What engineering leaders and their teams report about Devin Desktop (formerly Windsurf), with every finding tied to its evidence.
Read the findingsPractitioner sentiment, all practitioners
+17 Mixed from 67 sources that take a side
How this figure is madeFrom 63 authors in 42 discussions. Use a task pilot to measure accepted work and review time.
Jan 2025 to Sep 2026.
86 sources from practitioners, by month and sentiment.
The read is positive. Codebase context draws the most praise. Review burden draws the most concern, followed by agent reliability and usage limits after pricing changes. Of 30 sources comparing editors, 22 prefer another tool.
Positive, mixed and negative sources
The score counts them as 26 positive, 22 mixed and 15 negative votes, one per author. See How the numbers work.
See the evidenceBelow: 5 praise themes, 6 concerns, 3 close to agreement and 3 disagreements.
Counted for All practitioners, from 86 sources. Change group
Every capability practitioners describe positively, ranked by how many sources say so. Open one to read who said it.
Counted for All practitioners.
Positive, 15
Positive
gcbirzan, Hacker News commenter whose team moved from Copilot to Windsurf
Comment on: Ask HN: How can I get better at using AI for programming?Hacker News comment
Their team moved to Windsurf from Copilot, and the author finds it most useful for answering questions about a monorepo of a bit over 1M LoC; the supertab feature was also very useful. Almost never asks it to do complex tasks, since describing them takes almost as long as implementing.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes
Positive
Tripti Sheth, Senior Director of Software Engineering at Salesforce
Building AI-Powered Migration Tools: Compressing 4 Sprints Into 3 Days with Cursor, Windsurf, and ClaudeCompany engineering blog post, Salesforce Engineering
Sheth's backend team built a migration tracking UI with Cursor, Windsurf and Claude in three days, work that typically takes about six weeks; Windsurf's rules and memories let it refactor unwieldy code into modular structures.
"Windsurf: We used rules and memories, allowing it to learn from previous context and refactor code into more manageable structures."
Practitioner account, name and role on their organization's own page; direct use; counts toward Delivered outcomes
Positive
Bap (fernandezbaptiste), developer
IDE Comparison with Cursor, Windsurf & Copilot on GPT-5Personal blog post, DEV Community
Testing all three on GPT-5 over a weekend, the author found Windsurf explained an existing codebase best, kept context across the session and scaffolded a project unprompted, with slower inline suggestions.
Practitioner account, self-described; structured test; counts toward Delivered outcomes, Reliability, Operational control, Adoption and upkeep
Positive
Chizaram Ken, frontend developer writing for LogRocket
Windsurf vs. Cursor: When to choose the challengerIndependent review, LogRocket Blog
Building a Next.js todo app with Gemini 2.5 Pro in Cascade, Ken saw it update the correct files without being told and import Lucide-react without installing it. After a planned theme change, the theming "seems to have broken somewhere" until a few more iterations fixed it.
Practitioner account, self-described; direct use; counts toward Delivered outcomes, Operational control, Cost predictability and limits, Adoption and upkeep
Positive
whywhywhywhy, Hacker News commenter
Comment on: Ask HN: Cursor or Windsurf?Hacker News comment
Uses Cursor until hitting a wall, then switches to Windsurf, which the commenter says solves issues Cursor fails on after 40 minutes of retries, though its plans are limited.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes
Mixed
raymondgh, Hacker News commenter
Comment on: Ask HN: How much better are AI IDEs vs. copy pasting into chat apps?Hacker News comment
Vibe coding a web app, the commenter found Cascade good at locating where to change the codebase, but it could not solve performance or CSS animation problems without their direction.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
johntarter, Hacker News commenter
Comment on: OpenAI reaches agreement to buy Windsurf for US$3BHacker News comment
Moving between Windsurf and GitHub Copilot, the commenter credits Windsurf's whole-codebase indexing for completions across files and Cascade's terminal handling, while rating Cursor's checkpoint undo higher.
Practitioner account, pseudonymous; direct use; counts toward Adoption and upkeep
Positive
__jl__, Hacker News commenter
Comment on: OpenAI reaches agreement to buy Windsurf for US$3BHacker News comment
Comparing approaches, the commenter observes that Windsurf still loads more code into context than Cursor's search-tool approach, giving it an edge in some cases at a lower subscription price.
Practitioner account, pseudonymous; opinion; not counted, repeats a stronger account
Positive
Daniel Bergholz, developer
My AI-Powered Workflow for Writing Elixir and Phoenix with WindsurfPersonal blog post, DEV Community
After a month with each, Bergholz chose Windsurf for better codebase understanding and a smoother UI, and uses memories and a .windsurfrules file to correct it on a niche Elixir stack.
Practitioner account, self-described; direct use; counts toward Delivered outcomes, Cost predictability and limits
Positive
bicx, Hacker News commenter
Comment on: Claