience at Vanguard

    Augmented, accelerated, autonomized: How Vanguard is embedding AI across the product lifecycleConference session page with transcript, DX Annual via Engineering Enablement by DX

    Vanguard rolled Copilot out to its engineers and found pockets of power users, but faster coding alone did not speed delivery because the rest of each product team still worked at the old pace.

    "We had rolled out Copilot for engineers. There were some pockets of brilliance. We had some power users doing really incredible things."

    Published Jun 15, 2026Asset manager with more than 800 product teams; Copilot licensed for engineers

    Practitioner account, self-described; ran the rollout; counts toward Delivered outcomes How GILD checked it

    Origin

    Published practitioner account

    Checked

    Self-described; Customer of GitHub Copilot; Ran the rollout

    Role

    Name self-described; employer self-described; role self-described (Kelly Anne Pipe is Head of Developer Experience at Vanguard)

    Relationship

    customer (self disclosed); customer (self disclosed)

    How they know

    Ran the rollout

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-06-15

    Counts toward

    Delivered outcomes, measured outcomes: mixed, G1 (1 unit before bounds), capped: C-relevance, C-relationship, C-unknown-control

    Not counted

    Delivered outcomes, task completion: repeats a stronger account in its group

    Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Overall signal
    • Measuring Impact: Discussed without concern
    • Licenses alone do not produce the gain

    Published on

    Engineering Blogs, Studies & Reviews

    Who

    Technical leadership

    Evidence

    Reported

    Applies to

    Current

    Observed on

    Copilot rolled out to engineers across the firm

11. Mixed

    Brian Scanlan, Senior Principal Systems Engineer at Intercom

    Doubling the productivity of your engineering team using AIConference session page with transcript, DX Annual via Engineering Enablement by DX

    Intercom's first AI step was GitHub Copilot plus engineers trying Cursor; throughput metrics rose somewhat but the company was dissatisfied, set a goal to double throughput and standardized on Claude Code.

    "And we kind of went on a normal journey, I think, like of adopting the likes of GitHub Copilot and bottoms up kind of people using Cursor. And we kind of got like some results, some improvements, throughput metrics, different things will kind of go up, but we’re ultimately dissatisfied with the results."

    Published Jun 15, 2026Customer service software company; leads developer productivity and AI adoption on large Rails and JavaScript apps

    Practitioner account, self-described; ran the rollout; counts toward Delivered outcomes How GILD checked it

    Origin

    Published practitioner account

    Checked

    Self-described; Competitor staff: not counted as independent; Ran the rollout

    Role

    Name self-described; employer self-described; role unknown

    Relationship

    partner (confirmed); customer (self disclosed); customer (self disclosed)

    How they know

    Ran the rollout

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-06-15

    Counts toward

    Delivered outcomes, task completion: mixed, G1 (1 unit before bounds), capped: C-unknown-control

    Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Overall signal
    • Is Copilot the main AI coding tool, or the one engineers keep beside another?: Moved to another tool

    Published on

    Engineering Blogs, Studies & Reviews

    Who

    Technical leadership

    Evidence

    Reported

    Applies to

    Current

    Observed on

    GitHub Copilot alongside engineer-led Cursor use, before Intercom standardized on Claude Code

12. Mixed

    Jason Valentino, Head of Software Engineering Strategy at BNY

    Mapping the new SDLC at BNY: Codifying AI into every step of the delivery lifecycleConference session page with transcript, DX Annual via Engineering Enablement by DX

    Valentino reports that BNY rolled out Copilot in 2024 and brought the first strong use cases back to its community of practice. Once coding assistants were everywhere, BNY was "not exactly" making more money, so it looked across the whole delivery lifecycle; it now uses Windsurf, Copilot and Claude Code.

    "It’s 2024, we’re rolling out Copilot. As the first couple use cases show up where someone’s like, “Wow, this is really good at this.” Then come back to the community of practice, get those on there."

    Published Jun 8, 2026Bank with more than 8,000 engineers; owns developer tooling and DevEx

    Practitioner account, self-described; ran the rollout; not counted, takes no position How GILD checked it

    Origin

    Published practitioner account

    Checked

    Self-described; Customer of GitHub Copilot; Ran the rollout

    Role

    Name self-described; employer self-described; role self-described (Jason Valentino is Head of Software Engineering Strategy at BNY)

    Relationship

    customer (self disclosed); customer (self disclosed); customer (self disclosed); customer (self disclosed)

    How they know

    Ran the rollout

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-06-08

    Not counted

    no aspect: excluded, takes no position on any assessed aspect

    Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Measuring Impact: Discussed without concern
    • Licenses alone do not produce the gain

    Published on

    Engineering Blogs, Studies & Reviews

    Who

    Technical leadership

    Evidence

    Reported

    Applies to

    Current

    Observed on

    Copilot rollout in 2024, now one of several coding tools beside Windsurf and Claude Code

13. Positive

    Jesse Houwing, GitHub and Scrum trainer

    The why for Usage Based Billing for GitHub CopilotPersonal blog post, jessehouwing.net

    Explains the move to AI credits: request multipliers also steered load between models, and Copilot capped Claude Opus 4.6 at a 200K-token window where Anthropic offers 1M directly.

    "For example, if you're using Claude Opus 4.6 through GitHub Copilot, the maximum token window is 200k tokens. If you use the same model directly through Anthropic you have a maximum of 1m tokens."

    Published May 11, 2026self-described: Advises organizations on Copilot Business and Enterprise

    Practitioner account, self-described; opinion; not counted until substantiated How GILD checked it

    Origin

    Published practitioner account

    Checked

    Self-described; GitHub Copilot partner: not counted as independent; Opinion; Specific details

    Role

    Name self-described; employer unknown; role unknown

    Relationship

    program member (suspected)

    How they know

    Opinion, no use stated

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-05-11

    Not counted

    Cost predictability and limits, cost transparency: a lead (G0), counts toward nothing until substantiated

    Not counted

    Cost predictability and limits, cost predictability: a lead (G0), counts toward nothing until substantiated

    Not counted

    Operational control, admin control: excluded, praise from someone tied to the vendor, kept as an attributed position

    Not counted

    Operational control, models and deployment: a lead (G0), counts toward nothing until substantiated

    Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

    Not counted

    Its author is a Microsoft MVP, a program run by GitHub's parent Microsoft. Kept in the record; it takes part in no figure.

    Published on

    Engineering Blogs, Studies & Reviews

    Who

    Professional team

    Evidence

    Reported

    Applies to

    Current

    Observed on

    Premium requests before June 1, 2026, and the announced AI credits

14. Mixed

    on_the_train, Hacker News commenter

    Comment on: GitHub Copilot is moving to usage-based billingHacker News comment

    At a large semiconductor company Copilot is the only approved LLM tool and approving another would take years, so the company will pay the new price.

    Published Apr 28, 2026self-described: Large semiconductor company where Copilot is the only approved LLM tool

    Practitioner account, pseudonymous; opinion; not counted, takes no position How GILD checked it

    Origin

    Published practitioner account

    Checked

    Pseudonymous; No tie to GitHub Copilot found in GILD's check; Opinion

    Role

    Pseudonymous account; identity not pursued

    Relationship

    No tie to GitHub Copilot found in GILD's check (byline, content disclosure, platform api profile); that is not proof none exists

    How they know

    Opinion, no use stated

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-04-28

    Not counted

    no aspect: excluded, takes no position on any assessed aspect

    Checked Oct 29, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Usage / Cost: Discussed without concern
    • Chosen as the approved default
    • Is Copilot the main AI coding tool, or the one engineers keep beside another?: Main tool

    Published on

    Developer Communities

    Who

    Organization-wide

    Evidence

    Reported

    Applies to

    Current

15. Mixed

    sandos, Hacker News commenter

    Comment on: GitHub Copilot is moving to usage-based billingHacker News comment

    Has Copilot at work and personally. On their own and nearby teams, about one in three people use it much, so pooled credits depend on a few heavy users.

    Published Apr 28, 2026self-described: Copilot at work across several teams, plus a personal subscription

    Practitioner account, pseudonymous; second-hand; not counted, takes no position How GILD checked it

    Origin

    Published practitioner account

    Checked

    Pseudonymous; No tie to GitHub Copilot found in GILD's check; Second-hand

    Role

    Pseudonymous account; identity not pursued

    Relationship

    No tie to GitHub Copilot found in GILD's check (byline, content disclosure, platform api profile); that is not proof none exists

    How they know

    Reports others' experience

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-04-28

    Not counted

    no aspect: excluded, takes no position on any assessed aspect

    Checked Oct 30, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Usage / Cost: Discussed without concern
    • Measuring Impact: Discussed without concern
    • Licenses alone do not produce the gain

    Published on

    Developer Communities

    Who

    Professional team

    Evidence

    Reported

    Applies to

    Current

    Observed on

    Pooled credits under usage-based billing

16. Negative

    insane_dreamer, Hacker News commenter

    Comment on: GitHub Copilot is moving to usage-based billingHacker News comment

    The billing announcement showed their small team had been paying for Copilot through its GitHub bill while nobody used it, so they canceled it.

    Published Apr 27, 2026self-described: Small team paying for Copilot through the GitHub bill

    Practitioner account, pseudonymous; ran the rollout; not counted, repeats a stronger account How GILD checked it

    Origin

    Published practitioner account

    Checked

    Pseudonymous; Customer of GitHub Copilot; Ran the rollout; Specific details

    Role

    Pseudonymous account; identity not pursued

    Relationship

    customer (self disclosed)

    How they know

    Ran the rollout

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-04-27

    Not counted

    Cost predictability and limits, cost predictability: repeats a stronger account in its group

    Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Overall signal
    • Usage / Cost: Raised as a concern
    • Measuring Impact: Raised as a concern
    • Is Copilot the main AI coding tool, or the one engineers keep beside another?: Moved to another tool

    Published on

    Developer Communities

    Who

    Very small team

    Evidence

    Reported

    Applies to

    Current

17. Negative

    koyote, Hacker News commenter

    Comment on: GitHub Copilot is moving to usage-based billingHacker News comment

    At a Microsoft shop, data governance makes switching providers hard, and they expect the price change to make Copilot ROI negative and cut AI use in the company.

    Published Apr 27, 2026self-described: Microsoft-standardized company using Copilot

    Practitioner account, pseudonymous; opinion; not counted, takes no position How GILD checked it

    Origin

    Published practitioner account

    Checked

    Pseudonymous; No tie to GitHub Copilot found in GILD's check; Opinion

    Role

    Pseudonymous account; identity not pursued

    Relationship

    No tie to GitHub Copilot found in GILD's check (byline, content disclosure, platform api profile); that is not proof none exists

    How they know

    Opinion, no use stated

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-04-27

    Not counted

    no aspect: excluded, takes no position on any assessed aspect

    Checked Oct 29, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Value for Money: Mixed or negative
    • Usage / Cost: Raised as a concern
    • Data, IP & Security: Discussed without concern
    • Measuring Impact: Raised as a concern
    • Chosen as the approved default
    • Does Copilot's usage-based billing work for the people paying for it?: Unpredictable

    Published on

    Developer Communities

    Who

    Professional team

    Evidence

    Reported

    Applies to

    Current

    Observed on

    Announced move to usage-based billing

18. Negative

    isqueiros, Hacker News commenter

    Comment on: Changes to GitHub Copilot individual plansHacker News comment

    Their company holds company-wide Copilot, Windsurf and ChatGPT Enterprise accounts, yet nearly everyone uses Claude Code through Bedrock instead.

    Published Apr 22, 2026self-described: Company with company-wide Windsurf, Copilot, and ChatGPT Enterprise accounts

    Practitioner account, pseudonymous; second-hand; not counted, takes no position How GILD checked it

    Origin

    Published practitioner account

    Checked

    Pseudonymous; No tie to GitHub Copilot found in GILD's check; Second-hand

    Role

    Pseudonymous account; identity not pursued

    Relationship

    No tie to GitHub Copilot found in GILD's check (byline, content disclosure, platform api profile); that is not proof none exists

    How they know

    Reports others' experience

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-04-22

    Not counted

    no aspect: excluded, takes no position on any assessed aspect

    Checked Oct 29, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Measuring Impact: Raised as a concern
    • Is Copilot the main AI coding tool, or the one engineers keep beside another?: Moved to another tool

    Published on

    Developer Communities

    Who

    Organization-wide

    Evidence

    Reported

    Applies to

    Current

19. Mixed

    Mikhail Parakhin, chief technology officer, Shopify

    Latent Space: Shopify with Mikhail ParakhinPodcast episode with publisher transcript

    Shopify's usage chart shows IDE tools such as GitHub Copilot not shrinking but growing more slowly than command-line agents since December 2025.

    "…tools, uh, that require IDEs such as, uh, GitHub, Copilot or Cursor, they’re not exactly shrinking, but they’re not growing as fast. Like, uh, red, red line is, is the IDE kind of tools."

    Published Apr 22, 2026self-described: Shopify, every employee chooses their coding tool

    Published on

    Podcasts, Talks & Videos

    Who

    Organization-wide

    Evidence

    Reported

    Applies to

    Current

20. Negative

    qxxx, Hacker News commenter

    Comment on: How many products does Microsoft have named 'Copilot'?Hacker News comment

    Their boss told developers paying for personal Copilot to use the company's Copilot, which turned out to be the Microsoft 365 product, a different tool from GitHub Copilot.

    Published Apr 5, 2026self-described: Workplace where developers paid for personal Copilot accounts

    Practitioner account, pseudonymous; second-hand; not counted, takes no position How GILD checked it

    Origin

    Published practitioner account

    Checked

    Pseudonymous; No tie to GitHub Copilot found in GILD's check; Second-hand

    Role

    Pseudonymous account; identity not pursued

    Relationship

    No tie to GitHub Copilot found in GILD's check (byline, content disclosure, platform api profile); that is not proof none exists

    How they know

    Reports others' experience

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-04-05

    Not counted

    no aspect: excluded, takes no position on any assessed aspect

    Checked Oct 30, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Enterprise Administration: Mixed or negative

    Published on

    Developer Communities

    Who

    Professional team

    Evidence

    Reported

    Applies to

    Current

21. Negative

    kevcampb, Hacker News commenter

    Comment on: Updates to GitHub Copilot interaction data usage policyHacker News comment

    The 2026 interaction data policy change pushed their company into checking vendor by vendor that its code is not used to train models.

    Published Mar 29, 2026self-described: Company protecting its code from model training

    not counted, repeats a stronger account How GILD checked it

    Role

    Pseudonymous account; identity not pursued

    Relationship

    No tie to GitHub Copilot found in GILD's check (byline, content disclosure, platform api profile); that is not proof none exists

    How they know

    Opinion, no use stated

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-03-29

    Not counted

    Operational control, data control: repeats a stronger account in its group

    Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Data, IP & Security: Raised as a concern

    Published on

    Developer Communities

    Who

    Professional team

    Evidence

    Reported

    Applies to

    Current

    Observed on

    March 2026 interaction data policy change

22. Mixed

    parkersweb, Hacker News commenter who manages a GitHub organization

    Comment on: Updates to GitHub Copilot interaction data usage policyHacker News comment

    Managing an organization that licenses Copilot for some developers, they received the data policy email with no mention that Business seats were excluded.

    Published Mar 25, 2026self-described: Organization that gives Copilot to some of its developers

    Practitioner account, pseudonymous; ran the rollout; not counted, takes no position How GILD checked it

    Origin

    Published practitioner account

    Checked

    Pseudonymous; No tie to GitHub Copilot found in GILD's check; Ran the rollout; Observed on business copilot

    Role

    Pseudonymous account; identity not pursued

    Relationship

    No tie to GitHub Copilot found in GILD's check (byline, content disclosure, platform api profile); that is not proof none exists

    How they know

    Ran the rollout

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-03-25, business copilot

    Not counted

    no aspect: excluded, takes no position on any assessed aspect

    Checked Oct 29, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Enterprise Administration: Mixed or negative
    • Data, IP & Security: Discussed without concern

    Published on

    Developer Communities

    Who

    Technical leadership

    Evidence

    Reported

    Applies to

    Current

    Observed on

    March 2026 interaction data policy change

23. Negative

    mbenzakis, GitHub Community member on GitHub Enterprise Cloud

    Title: Copilot Code Review generates new comments on every push, creating an endless fix-push-review loopGitHub Community discussion, Copilot Conversations

    On GitHub Enterprise Cloud, Copilot code review raised new comments on unchanged code after every fix push, taking one pull request through 5 rounds of review, each 15-30 minutes. Of ~24 comments across all rounds, ~3 were genuinely useful.

    "Total: 5 rounds of review for a single PR. Each round took 15-30 minutes. The same pattern repeated on PR B."

    Published Mar 16, 2026self-described: Organization with Copilot code review enabled, Python tooling repositories

    Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes, Reliability How GILD checked it

    Origin

    Published practitioner account

    Checked

    Pseudonymous; No tie to GitHub Copilot found in GILD's check; Used it directly; Specific details; Observed on GitHub Enterprise Cloud

    Role

    Pseudonymous account; identity not pursued

    Relationship

    No tie to GitHub Copilot found in GILD's check (byline, content disclosure, platform api profile); that is not proof none exists

    How they know

    Used it directly

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-03-16, GitHub Enterprise Cloud

    Counts toward

    Reliability, run stability: unfavorable, G2 (2 units before bounds)

    Counts toward

    Delivered outcomes, task completion: unfavorable, G2 (2 units before bounds)

    Counts toward

    Delivered outcomes, correctness: unfavorable, G2 (2 units before bounds)

    Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Overall signal
    • GitHub Integration: Mixed or negative
    • Code Quality: Mixed or negative

    Published on

    Developer Communities

    Who

    Professional team

    Evidence

    Measured

    Applies to

    Current

    Observed on

    Copilot code review on GitHub Enterprise Cloud; fixes made with Copilot in VS Code on Claude Opus 4.6

24. No opinion stated

    hades200082, GitHub Community member writing for their company

    The #1 reason my company won't use Github Copilot right nowGitHub Community discussion, Copilot Conversations

    Their company will not roll Copilot out because assigning a company seat cancels a developer's personal Copilot subscription on the same account, which ties side projects to company tooling and IP terms.

    "When our company assigns a Copilot seat to a developer, GitHub automatically cancels their individual Copilot subscription on that same account."

    Published Mar 6, 2026self-described: Company on GitHub for all repositories, already using Claude Code

    Practitioner account, pseudonymous; opinion; not counted until substantiated How GILD checked it

    Origin

    Published practitioner account

    Checked

    Pseudonymous; No tie to GitHub Copilot found in GILD's check; Opinion; Specific details

    Role

    Pseudonymous account; identity not pursued

    Relationship

    customer (self disclosed)

    How they know

    Opinion, no use stated

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-03-06

    Not counted

    Operational control, admin control: a lead (G0), counts toward nothing until substantiated

    Checked Oct 30, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Enterprise Administration: Mixed or negative
    • Data, IP & Security: Raised as a concern

    Published on

    Developer Communities

    Who

    Professional team

    Evidence

    Reported

    Applies to

    Current

    Observed on

    Copilot Business and Enterprise seat assignment

25. Mixed

    Viktoria Stray, University of Oslo and SINTEF, with Elias Goldmann Brandtzæg, Viggo Tellefsen Wivestad, Astri Barbala, and Nils Brede Moe

    Developer Productivity With and Without GitHub Copilot: A Longitudinal Mixed-Methods Case StudyResearch paper, arXiv (University of Oslo and SINTEF)

    Analyzing 26,317 commits from 25 Copilot users and 14 non-users at NAV IT over two years, the authors found no statistically significant change in commit activity after adoption, though minor increases appeared. Developers mainly used Copilot for boilerplate, documentation and repetitive tasks.

    "We did not find any statistically significant changes in commit-based activity for Copilot users after they adopted the tool, although minor increases were observed."

    Published Jan 28, 2026self-described: NAV IT, a large Norwegian public sector agile organization; 25 Copilot users and 14 non-users across 703 repositories

    Practitioner account; structured test; counts toward Delivered outcomes How GILD checked it

    Origin

    Published practitioner account

    Checked

    Ran a structured test; Artifact checked (benchmark output)

    Role

    Name unknown; employer unknown; role unknown

    Relationship

    Checked only from the record itself; an outside check is open

    How they know

    Ran a structured test

    Support

    1 artifact (benchmark output)

    Date and version

    2026-01-28

    Counts toward

    Delivered outcomes, measured outcomes: mixed, G1 (1 unit before bounds), capped: C-relationship, C-unknown-control

    Counts toward

    Delivered outcomes, correctness: favorable, G1 (1 unit before bounds), capped: C-relevance, C-relationship, C-unknown-control

    Counts toward

    Delivered outcomes, task completion: mixed, G1 (1 unit before bounds), capped: C-relationship, C-unknown-control

    Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Overall signal
    • Measuring Impact: Raised as a concern
    • Largest gains on routine code and tests

    Published on

    Engineering Blogs, Studies & Reviews

    Who

    Organization-wide

    Evidence

    Measured

    Applies to

    Still applicable

    Observed on

    GitHub Copilot in daily development, two years of commit history

26. Positive

    Grégoire Dauré, Sogeti (Capgemini), led Copilot adoption programs at Michelin and CA-CIB

    Adopting GitHub Copilot at Enterprise ScalePersonal blog post, gdaure.io

    Leading Copilot rollouts at Michelin and CA-CIB, found coaching mattered more than licenses, unit testing gained most, and CA-CIB test coverage rose 18 points in six months.

    Published Jan 15, 2026self-described: Michelin and CA-CIB, hundreds of developers

    Practitioner account, self-described; ran the rollout; counts toward Delivered outcomes, Adoption and upkeep How GILD checked it

    Origin

    Published practitioner account

    Checked

    Self-described; GitHub Copilot partner: not counted as independent; Ran the rollout; Observed on GitHub Copilot Business/Enterprise; Specific details

    Role

    Name self-described; employer unknown; role unknown

    Relationship

    consultant or reseller (suspected); consultant or reseller (self disclosed)

    How they know

    Ran the rollout

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-01-15, GitHub Copilot Business/Enterprise

    Counts toward

    Delivered outcomes, correctness: mixed, G1 (1 unit before bounds), capped: C-relationship

    Counts toward

    Delivered outcomes, context handling: unfavorable, G1 (1 unit before bounds), capped: C-relationship

    Counts toward

    Adoption and upkeep, setup: mixed, G1 (1 unit before bounds), capped: C-relationship

    Not counted

    Operational control, data control: excluded, praise from someone tied to the vendor, kept as an attributed position

    Not counted

    Operational control, admin control: excluded, praise from someone tied to the vendor, kept as an attributed position

    Not counted

    Delivered outcomes, measured outcomes: excluded, praise from someone tied to the vendor, kept as an attributed position

    Not counted

    Cost predictability and limits, cost transparency: excluded, praise from someone tied to the vendor, kept as an attributed position

    Checked Oct 29, 2026 under gild-policy/1.0.0. The scorecard

    Not counted

    Its author is a Sogeti consultant who sells GitHub Copilot adoption programs. Kept in the record; it takes part in no figure.

    Published on

    Engineering Blogs, Studies & Reviews

    Who

    Technical leadership

    Evidence

    Measured

    Applies to

    Current

    Observed on

    Copilot Business and Enterprise

27. Mixed

    alexander-jether, GitHub Community member

    How to prevent Copilot Coding Agent from being the main author of commits in masterGitHub Community discussion, Copilot Conversations

    When they squash-merge pull requests from the Copilot coding agent, GitHub sets Copilot as commit author, so git blame and history stop showing who requested and approved the work.

    "…when I merge it to master using "Squash and merge" via the GitHub UI, the resulting commit author is set to Copilot rather than the person who requested the task."

    Published Nov 19, 2025self-described: Team merging Copilot coding agent pull requests with squash and merge

    Practitioner account, pseudonymous; direct use; not counted, takes no position How GILD checked it

    Origin

    Published practitioner account

    Checked

    Pseudonymous; No tie to GitHub Copilot found in GILD's check; Used it directly

    Role

    Pseudonymous account; identity not pursued

    Relationship

    No tie to GitHub Copilot found in GILD's check (byline, content disclosure, platform api profile); that is not proof none exists

    How they know

    Used it directly

    Support

    No artifact or independent corroboration recorded

    Date and version

    2025-11-19

    Not counted

    no aspect: excluded, takes no position on any assessed aspect

    Checked Oct 29, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • GitHub Integration: Mixed or negative

    Published on

    Developer Communities

    Who

    Professional team

    Evidence

    Reported

    Applies to

    Still applicable

    Observed on

    Copilot coding agent, New Agent Task

28. Positive

    Angie Jones, maintainer of goose, Block's open source AI agent

    How I Taught GitHub Copilot Code Review to Think Like a MaintainerPersonal blog post, angiejones.tech

    goose maintainers found Copilot code review noisy, with about one comment in five a real catch, until repository instructions on confidence, priorities and CI coverage cut the noise.

    Published Nov 10, 2025self-described: goose maintainers, open source Rust project with a high volume of contributions

    Practitioner account, self-described; ran the rollout; counts toward Delivered outcomes, Adoption and upkeep How GILD checked it

    Origin

    Published practitioner account

    Checked

    Self-described; Ran the rollout; Specific details

    Role

    Name self-described; employer unknown; role unknown

    Relationship

    Checked only from the record itself; an outside check is open

    How they know

    Ran the rollout

    Support

    No artifact or independent corroboration recorded

    Date and version

    2025-11-11

    Counts toward

    Delivered outcomes, correctness: unfavorable, G2 (2 units before bounds), capped: C-relationship

    Counts toward

    Adoption and upkeep, ongoing upkeep: unfavorable, G1 (1 unit before bounds), capped: C-relevance, C-relationship

    Not counted

    Delivered outcomes, task completion: repeats a stronger account in its group

    Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Overall signal
    • GitHub Integration: Positive
    • Requires Strong Review: Discussed without concern
    • Written instructions improve results

    Published on

    Engineering Blogs, Studies & Reviews

    Who

    Professional team

    Evidence

    Reported

    Applies to

    Still applicable

    Observed on

    Copilot code review with .github/copilot-instructions.md

29. Mixed

    Fabien Deshayes, leads platform engineering teams at Monzo

    How Monzo runs data-driven AI experimentationPodcast episode page with transcript, Engineering Enablement by DX

    Monzo's AI coding work began when GitHub switched Copilot on in engineers' VS Code; many tried it but adoption stayed low, so the bank moved to small structured trials with set adoption and satisfaction metrics.

    "…our journey with AI started with GitHub turning on Copilot for us and suddenly this landed in our VS code. And that was quite interesting and nice and a lot of people starting trying it, but there was not a lot of adoption."

    Published Oct 31, 2025UK digital bank; regulated engineering organization running structured AI tool trials

    Practitioner account, self-described; ran the rollout; counts toward Cost predictability and limits, Adoption and upkeep How GILD checked it

    Origin

    Published practitioner account

    Checked

    Self-described; Customer of GitHub Copilot; Ran the rollout; Specific details

    Role

    Name self-described; employer self-described; role self-described (leads multiple platform engineering teams at Monzo Bank)

    Relationship

    customer (self disclosed)

    How they know

    Ran the rollout

    Support

    No artifact or independent corroboration recorded

    Date and version

    2025-10-31

    Counts toward

    Adoption and upkeep, setup: unfavorable, G1 (1 unit before bounds), capped: C-relationship, C-unknown-control

    Counts toward

    Cost predictability and limits, cost transparency: favorable, G1 (1 unit before bounds), capped: C-relationship, C-unknown-control

    Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Overall signal
    • Licenses alone do not produce the gain
    • Chosen as the approved default

    Published on

    Engineering Blogs, Studies & Reviews

    Who

    Technical leadership

    Evidence

    Reported

    Applies to

    Still applicable

    Observed on

    Copilot in VS Code, switched on for the whole engineering organization

30. Mixed

    James Camilleri, Lead Developer at Scott Logic

    Rapid web app development with Devin - A Developer’s PerspectiveEngineering blog post, Scott Logic blog

    Switching between Copilot and Devin on one app, found Copilot needed tight prompts, limited context and line-by-line review like a junior's work. Advises small pull requests so AI code does not swamp reviewers.

    "When working with Copilot I found I needed to rein it in to get good output. Given too much freedom it would overproduce inappropriate code and go down paths I wanted to avoid."

    Published Oct 20, 2025Scott Logic, UK software consultancy; carbon-calculator web app built with Copilot and Devin

    Practitioner account, name and role on their organization's own page; direct use; counts toward Delivered outcomes How GILD checked it

    Origin

    Published practitioner account

    Checked

    Name and role on their organization's own page; Used it directly; Needs recheck after 12 months

    Role

    Name on a page the person does not control; employer on a page the person does not control; role on a page the person does not control (I'm a Lead Developer at Scott logic based in the Bristol office.)

    Relationship

    Checked only from the record itself; an outside check is open

    How they know

    Used it directly

    Support

    No artifact or independent corroboration recorded

    Date and version

    2025-10-20, needs a recheck

    Counts toward

    Delivered outcomes, task completion: favorable, G1 (1 unit before bounds), capped: C-relationship

    Counts toward

    Delivered outcomes, correctness: unfavorable, G1 (1 unit before bounds), capped: C-relationship

    Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Overall signal
    • Inline Completions: Positive
    • Requires Strong Review: Raised as a concern
    • Needs Close Steering: Raised as a concern
    • Quality Against Alternatives: Discussed without concern

    Published on

    Engineering Blogs, Studies & Reviews

    Who

    Professional team

    Evidence

    Reported

    Applies to

    Still applicable

    Observed on

    Copilot alongside Devin on the same project

31. No opinion stated

    Suhail Patel, leads the platform teams, Monzo

    The Pragmatic Engineer: How tech companies measure the impact of AI on software developmentNewsletter account

    Monzo's first AI coding tool was GitHub Copilot, through VS Code, and the bank keeps investing in it while trying Cursor, Windsurf and Claude Code.

    "The very first one was GitHub Copilot. That's going to be quite a common answer for most companies, mostly because I think GitHub played their strategy very well, as Copilot ‘sneaked’ its way into VS Code."

    Published Sep 16, 2025self-described: Monzo engineering

    Published on

    Engineering Blogs, Studies & Reviews

    Who

    Organization-wide

    Evidence

    Reported

    Applies to

    Current

32. Positive

    Government Digital Service, Department for Science, Innovation and Technology

    AI coding assistant trial: UK public sector findings reportGovernment trial report, GOV.UK (DSIT and GDS)

    A three-month cross-government trial with 1,100 redeemed Copilot Business licenses reported 56 minutes saved per day, a 15.8% line acceptance rate, and 39% of users committing suggested code.

    Published Sep 12, 2025self-described: More than 50 UK public sector organizations; 1,600 GitHub Copilot Business licenses assigned, 1,100 redeemed

    Practitioner account, self-described; structured test; counts toward Delivered outcomes, Adoption and upkeep How GILD checked it

    Origin

    Published practitioner account

    Checked

    Self-described; Customer of GitHub Copilot; Ran a structured test; Needs recheck after 12 months; Artifact checked (dataset)

    Role

    Name self-described; employer unknown; role unknown

    Relationship

    customer (self disclosed)

    How they know

    Ran a structured test

    Support

    1 artifact (dataset)

    Date and version

    2025-09-12, GitHub Copilot business, needs a recheck

    Counts toward

    Delivered outcomes, measured outcomes: mixed, G2 (2 units before bounds), capped: C-relationship

    Counts toward

    Delivered outcomes, task completion: favorable, G2 (2 units before bounds), capped: C-relationship

    Counts toward

    Adoption and upkeep, setup: unfavorable, G1 (1 unit before bounds), capped: C-relationship

    Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Overall signal
    • Inline Completions: Mixed or negative
    • Measuring Impact: Discussed without concern
    • Licenses alone do not produce the gain
    • Once review and rework are counted, does Copilot reduce engineering time?: Net time saved

    Published on

    Engineering Blogs, Studies & Reviews

    Who

    Organization-wide

    Evidence

    Measured

    Applies to

    Still applicable

    Observed on

    Copilot Business, November 2024 to February 2025, alongside Gemini Code Assist

33. Mixed

    Jesse Adametz, Senior Engineering Leader on the Developer Platform at Twilio

    Lessons from Twilio’s multi-year platform consolidationPodcast episode page with transcript, Engineering Enablement by DX

    Adametz argues that Twilio has used Copilot for a long time and now tests Claude Code against what people did with Copilot, without knowing yet whether it is the right tool. Agrees that handing out a Copilot license is "a piece of the puzzle" of enablement, with a large way still to go.

    "We’ve used Copilot for a long time and now we’re experimenting with Claude Code, and that took a while to build up to, but as we evaluate Claude Code, we have things to test it against."

    Published Sep 12, 2025Communications platform company; leads platform consolidation and AI tool evaluation for R&D

    Practitioner account, self-described; opinion; not counted, takes no position How GILD checked it

    Origin

    Published practitioner account

    Checked

    Self-described; Opinion; Needs recheck after 12 months

    Role

    Name self-described; employer self-described; role self-described (Jesse Adametz, Senior Engineering Leader on the Developer Platform at Twilio)

    Relationship

    Checked only from the record itself; an outside check is open

    How they know

    Opinion, no use stated

    Support

    No artifact or independent corroboration recorded

    Date and version

    2025-09-12, needs a recheck

    Not counted

    no aspect: excluded, takes no position on any assessed aspect

    Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Licenses alone do not produce the gain

    Published on

    Engineering Blogs, Studies & Reviews

    Who

    Technical leadership

    Evidence

    Reported

    Applies to

    Still applicable

    Observed on

    Long-running Copilot use, now the baseline for evaluating Claude Code

34. Positive

    Marco Argenti, Chief Information Officer, Goldman Sachs

    American Banker: Goldman Sachs staff now write a million gen AI prompts a monthNews report quoting the CIO

    All Goldman Sachs developers have GitHub Copilot, and teams are getting Devin as a colleague to delegate work to.

    "All the developers have Github Copilot. They are getting the Devin technology. There's a shift from the developer assistant that helps you auto complete your code or generate code for a specific task, so I now have a virtual colleague that I can delegate work to."

    Published Jul 25, 2025self-described: Goldman Sachs, 12,000 developers

    Counts toward

    • Overall signal
    • Chosen as the approved default

    Published on

    Engineering Blogs, Studies & Reviews

    Who

    Organization-wide

    Evidence

    Reported

    Applies to

    Current

35. No opinion stated

    Hao Li, Haoxiang Zhang, and Ahmed E. Hassan, Queen's University

    The Rise of AI Teammates in Software Engineering (SE) 3.0: How Autonomous Coding Agents Are Reshaping Software EngineeringResearch paper, arXiv (Queen's University)

    Studying agentic pull requests in the AIDev dataset (456,535 PRs), the authors report OpenAI Codex with the highest acceptance rate at 64%, then Devin at 49% and GitHub Copilot at 35%. Accepted Copilot PRs took 17.2 hours to review, over 4 times the 3.9 hours for accepted human PRs.

    "In contrast, GitHub Copilot shows the opposite pattern: slow to accept, fast to reject. Accepted GitHub Copilot PRs take 17.2 hours to review, which is over 4 times longer than accepted Human-PRs (3.9 hours)…"

    Published Jul 20, 2025self-described: AIDev dataset: 456,535 agent pull requests across 61,453 GitHub repositories from Codex, Devin, Copilot, Cursor, and Claude Code

    Practitioner account; structured test; counts toward Delivered outcomes How GILD checked it

    Origin

    Published practitioner account

    Checked

    Ran a structured test; Artifact checked (benchmark output); Needs recheck after 12 months

    Role

    Name unknown; employer unknown; role unknown

    Relationship

    Checked only from the record itself; an outside check is open

    How they know

    Ran a structured test

    Support

    1 artifact (benchmark output)

    Date and version

    2025-07-20, needs a recheck

    Counts toward

    Delivered outcomes, measured outcomes: mixed, G1 (1 unit before bounds), capped: C-relationship, C-unknown-control

    Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward, pending a recheck

    • Agent Execution: Mixed or negative
    • Requires Strong Review: Raised as a concern
    • Quality Against Alternatives: Raised as a concern

    Published on

    Engineering Blogs, Studies & Reviews

    Who

    Professional team

    Evidence

    Measured

    Applies to

    Needs a recheck

    Observed on

    Copilot coding agent pull requests from January 2025

36. Positive

    David Griffiths, Chief Technology Officer, Citi

    American Banker: Citi is rolling out agentic AI to its 40,000 developersNews report quoting the CTO

    Citi's CTO estimates regular Copilot users get 5% to 15% more productivity and calls Copilot a good assistive tool, while Devin takes simple, repeatable tasks.

    ""We still very much believe in GitHub Copilot," Citi's Chief Technology Officer David Griffiths told American Banker. "Copilot is a good assistive tool for developers. It can help them complete tasks faster, because it effectively auto-completes parts of those tasks for them."…"

    Published Jul 14, 2025self-described: Citi, 40,000 developers

    Counts toward

    • Overall signal
    • Inline Completions: Positive

    Published on

    Engineering Blogs, Studies & Reviews

    Who

    Organization-wide

    Evidence

    Reported

    Applies to

    Current

37. Mixed

    kookamamie, Hacker News commenter

    Comment on: GitHub Copilot Coding AgentHacker News comment

    Their organization on GitHub Business could not work out how to enable the coding agent, and they want tighter control than write access over who can assign it work.

    Published May 20, 2025self-described: Organization on GitHub Business with Copilot

    Practitioner account, pseudonymous; direct use; counts toward Adoption and upkeep How GILD checked it

    Origin

    Published practitioner account

    Checked

    Pseudonymous; No tie to GitHub Copilot found in GILD's check; Used it directly; Needs recheck after 12 months; Specific details

    Role

    Pseudonymous account; identity not pursued

    Relationship

    No tie to GitHub Copilot found in GILD's check (byline, content disclosure, platform api profile); that is not proof none exists

    How they know

    Used it directly

    Support

    No artifact or independent corroboration recorded

    Date and version

    2025-05-20, GH Business, needs a recheck

    Counts toward

    Adoption and upkeep, setup: unfavorable, G2 (2 units before bounds)

    Not counted

    Cost predictability and limits, cost transparency: repeats a stronger account in its group

    Not counted

    Operational control, admin control: repeats a stronger account in its group

    Checked Oct 29, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward, pending a recheck

    • Enterprise Administration: Mixed or negative
    • Data, IP & Security: Discussed without concern

    Published on

    Developer Communities

    Who

    Professional team

    Evidence

    Reported

    Applies to

    Needs a recheck

    Observed on

    Copilot coding agent launch, May 2025

38. Mixed

    Ostap Elyashevskyy, Test Automation Competence Manager at ELEKS

    AI Code Review in Practice: Expert Evaluating GitHub CopilotCompany engineering blog post, ELEKS

    ELEKS found Copilot code review good at naming, unused imports and common best practices, but it missed a null check and stays at surface-level feedback that cannot replace human review.

    "GitHub Copilot (as of now) tends to focus more on surface-level suggestions, like syntax, naming, formatting, and common best practices, rather than deep code review feedback. It cannot substitute human code review."

    Published Apr 15, 2025ELEKS, software engineering company; Copilot code review tested across several client projects

    Practitioner account, name and role on their organization's own page; direct use; counts toward Delivered outcomes How GILD checked it

    Origin

    Published practitioner account

    Checked

    Name and role on their organization's own page; GitHub Copilot partner: not counted as independent; Used it directly; Needs recheck after 12 months; Specific details

    Role

    Name on a page the person does not control; employer on a page the person does not control; role on a page the person does not control (ELEKS' Test Automation Competence Manager)

    Relationship

    partner (confirmed); consultant or reseller (confirmed)

    How they know

    Used it directly

    Support

    No artifact or independent corroboration recorded

    Date and version

    2025-04-15, needs a recheck

    Counts toward

    Delivered outcomes, correctness: mixed, G2 (2 units before bounds)

    Not counted

    Delivered outcomes, task completion: repeats a stronger account in its group

    Not counted

    Delivered outcomes, context handling: repeats a stronger account in its group

    Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Overall signal
    • Code Quality: Mixed or negative
    • Requires Strong Review: Raised as a concern
    • Largest gains on routine code and tests

    Published on

    Engineering Blogs, Studies & Reviews

    Who

    Professional team

    Evidence

    Reported

    Applies to

    Still applicable

    Observed on

    Copilot code review on GitHub pull requests, early 2025

39. Positive

    Birlla Packiasamy, Thoughtworks

    Our experiment with GitHub Copilot: A practical guide for development teamsEngineering blog post, Thoughtworks

    Over a 10-week experiment a four-developer team went from 20 to 23 stories per sprint, after a slow start and a drift away from test-driven development they had to correct.

    Published Apr 3, 2025self-described: Thoughtworks delivery team of four developers, 10-week experiment

    Practitioner account, self-described; structured test; counts toward Delivered outcomes, Adoption and upkeep How GILD checked it

    Origin

    Published practitioner account

    Checked

    Self-described; Customer of GitHub Copilot; Ran a structured test; Needs recheck after 12 months; Specific details

    Role

    Name self-described; employer self-described; role unknown

    Relationship

    customer (self disclosed)

    How they know

    Ran a structured test

    Support

    No artifact or independent corroboration recorded

    Date and version

    2025-04-03, needs a recheck

    Counts toward

    Delivered outcomes, task completion: favorable, G2 (2 units before bounds), capped: C-relationship

    Counts toward

    Adoption and upkeep, setup: mixed, G1 (1 unit before bounds), capped: C-relationship

    Not counted

    Delivered outcomes, measured outcomes: repeats a stronger account in its group

    Not counted

    Delivered outcomes, context handling: repeats a stronger account in its group

    Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Overall signal
    • Inline Completions: Positive
    • Requires Strong Review: Discussed without concern
    • Measuring Impact: Discussed without concern
    • Licenses alone do not produce the gain
    • Written instructions improve results
    • Once review and rework are counted, does Copilot reduce engineering time?: Net time saved

    Published on

    Engineering Blogs, Studies & Reviews

    Who

    Very small team

    Evidence

    Measured

    Applies to

    Still applicable

    Observed on

    Copilot inline and chat assistance

40. Positive

    Ali Dasdan, with Gal Bakal, Guy Levin, Michael Kaufman, and Yaniv Katz, ZoomInfo Engineering

    Experience with GitHub Copilot for Developer Productivity at ZoominfoEngineering blog post, ZoomInfo Engineering Blog

    ZoomInfo rolled Copilot out to over 400 developers in four phases and measured a 33% suggestion acceptance rate, 20% of lines accepted, 72% satisfaction, and about 20% time saved.

    Published Feb 13, 2025self-described: ZoomInfo engineering organization, over 400 developers

    Practitioner account, self-described; structured test; counts toward Delivered outcomes, Adoption and upkeep How GILD checked it

    Origin

    Published practitioner account

    Checked

    Self-described; Customer of GitHub Copilot; Ran a structured test; Needs recheck after 12 months; Artifact checked (dataset)

    Role

    Name self-described; employer self-described; role unknown

    Relationship

    customer (self disclosed)

    How they know

    Ran a structured test

    Support

    1 artifact (dataset)

    Date and version

    2025-02-13, GitHub Copilot for Business, needs a recheck

    Counts toward

    Delivered outcomes, measured outcomes: favorable, G1 (1 unit before bounds), capped: C-relationship, C-unknown-control

    Counts toward

    Delivered outcomes, task completion: favorable, G1 (1 unit before bounds), capped: C-relationship, C-unknown-control

    Counts toward

    Delivered outcomes, correctness: mixed, G1 (1 unit before bounds), capped: C-relationship, C-unknown-control

    Counts toward

    Delivered outcomes, context handling: unfavorable, G1 (1 unit before bounds), capped: C-relationship, C-unknown-control

    Counts toward

    Adoption and upkeep, setup: favorable, G1 (1 unit before bounds), capped: C-relationship, C-unknown-control

    Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Overall signal
    • Inline Completions: Positive
    • Enterprise Administration: Positive
    • Code Quality: Mixed or negative
    • Requires Strong Review: Raised as a concern
    • Quality Against Alternatives: Discussed without concern
    • Measuring Impact: Discussed without concern
    • Largest gains on routine code and tests
    • Is Copilot the main AI coding tool, or the one engineers keep beside another?: Main tool
    • Once review and rework are counted, does Copilot reduce engineering time?: Net time saved

    Published on

    Engineering Blogs, Studies & Reviews

    Who

    Organization-wide

    Evidence

    Measured

    Applies to

    Still applicable

    Observed on

    GitHub Copilot code completion, rollout from September 2023

41. Positive

    David Betts, leader of the developer platform team at Twilio

    Leveraging sentiment data, driving org-wide action, and executive engagementPodcast episode page with transcript, Engineering Enablement by DX

    Betts cites Twilio engineers self-reporting that GitHub Copilot saves them two to four hours a week, which Betts frames as four full-time engineers of time saved across the organization, and says high-frequency Copilot users correlate with high sentiment on a variety of other drivers.

    "…people are self-reporting that they’re saving two to four hours by using GitHub Copilot per week, and that equates to four full-time engineers of time saved across our organization."

    Published Jan 31, 2025Communications platform company; platform team runs developer sentiment surveys across engineering

    not counted until substantiated How GILD checked it

    Role

    Name self-described; employer self-described; role self-described (David Betts, leader of Twilio's developer platform team)

    Relationship

    customer (self disclosed)

    How they know

    Reports others' experience

    Support

    No artifact or independent corroboration recorded

    Date and version

    2025-01-31, needs a recheck

    Not counted

    Delivered outcomes, measured outcomes: a lead (G0), counts toward nothing until substantiated

    Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Measuring Impact: Discussed without concern
    • Once review and rework are counted, does Copilot reduce engineering time?: Net time saved

    Published on

    Engineering Blogs, Studies & Reviews

    Who

    Technical leadership

    Evidence

    Measured

    Applies to

    Still applicable

    Observed on

    GitHub Copilot as Twilio's coding assistant, tracked through surveys and system data

42. Positive

    Jesse Houwing, GitHub and Scrum trainer

    Fixing complex bugs with AI - A deepdive into the PowerShellScriptAnalyzerPersonal blog post, jessehouwing.net

    Houwing ran eleven Copilot agent tasks on a fork while chasing a PowerShell analyzer build failure; eight were merged and three closed as superseded. Copilot was very good at bounded, well-specified work, but did not say when to stop and installed .NET 8 into a temporary folder at least five times.

    "I ran eleven separate agent tasks on my fork; eight were merged and three were closed as superseded."

    Published Sep 17, 2026self-described: Side-project contribution to the PSScriptAnalyzer open source project

    Practitioner account, self-described; direct use; counts toward Delivered outcomes, Reliability How GILD checked it

    Origin

    Published practitioner account

    Checked

    Self-described; GitHub Copilot partner: not counted as independent; Used it directly; Specific details

    Role

    Name self-described; employer unknown; role unknown

    Relationship

    program member (suspected)

    How they know

    Used it directly

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-09-17

    Counts toward

    Delivered outcomes, correctness: mixed, G1 (1 unit before bounds)

    Counts toward

    Reliability, run stability: unfavorable, G2 (2 units before bounds)

    Not counted

    Delivered outcomes, task completion: excluded, praise from someone tied to the vendor, kept as an attributed position

    Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

    Not counted

    Its author is a Microsoft MVP, a program run by GitHub's parent Microsoft. Kept in the record; it takes part in no figure.

    Published on

    Engineering Blogs, Studies & Reviews

    Who

    Individual practitioner

    Evidence

    Measured

    Applies to

    Current

    Observed on

    Copilot agent tasks on a fork, with Copilot in VS Code

43. Positive

    DHH, chief technology officer, 37signals

    Lex Fridman Podcast #501: DHH on the future of programming and agentic engineeringPodcast episode with publisher transcript

    DHH reports that Copilot's pull request review has become good and keeps finding things that are legitimately broken after his Claude Code and Codex agents push to GitHub.

    "And then Copilot, I kid you not, has actually gotten good. Copilot keeps finding stuff that’s legitimately broken,…"

    Published Aug 26, 2026self-described: 37signals and Omarchy, his own agent work

    Counts toward

    • Overall signal
    • GitHub Integration: Positive
    • Code Quality: Positive

    Published on

    Podcasts, Talks & Videos

    Who

    Individual practitioner

    Evidence

    Reported

    Applies to

    Current

44. Negative

    James Faris, reporter, Business Insider, reporting a Disney internal message

    Business Insider: Disney is ditching Microsoft's GitHub Copilot and adding OpenAI's CodexNews report

    Disney is dropping GitHub Copilot for US staff in August 2026, and eight Disney tech employees said they rarely or never used it.

    "Eight Disney tech employees told Business Insider that they rarely or never used GitHub Copilot."

    Published Jul 29, 2026self-described: Disney's US technology organization

    Counts toward

    • Overall signal
    • Quality Against Alternatives: Raised as a concern
    • Is Copilot the main AI coding tool, or the one engineers keep beside another?: Moved to another tool

    Published on

    Engineering Blogs, Studies & Reviews

    Who

    Individual practitioner

    Evidence

    Reported

    Applies to

    Current

45. Positive

    lanthissa, Hacker News commenter

    Comment on: Kimi K2.7 Code is generally available in GitHub CopilotHacker News comment

    Falling back to Copilot at work after running out of Claude credits, found choosing subagent models across vendors more useful than Claude's tooling and cheaper.

    Published Jul 2, 2026self-described: Developer at work with a dollar budget for AI use

    Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account How GILD checked it

    Origin

    Published practitioner account

    Checked

    Pseudonymous; No tie to GitHub Copilot found in GILD's check; Used it directly

    Role

    Pseudonymous account; identity not pursued

    Relationship

    No tie to GitHub Copilot found in GILD's check (byline, content disclosure, platform api profile); that is not proof none exists

    How they know

    Used it directly

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-07-02

    Not counted

    Operational control, models and deployment: repeats a stronger account in its group

    Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Overall signal
    • Agent Execution: Positive
    • Usage / Cost: Discussed without concern
    • Does Copilot's usage-based billing work for the people paying for it?: Workable

    Published on

    Developer Communities

    Who

    Individual practitioner

    Evidence

    Reported

    Applies to

    Current

46. Negative

    KarimWajihKMW, GitHub Community member on Copilot Pro+

    GitHub Copilot AI Credits Are Unfair, Expensive, and Killing Real Development WorkflowsGitHub Community discussion, Copilot Conversations

    After the June 2026 move to AI credits, used all of their Copilot Pro+ allowance in two days of normal work. They had no way to see what a task would cost before running it.

    "I used 100% of copilot pro+ in just 2 days of normal work which I used to do months and years ago."

    Published Jun 4, 2026self-described: Copilot Pro+ subscriber using agent mode and code review

    Practitioner account, pseudonymous; direct use; counts toward Cost predictability and limits How GILD checked it

    Origin

    Published practitioner account

    Checked

    Pseudonymous; No tie to GitHub Copilot found in GILD's check; Used it directly; Observed on copilot pro+; Specific details

    Role

    Pseudonymous account; identity not pursued

    Relationship

    No tie to GitHub Copilot found in GILD's check (byline, content disclosure, platform api profile); that is not proof none exists

    How they know

    Used it directly

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-06-04, copilot pro+

    Counts toward

    Cost predictability and limits, cost predictability: unfavorable, G1 (1 unit before bounds), capped: C-relevance

    Counts toward

    Cost predictability and limits, limits in practice: unfavorable, G1 (1 unit before bounds), capped: C-relevance

    Counts toward

    Cost predictability and limits, cost transparency: unfavorable, G1 (1 unit before bounds), capped: C-relevance

    Checked Oct 30, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Overall signal
    • Value for Money: Mixed or negative
    • Usage / Cost: Raised as a concern
    • Does Copilot's usage-based billing work for the people paying for it?: Unpredictable

    Published on

    Developer Communities

    Who

    Individual practitioner

    Evidence

    Reported

    Applies to

    Current

    Observed on

    Copilot Pro+ under AI credits billing from June 1, 2026

47. Mixed

    webworker, Hacker News commenter, self-described

    Comment on: GitHub Copilot is moving to usage-based billingHacker News comment

    Used Copilot to write issues and hand them to agents, which worked well for contained work, but will let it lapse after the billing change and move to Codex and JetBrains AI.

    Published Apr 28, 2026

    Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account How GILD checked it

    Origin

    Published practitioner account

    Checked

    Pseudonymous; No tie to GitHub Copilot found in GILD's check; Used it directly

    Role

    Pseudonymous account; identity not pursued

    Relationship

    No tie to GitHub Copilot found in GILD's check (byline, content disclosure, platform api profile); that is not proof none exists

    How they know

    Used it directly

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-04-28

    Not counted

    Delivered outcomes, task completion: repeats a stronger account in its group

    Checked Oct 30, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Overall signal
    • Agent Execution: Positive
    • GitHub Integration: Positive
    • Needs Close Steering: Discussed without concern
    • Is Copilot the main AI coding tool, or the one engineers keep beside another?: Moved to another tool
    • Does Copilot's usage-based billing work for the people paying for it?: Unpredictable

    Published on

    Developer Communities

    Who

    Individual practitioner

    Evidence

    Reported

    Applies to

    Current

48. Mixed

    KronisLV, Hacker News commenter

    Comment on: GitHub Copilot is moving to usage-based billingHacker News comment

    Uses Claude Code as their main agent but kept Copilot for cheap access to other models and for VS Code autocomplete that tabs through edits across files.

    Published Apr 27, 2026self-described: Uses Claude Code as the main agent

    Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account How GILD checked it

    Origin

    Published practitioner account

    Checked

    Pseudonymous; No tie to GitHub Copilot found in GILD's check; Used it directly

    Role

    Pseudonymous account; identity not pursued

    Relationship

    No tie to GitHub Copilot found in GILD's check (byline, content disclosure, platform api profile); that is not proof none exists

    How they know

    Used it directly

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-04-27

    Not counted

    Delivered outcomes, context handling: repeats a stronger account in its group

    Not counted

    Operational control, models and deployment: repeats a stronger account in its group

    Checked Oct 30, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Overall signal
    • Inline Completions: Positive
    • Value for Money: Positive
    • Usage / Cost: Discussed without concern
    • Is Copilot the main AI coding tool, or the one engineers keep beside another?: Moved to another tool

    Published on

    Developer Communities

    Who

    Individual practitioner

    Evidence

    Reported

    Applies to

    Current

49. Mixed

    0xffff2, Hacker News commenter

    Comment on: GitHub Copilot is moving to usage-based billingHacker News comment

    Prefers Copilot because in VS Code it sees what the developer sees, including linter output, but says the pricing change leaves no reason to stay beyond that integration.

    Published Apr 27, 2026self-described: Works mainly in VS Code

    Practitioner account, pseudonymous; direct use; not counted, takes no position How GILD checked it

    Origin

    Published practitioner account

    Checked

    Pseudonymous; No tie to GitHub Copilot found in GILD's check; Used it directly

    Role

    Pseudonymous account; identity not pursued

    Relationship

    No tie to GitHub Copilot found in GILD's check (byline, content disclosure, platform api profile); that is not proof none exists

    How they know

    Used it directly

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-04-27

    Not counted

    no aspect: excluded, takes no position on any assessed aspect

    Checked Oct 29, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Overall signal
    • Editor Integration: Positive
    • Value for Money: Mixed or negative
    • Does Copilot's usage-based billing work for the people paying for it?: Unpredictable

    Published on

    Developer Communities

    Who

    Individual practitioner

    Evidence

    Reported

    Applies to

    Current

50. Mixed

    Merad, Hacker News commenter

    Comment on: GitHub Copilot is moving to usage-based billingHacker News comment

    Used Copilot at work because it was the only approved tool and rates it about 80% as capable as Claude Code; expects their company to drop it after the billing change.

    Published Apr 27, 2026self-described: Corporate developer, Copilot was the only approved AI tool

    Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account How GILD checked it

    Origin

    Published practitioner account

    Checked

    Pseudonymous; No tie to GitHub Copilot found in GILD's check; Used it directly

    Role

    Pseudonymous account; identity not pursued

    Relationship

    No tie to GitHub Copilot found in GILD's check (byline, content disclosure, platform api profile); that is not proof none exists

    How they know

    Used it directly

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-04-27

    Not counted

    Delivered outcomes, task completion: repeats a stronger account in its group

    Checked Oct 30, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Overall signal
    • Quality Against Alternatives: Raised as a concern
    • Chosen as the approved default
    • Is Copilot the main AI coding tool, or the one engineers keep beside another?: Moved to another tool
    • Does Copilot's usage-based billing work for the people paying for it?: Unpredictable

    Published on

    Developer Communities

    Who

    Individual practitioner

    Evidence

    Reported

    Applies to

    Current

51. Positive

    zmj, Hacker News commenter

    Comment on: GitHub Copilot is moving to usage-based billingHacker News comment

    Using Copilot CLI heavily at work, finds it keeps pace with Claude Code features; Claude models need about 20% more turns there, without a visible difference in output.

    Published Apr 27, 2026self-described: Heavy use at work since January 2026, personal Anthropic subscription for comparison

    Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account How GILD checked it

    Origin

    Published practitioner account

    Checked

    Pseudonymous; No tie to GitHub Copilot found in GILD's check; Used it directly

    Role

    Pseudonymous account; identity not pursued

    Relationship

    customer (self disclosed)

    How they know

    Used it directly

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-04-27

    Not counted

    Delivered outcomes, task completion: repeats a stronger account in its group

    Not counted

    Adoption and upkeep, continuity: repeats a stronger account in its group

    Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Overall signal
    • Agent Execution: Positive
    • Code Quality: Positive
    • Feature Lag: Discussed without concern
    • Quality Against Alternatives: Discussed without concern
    • Is Copilot the main AI coding tool, or the one engineers keep beside another?: Main tool

    Published on

    Developer Communities

    Who

    Individual practitioner

    Evidence

    Reported

    Applies to

    Current

    Observed on

    Copilot CLI with Claude models

52. Mixed

    cityofdelusion, Hacker News commenter, self-described

    Comment on: GitHub Copilot is moving to usage-based billingHacker News comment

    Finds the Copilot IDE plugin's agents glitchy, with mangled shells and hung terminals, though the IDE integration makes reviewing changes quick.

    Published Apr 27, 2026

    Practitioner account, pseudonymous; direct use; counts toward Reliability How GILD checked it

    Origin

    Published practitioner account

    Checked

    Pseudonymous; No tie to GitHub Copilot found in GILD's check; Used it directly

    Role

    Pseudonymous account; identity not pursued

    Relationship

    No tie to GitHub Copilot found in GILD's check (byline, content disclosure, platform api profile); that is not proof none exists

    How they know

    Used it directly

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-04-27

    Counts toward

    Reliability, run stability: unfavorable, G1 (1 unit before bounds)

    Checked Oct 29, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Overall signal
    • Agent Execution: Mixed or negative
    • Editor Integration: Positive
    • Quality Against Alternatives: Raised as a concern

    Published on

    Developer Communities

    Who

    Individual practitioner

    Evidence

    Reported

    Applies to

    Current

    Observed on

    Copilot IDE plugin agents

53. Negative

    to11mtm, Hacker News commenter

    Comment on: Changes to GitHub Copilot individual plansHacker News comment

    Copilot is the only real option at their workplace. After the April 2026 changes, their personal plan lost models the enterprise account still had, at higher multipliers.

    Published Apr 22, 2026self-described: Uses the Copilot plugin at work, where it is the only real option, and at home

    Practitioner account, pseudonymous; direct use; counts toward Cost predictability and limits How GILD checked it

    Origin

    Published practitioner account

    Checked

    Pseudonymous; No tie to GitHub Copilot found in GILD's check; Used it directly; Specific details; Observed on Opus 4.7

    Role

    Pseudonymous account; identity not pursued

    Relationship

    No tie to GitHub Copilot found in GILD's check (byline, content disclosure, platform api profile); that is not proof none exists

    How they know

    Used it directly

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-04-22, Opus 4.7

    Counts toward

    Cost predictability and limits, cost transparency: unfavorable, G2 (2 units before bounds)

    Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward, pending a recheck

    • Overall signal
    • Usage / Cost: Raised as a concern
    • Chosen as the approved default
    • Does Copilot's usage-based billing work for the people paying for it?: Unpredictable

    Published on

    Developer Communities

    Who

    Individual practitioner

    Evidence

    Reported

    Applies to

    Needs a recheck

    Observed on

    Pro+ and Enterprise model multipliers, April 2026

54. Negative

    chillfox, Hacker News commenter

    Comment on: Changes to GitHub Copilot individual plansHacker News comment

    Trying Claude through Copilot after using the API directly, saw failed file edits and wasted turns on work the API gets right in one shot.

    Published Apr 22, 2026self-described: Weekend trial after using the Anthropic API directly

    Practitioner account, pseudonymous; direct use; counts toward Reliability How GILD checked it

    Origin

    Published practitioner account

    Checked

    Pseudonymous; No tie to GitHub Copilot found in GILD's check; Used it directly; Specific details

    Role

    Pseudonymous account; identity not pursued

    Relationship

    No tie to GitHub Copilot found in GILD's check (byline, content disclosure, platform api profile); that is not proof none exists

    How they know

    Used it directly

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-04-22

    Counts toward

    Reliability, run stability: unfavorable, G1 (1 unit before bounds), capped: C-relevance

    Not counted

    Delivered outcomes, task completion: repeats a stronger account in its group

    Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Overall signal
    • Agent Execution: Mixed or negative
    • Needs Close Steering: Raised as a concern
    • Quality Against Alternatives: Raised as a concern

    Published on

    Developer Communities

    Who

    Individual practitioner

    Evidence

    Reported

    Applies to

    Current

    Observed on

    Claude models through Copilot

55. Mixed

    shmolyneaux, Hacker News commenter

    Comment on: Changes to GitHub Copilot individual plansHacker News comment

    Assigned detailed issues to the Copilot agent on Opus for 30-minute runs at three premium requests each, called that unsustainable, and canceled when Opus left Pro.

    Published Apr 22, 2026self-described: Solo, Copilot Pro

    Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes, Cost predictability and limits, Adoption and upkeep How GILD checked it

    Origin

    Published practitioner account

    Checked

    Pseudonymous; Customer of GitHub Copilot; Used it directly; Specific details; Observed on Opus 4.5

    Role

    Pseudonymous account; identity not pursued

    Relationship

    customer (self disclosed)

    How they know

    Used it directly

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-04-22, Opus 4.5

    Counts toward

    Delivered outcomes, task completion: favorable, G2 (2 units before bounds)

    Counts toward

    Cost predictability and limits, cost transparency: mixed, G2 (2 units before bounds)

    Counts toward

    Adoption and upkeep, continuity: unfavorable, G2 (2 units before bounds)

    Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward, pending a recheck

    • Overall signal
    • Agent Execution: Positive
    • GitHub Integration: Positive
    • Usage / Cost: Discussed without concern
    • Needs Close Steering: Discussed without concern
    • Is Copilot the main AI coding tool, or the one engineers keep beside another?: Moved to another tool
    • Does Copilot's usage-based billing work for the people paying for it?: Workable

    Published on

    Developer Communities

    Who

    Individual practitioner

    Evidence

    Reported

    Applies to

    Needs a recheck

    Observed on

    Cloud agent with Opus 4.5 and 4.6 on 300 premium requests a month

56. Positive

    WorldMaker, Hacker News commenter

    Comment on: Changes to GitHub Copilot individual plansHacker News comment

    Values Copilot's integration with GitHub, where agent work starts from a phone, and with VS Code on Windows, plus GitHub picking sensible default models.

    Published Apr 22, 2026self-described: Solo development on Windows with VS Code

    Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account How GILD checked it

    Origin

    Published practitioner account

    Checked

    Pseudonymous; No tie to GitHub Copilot found in GILD's check; Used it directly

    Role

    Pseudonymous account; identity not pursued

    Relationship

    No tie to GitHub Copilot found in GILD's check (byline, content disclosure, platform api profile); that is not proof none exists

    How they know

    Used it directly

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-04-22

    Not counted

    Adoption and upkeep, setup: repeats a stronger account in its group

    Not counted

    Operational control, models and deployment: repeats a stronger account in its group

    Checked Oct 30, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Overall signal
    • Editor Integration: Positive
    • GitHub Integration: Positive
    • Is Copilot the main AI coding tool, or the one engineers keep beside another?: Main tool

    Published on

    Developer Communities

    Who

    Individual practitioner

    Evidence

    Reported

    Applies to

    Current

57. Mixed

    benwills, Hacker News commenter

    Comment on: Changes to GitHub Copilot individual plansHacker News comment

    Settled on Copilot after four weeks testing the major tools, then saw Opus 4.6 removed and Opus 4.7 priced at 7.5 requests per prompt with no refund path.

    Published Apr 21, 2026self-described: Solo, prepaid annual plan

    Practitioner account, pseudonymous; direct use; counts toward Cost predictability and limits, Adoption and upkeep How GILD checked it

    Origin

    Published practitioner account

    Checked

    Pseudonymous; No tie to GitHub Copilot found in GILD's check; Used it directly; Specific details; Observed on Opus 4.7

    Role

    Pseudonymous account; identity not pursued

    Relationship

    No tie to GitHub Copilot found in GILD's check (byline, content disclosure, platform api profile); that is not proof none exists

    How they know

    Used it directly

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-04-21, Opus 4.7

    Counts toward

    Adoption and upkeep, continuity: unfavorable, G2 (2 units before bounds)

    Counts toward

    Cost predictability and limits, cost transparency: unfavorable, G2 (2 units before bounds)

    Not counted

    Cost predictability and limits, cost predictability: repeats a stronger account in its group

    Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward, pending a recheck

    • Overall signal
    • Usage / Cost: Raised as a concern
    • Is Copilot the main AI coding tool, or the one engineers keep beside another?: Main tool
    • Does Copilot's usage-based billing work for the people paying for it?: Unpredictable

    Published on

    Developer Communities

    Who

    Individual practitioner

    Evidence

    Reported

    Applies to

    Needs a recheck

    Observed on

    Premium request multipliers for Opus 4.6 and 4.7

58. Mixed

    sunaookami, Hacker News commenter

    Comment on: Changes to GitHub Copilot individual plansHacker News comment

    Liked Copilot Pro because monthly allowances made usage easy to plan, unlike weekly limits elsewhere, then lost Opus access without warning in the April 2026 changes.

    Published Apr 21, 2026self-described: Copilot Pro with the Pi agent harness

    Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account How GILD checked it

    Origin

    Published practitioner account

    Checked

    Pseudonymous; Customer of GitHub Copilot; Used it directly; Observed on Copilot Pro

    Role

    Pseudonymous account; identity not pursued

    Relationship

    customer (self disclosed)

    How they know

    Used it directly

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-04-21, Copilot Pro

    Not counted

    Adoption and upkeep, continuity: repeats a stronger account in its group

    Not counted

    Cost predictability and limits, cost predictability: repeats a stronger account in its group

    Not counted

    Cost predictability and limits, limits in practice: repeats a stronger account in its group

    Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward, pending a recheck

    • Overall signal
    • Usage / Cost: Raised as a concern
    • Does Copilot's usage-based billing work for the people paying for it?: Unpredictable

    Published on

    Developer Communities

    Who

    Individual practitioner

    Evidence

    Reported

    Applies to

    Needs a recheck

    Observed on

    Copilot Pro, Sonnet and Opus, April 2026 plan changes

59. Positive

    aurareturn, Hacker News commenter, self-described

    Comment on: Ask HN: Why does it look like everyone is abandoning GitHub Copilot?Hacker News comment

    Keeps Copilot as their main tool because the default VS Code integration shows AI changes to code they know, and rates the US$10 plan good value.

    Published Apr 7, 2026

    Practitioner account, pseudonymous; direct use; counts toward Operational control How GILD checked it

    Origin

    Published practitioner account

    Checked

    Pseudonymous; No tie to GitHub Copilot found in GILD's check; Used it directly; Specific details; Observed on Opus 4.6 and GPT 5.4

    Role

    Pseudonymous account; identity not pursued

    Relationship

    No tie to GitHub Copilot found in GILD's check (byline, content disclosure, platform api profile); that is not proof none exists

    How they know

    Used it directly

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-04-07, Opus 4.6 and GPT 5.4

    Counts toward

    Operational control, models and deployment: favorable, G2 (2 units before bounds)

    Not counted

    Delivered outcomes, context handling: repeats a stronger account in its group

    Not counted

    Cost predictability and limits, limits in practice: repeats a stronger account in its group

    Checked Oct 29, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward, pending a recheck

    • Overall signal
    • Editor Integration: Positive
    • Value for Money: Positive
    • Usage / Cost: Discussed without concern
    • Is Copilot the main AI coding tool, or the one engineers keep beside another?: Main tool
    • Does Copilot's usage-based billing work for the people paying for it?: Workable

    Published on

    Developer Communities

    Who

    Individual practitioner

    Evidence

    Reported

    Applies to

    Needs a recheck

    Observed on

    US$10 Pro plan with Opus 4.6 and GPT 5.4, before AI credits

60. Positive

    fabev, Hacker News commenter

    Comment on: Ask HN: Why does it look like everyone is abandoning GitHub Copilot?Hacker News comment

    Stays on Copilot after a Cursor subscription ran out of its US$20 limit far faster, and finds Copilot's tools better tuned.

    Published Apr 7, 2026self-described: Founder of a hobby roleplay AI platform

    Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account How GILD checked it

    Origin

    Published practitioner account

    Checked

    Pseudonymous; No tie to GitHub Copilot found in GILD's check; Used it directly

    Role

    Pseudonymous account; identity not pursued

    Relationship

    No tie to GitHub Copilot found in GILD's check (byline, content disclosure, platform api profile); that is not proof none exists

    How they know

    Used it directly

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-04-07

    Not counted

    Cost predictability and limits, limits in practice: repeats a stronger account in its group

    Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward, pending a recheck

    • Overall signal
    • Value for Money: Positive
    • Usage / Cost: Discussed without concern
    • Is Copilot the main AI coding tool, or the one engineers keep beside another?: Main tool
    • Does Copilot's usage-based billing work for the people paying for it?: Workable

    Published on

    Developer Communities

    Who

    Individual practitioner

    Evidence

    Reported

    Applies to

    Needs a recheck

    Observed on

    Before AI credits

61. Positive

    gWPVhyxPHqvk, Hacker News commenter

    Comment on: How many products does Microsoft have named 'Copilot'?Hacker News comment

    Per-request billing gave them 10 to 20 times the value of Claude Code, with model choice across vendors and a Copilot CLI near feature parity.

    Published Apr 4, 2026self-described: Uses Copilot and Claude Code at work and at home

    Practitioner account, pseudonymous; direct use; counts toward Reliability, Cost predictability and limits How GILD checked it

    Origin

    Published practitioner account

    Checked

    Pseudonymous; No tie to GitHub Copilot found in GILD's check; Used it directly; Observed on Github Copilot CLI; Specific details

    Role

    Pseudonymous account; identity not pursued

    Relationship

    No tie to GitHub Copilot found in GILD's check (byline, content disclosure, platform api profile); that is not proof none exists

    How they know

    Used it directly

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-04-04, Github Copilot CLI

    Counts toward

    Cost predictability and limits, cost transparency: favorable, G2 (2 units before bounds)

    Counts toward

    Reliability, run stability: favorable, G1 (1 unit before bounds)

    Not counted

    Operational control, models and deployment: repeats a stronger account in its group

    Checked Oct 29, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward, pending a recheck

    • Overall signal
    • Agent Execution: Positive
    • Value for Money: Positive
    • Usage / Cost: Discussed without concern
    • Feature Lag: Discussed without concern
    • Is Copilot the main AI coding tool, or the one engineers keep beside another?: Main tool
    • Does Copilot's usage-based billing work for the people paying for it?: Workable

    Published on

    Developer Communities

    Who

    Individual practitioner

    Evidence

    Reported

    Applies to

    Needs a recheck

    Observed on

    Per-request billing before AI credits; Copilot CLI

62. Positive

    Mike Cann, Convex

    Comparing 9 Code Review Tools - Which is Best?Engineering blog post, Convex Stack (Convex engineering blog)

    Scoring nine review bots on ten seeded pull requests, Mike Cann placed Copilot second, close behind Qodo. He chose it for his own reviews because it is built into GitHub with billing already set up.

    "I will probably just start to use Copilot for my code review needs. It's built into GitHub automatically, so I don't need to mess around with anything else. Billing's already set up."

    Published Mar 10, 2026self-described: Ten seeded pull requests on a React, Vite, and Convex project, scored against nine review bots on default settings

    Practitioner account; structured test; counts toward Delivered outcomes, Adoption and upkeep How GILD checked it

    Origin

    Published practitioner account

    Checked

    Ran a structured test; Specific details

    Role

    Name unknown; employer unknown; role unknown

    Relationship

    Checked only from the record itself; an outside check is open

    How they know

    Ran a structured test

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-03-10

    Counts toward

    Delivered outcomes, measured outcomes: favorable, G1 (1 unit before bounds), capped: C-relationship, C-unknown-control

    Counts toward

    Adoption and upkeep, setup: favorable, G1 (1 unit before bounds), capped: C-relationship, C-unknown-control

    Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Overall signal
    • GitHub Integration: Positive
    • Code Quality: Positive

    Published on

    Engineering Blogs, Studies & Reviews

    Who

    Individual practitioner

    Evidence

    Measured

    Applies to

    Current

    Observed on

    Copilot code review on GitHub pull requests

63. Mixed

    ZainDevX, GitHub Community member

    Copilot Context Window Showing ~40% Reserved Output Even With Minimal PromptGitHub Community discussion, Copilot Conversations

    After Copilot's context window grew to 192K tokens, an empty chat that only said hi already showed about 40% of the window in use, mostly reserved for output.

    "Even when I open a completely empty chat and send a very simple message like: hi the Context Window indicator already shows around 40% usage, and most of that appears to be labeled as Reserved Output."

    Published Mar 5, 2026self-described: Copilot Chat user

    Practitioner account, pseudonymous; direct use; not counted, takes no position How GILD checked it

    Origin

    Published practitioner account

    Checked

    Pseudonymous; No tie to GitHub Copilot found in GILD's check; Used it directly

    Role

    Pseudonymous account; identity not pursued

    Relationship

    No tie to GitHub Copilot found in GILD's check (byline, content disclosure, platform api profile); that is not proof none exists

    How they know

    Used it directly

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-03-05

    Not counted

    no aspect: excluded, takes no position on any assessed aspect

    Checked Oct 29, 2026 under gild-policy/1.0.0. The scorecard

    Published on

    Developer Communities

    Who

    Individual practitioner

    Evidence

    Measured

    Applies to

    Current

    Observed on

    Copilot Chat after the context window grew from 128K to 192K tokens

64. Positive

    Terence Tham, developer and paying user of all three tools

    38 Issues: Code Review Agent Showdown between BugBot, Copilot and ClaudeIndependent review, DEV Community

    On a .NET 10 pull request seeded with 38 issues, Copilot code review caught 34/38 on the first pass, ahead of Claude (32/38) and BugBot (29/38), and reached 38/38 over three passes, as Claude also did. Copilot never gives an overall ready-to-merge verdict.

    Published Feb 24, 2026self-described: Controlled .NET 10 test repository with 38 planted issues

    Practitioner account, self-described; structured test; counts toward Delivered outcomes How GILD checked it

    Origin

    Published practitioner account

    Checked

    Self-described; Customer of GitHub Copilot; Ran a structured test; Artifact checked (benchmark output)

    Role

    Name self-described; employer unknown; role unknown

    Relationship

    customer (self disclosed); customer (self disclosed); customer (self disclosed)

    How they know

    Ran a structured test

    Support

    1 artifact (benchmark output)

    Date and version

    2026-02-24

    Counts toward

    Delivered outcomes, measured outcomes: favorable, G3 (3 units before bounds)

    Not counted

    Delivered outcomes, task completion: repeats a stronger account in its group

    Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Overall signal
    • GitHub Integration: Positive
    • Code Quality: Positive
    • Requires Strong Review: Discussed without concern

    Published on

    Engineering Blogs, Studies & Reviews

    Who

    Individual practitioner

    Evidence

    Measured

    Applies to

    Current

    Observed on

    Copilot code review on GitHub pull requests

65. Mixed

    Maxim Saplin, developer and heavy Cursor user

    Ran out of Cursor tokens and switched to GitHub Copilot: Side-by-SidePersonal blog post, DEV Community

    Moving from Cursor to Copilot after hitting Cursor limits, found plan mode and chat weaker and the experience irritating, though still productive; a later update says most issues were fixed.

    Published Feb 18, 2026self-described: Solo use after a Cursor token limit ran out

    Practitioner account, self-described; direct use; counts toward Delivered outcomes, Reliability, Operational control, Cost predictability and limits, Adoption and upkeep How GILD checked it

    Origin

    Published practitioner account

    Checked

    Self-described; No tie to GitHub Copilot found in GILD's check; Used it directly; Observed on Insider Preview version; Specific details

    Role

    Name self-described; employer unknown; role unknown

    Relationship

    No tie to GitHub Copilot found in GILD's check (byline, content disclosure, author profile); that is not proof none exists

    How they know

    Used it directly

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-02-18, Insider Preview version

    Counts toward

    Adoption and upkeep, continuity: favorable, G1 (1 unit before bounds)

    Counts toward

    Cost predictability and limits, cost transparency: unfavorable, G1 (1 unit before bounds)

    Counts toward

    Cost predictability and limits, limits in practice: favorable, G1 (1 unit before bounds)

    Counts toward

    Reliability, run stability: unfavorable, G2 (2 units before bounds)

    Counts toward

    Delivered outcomes, context handling: unfavorable, G2 (2 units before bounds)

    Counts toward

    Delivered outcomes, task completion: unfavorable, G2 (2 units before bounds)

    Counts toward

    Operational control, models and deployment: unfavorable, G2 (2 units before bounds)

    Counts toward

    Adoption and upkeep, setup: favorable, G1 (1 unit before bounds)

    Counts toward

    Adoption and upkeep, ongoing upkeep: unfavorable, G1 (1 unit before bounds)

    Not counted

    Reliability, service availability: repeats a stronger account in its group

    Not counted

    Delivered outcomes, correctness: repeats a stronger account in its group

    Not counted

    Operational control, execution control: repeats a stronger account in its group

    Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward, pending a recheck

    • Overall signal
    • Agent Execution: Mixed or negative
    • Needs Close Steering: Raised as a concern
    • Feature Lag: Raised as a concern
    • Quality Against Alternatives: Raised as a concern
    • Is Copilot the main AI coding tool, or the one engineers keep beside another?: Moved to another tool

    Published on

    Engineering Blogs, Studies & Reviews

    Who

    Individual practitioner

    Evidence

    Reported

    Applies to

    Needs a recheck

    Observed on

    VS Code Insiders with Copilot agent and plan modes

66. Mixed

    Domenico Tenace, developer

    Google Antigravity vs GitHub Copilot: Why I'm Team Antigravity NowPersonal blog post, DEV Community (Playful Programming)

    Having used both extensively, rates Copilot a fast, capable assistant for completions but moved their main work to Antigravity for agent-driven tasks.

    Published Jan 23, 2026self-described: Solo, open source utility library refactoring

    Practitioner account, self-described; direct use; counts toward Delivered outcomes, Operational control, Adoption and upkeep How GILD checked it

    Origin

    Published practitioner account

    Checked

    Self-described; No tie to GitHub Copilot found in GILD's check; Used it directly

    Role

    Name self-described; employer unknown; role unknown

    Relationship

    No tie to GitHub Copilot found in GILD's check (byline, content disclosure, author profile); that is not proof none exists

    How they know

    Used it directly

    Support

    No artifact or independent corroboration recorded

    Date and version

    2026-01-23

    Counts toward

    Delivered outcomes, task completion: favorable, G1 (1 unit before bounds)

    Counts toward

    Delivered outcomes, context handling: unfavorable, G1 (1 unit before bounds), capped: C-relevance

    Counts toward

    Operational control, execution control: favorable, G1 (1 unit before bounds)

    Counts toward

    Adoption and upkeep, setup: favorable, G1 (1 unit before bounds)

    Not counted

    Operational control, admin control: repeats a stronger account in its group

    Not counted

    Operational control, data control: repeats a stronger account in its group

    Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Overall signal
    • Inline Completions: Positive
    • Agent Execution: Mixed or negative
    • Is Copilot the main AI coding tool, or the one engineers keep beside another?: Moved to another tool

    Published on

    Engineering Blogs, Studies & Reviews

    Who

    Individual practitioner

    Evidence

    Reported

    Applies to

    Current

67. Positive

    Ashley Childress, developer using Copilot Pro and Enterprise

    Copilot Premium Requests: More Than Asked, Exactly What You NeedPersonal blog post, DEV Community

    A daily Copilot user on Pro and Enterprise explains premium request limits and the workflow habits that make each monthly allowance last.

    Published Oct 22, 2025self-described: Daily use on Pro and Enterprise plans

    Practitioner account, self-described; direct use; counts toward Delivered outcomes, Reliability, Operational control, Cost predictability and limits How GILD checked it

    Origin

    Published practitioner account

    Checked

    Self-described; No tie to GitHub Copilot found in GILD's check; Used it directly; Needs recheck after 12 months; Specific details

    Role

    Name self-described; employer unknown; role unknown

    Relationship

    No tie to GitHub Copilot found in GILD's check (byline, content disclosure, author profile); that is not proof none exists

    How they know

    Used it directly

    Support

    No artifact or independent corroboration recorded

    Date and version

    2025-10-22, Pro and Enterprise, needs a recheck

    Counts toward

    Cost predictability and limits, limits in practice: unfavorable, G2 (2 units before bounds)

    Counts toward

    Cost predictability and limits, cost predictability: unfavorable, G2 (2 units before bounds)

    Counts toward

    Cost predictability and limits, cost transparency: mixed, G2 (2 units before bounds)

    Counts toward

    Operational control, models and deployment: mixed, G2 (2 units before bounds)

    Counts toward

    Delivered outcomes, correctness: mixed, G2 (2 units before bounds)

    Counts toward

    Reliability, service availability: unfavorable, G1 (1 unit before bounds)

    Not counted

    Delivered outcomes, task completion: repeats a stronger account in its group

    Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward, pending a recheck

    • Overall signal
    • Usage / Cost: Discussed without concern
    • Needs Close Steering: Discussed without concern
    • Written instructions improve results
    • Does Copilot's usage-based billing work for the people paying for it?: Workable

    Published on

    Engineering Blogs, Studies & Reviews

    Who

    Individual practitioner

    Evidence

    Reported

    Applies to

    Needs a recheck

    Observed on

    Premium requests, before AI credits

68. Mixed

    Zardoz84, Hacker News commenter

    Comment on: I am a programmer, not a rubber-stamp that approves Copilot generated codeHacker News comment

    In Eclipse on a legacy Java codebase, completions help with refactoring chores and tests, but agent chat is a coin toss and sometimes breaks file encoding.

    Published Oct 15, 2025self-described: Legacy Java servlet codebase with few tests, using Eclipse

    Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes How GILD checked it

    Origin

    Published practitioner account

    Checked

    Pseudonymous; No tie to GitHub Copilot found in GILD's check; Used it directly; Needs recheck after 12 months; Specific details

    Role

    Pseudonymous account; identity not pursued

    Relationship

    No tie to GitHub Copilot found in GILD's check (byline, content disclosure, platform api profile); that is not proof none exists

    How they know

    Used it directly

    Support

    No artifact or independent corroboration recorded

    Date and version

    2025-10-15, needs a recheck

    Counts toward

    Delivered outcomes, correctness: unfavorable, G2 (2 units before bounds)

    Not counted

    Delivered outcomes, task completion: repeats a stronger account in its group

    Not counted

    Cost predictability and limits, limits in practice: repeats a stronger account in its group

    Checked Oct 30, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Overall signal
    • Inline Completions: Positive
    • Agent Execution: Mixed or negative
    • Editor Integration: Positive
    • Requires Strong Review: Raised as a concern
    • Largest gains on routine code and tests

    Published on

    Developer Communities

    Who

    Individual practitioner

    Evidence

    Reported

    Applies to

    Still applicable

    Observed on

    Copilot in Eclipse, completions and agent chat

69. Positive

    brushfoot, Hacker News commenter, solopreneur

    Comment on: I am a programmer, not a rubber-stamp that approves Copilot generated codeHacker News comment

    Fed a detailed Markdown spec to the Copilot agent on Sonnet 4.5 and reviewed the pull request after lunch; 90 minutes replaced a day, though one generated test was wrong.

    Published Oct 15, 2025self-described: Solo business, customer-facing tools

    Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes How GILD checked it

    Origin

    Published practitioner account

    Checked

    Pseudonymous; No tie to GitHub Copilot found in GILD's check; Used it directly; Specific details; Needs recheck after 12 months

    Role

    Pseudonymous account; identity not pursued

    Relationship

    No tie to GitHub Copilot found in GILD's check (byline, content disclosure, platform api profile); that is not proof none exists

    How they know

    Used it directly

    Support

    No artifact or independent corroboration recorded

    Date and version

    2025-10-15, Sonnet 4.5, needs a recheck

    Counts toward

    Delivered outcomes, task completion: favorable, G2 (2 units before bounds)

    Counts toward

    Delivered outcomes, correctness: unfavorable, G2 (2 units before bounds)

    Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Overall signal
    • Agent Execution: Positive
    • GitHub Integration: Positive
    • Code Quality: Mixed or negative
    • Requires Strong Review: Discussed without concern
    • Needs Close Steering: Discussed without concern
    • Written instructions improve results
    • Once review and rework are counted, does Copilot reduce engineering time?: Net time saved

    Published on

    Developer Communities

    Who

    Individual practitioner

    Evidence

    Reported

    Applies to

    Still applicable

    Observed on

    Copilot cloud agent with Sonnet 4.5

70. Negative

    Nathan Arthur, independent contractor

    AI for Code Review: CodeRabbitNewsletter post, Narthur Online (Substack)

    A contractor who relies on AI review across client repositories finds Copilot's review suggestions limited and less reliable than CodeRabbit's.

    "GitHub Copilot has a similar feature. But in my experience its recommendations are limited and less reliable than CodeRabbit’s."

    Published Aug 19, 2025self-described: Contract work across several client GitHub organizations

    Practitioner account, self-described; direct use; not counted, from a competitor's side How GILD checked it

    Origin

    Published practitioner account

    Checked

    Self-described; Competitor staff: not counted as independent; Used it directly; Needs recheck after 12 months

    Role

    Name self-described; employer unknown; role unknown

    Relationship

    affiliate (self disclosed); affiliate (self disclosed)

    How they know

    Used it directly

    Support

    No artifact or independent corroboration recorded

    Date and version

    2025-08-19, needs a recheck

    Not counted

    Delivered outcomes, correctness: excluded, criticism from a competitor's side, kept as an attributed position

    Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Overall signal
    • Code Quality: Mixed or negative
    • Quality Against Alternatives: Raised as a concern

    Published on

    Engineering Blogs, Studies & Reviews

    Who

    Individual practitioner

    Evidence

    Reported

    Applies to

    Still applicable

    Observed on

    Copilot code review on pull requests

71. Negative

    nodja, Hacker News commenter, self-described

    Comment on: GitHub Copilot Coding AgentHacker News comment

    Found only autocomplete fast; ask, edit and agent modes took minutes on small files, so they canceled and plan to use local models for autocomplete.

    Published May 19, 2025

    Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account How GILD checked it

    Origin

    Published practitioner account

    Checked

    Pseudonymous; No tie to GitHub Copilot found in GILD's check; Used it directly; Needs recheck after 12 months

    Role

    Pseudonymous account; identity not pursued

    Relationship

    No tie to GitHub Copilot found in GILD's check (byline, content disclosure, platform api profile); that is not proof none exists

    How they know

    Used it directly

    Support

    No artifact or independent corroboration recorded

    Date and version

    2025-05-19, needs a recheck

    Not counted

    Delivered outcomes, task completion: repeats a stronger account in its group

    Checked Oct 29, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward, pending a recheck

    • Overall signal
    • Inline Completions: Positive
    • Agent Execution: Mixed or negative
    • Is Copilot the main AI coding tool, or the one engineers keep beside another?: Moved to another tool

    Published on

    Developer Communities

    Who

    Individual practitioner

    Evidence

    Reported

    Applies to

    Needs a recheck

72. Positive

    quantadev, Hacker News commenter

    Comment on: GitHub Copilot Coding AgentHacker News comment

    Uses Copilot chat in VS Code with Claude, accepting edits, then reviewing and testing them; MCP support lets them add their own tools.

    Published May 19, 2025self-described: Solo, VS Code

    Practitioner account, pseudonymous; direct use; counts toward Operational control How GILD checked it

    Origin

    Published practitioner account

    Checked

    Pseudonymous; No tie to GitHub Copilot found in GILD's check; Used it directly; Specific details; Needs recheck after 12 months

    Role

    Pseudonymous account; identity not pursued

    Relationship

    No tie to GitHub Copilot found in GILD's check (byline, content disclosure, platform api profile); that is not proof none exists

    How they know

    Used it directly

    Support

    No artifact or independent corroboration recorded

    Date and version

    2025-05-19, Claude, needs a recheck

    Counts toward

    Operational control, execution control: favorable, G2 (2 units before bounds)

    Counts toward

    Operational control, models and deployment: favorable, G2 (2 units before bounds)

    Not counted

    Delivered outcomes, task completion: repeats a stronger account in its group

    Not counted

    Delivered outcomes, correctness: repeats a stronger account in its group

    Checked Oct 31, 2026 under gild-policy/1.0.0. The scorecard

    Counts toward

    • Overall signal
    • Editor Integration: Positive
    • Code Quality: Positive
    • Requires Strong Review: Discussed without concern
    • Is Copilot the main AI coding tool, or the one engineers keep beside another?: Main tool

    Published on

    Developer Communities

    Who

    Individual practitioner

    Evidence

    Reported

    Applies to

    Still applicable

    Observed on

    Copilot chat in VS Code with Claude models and MCP tools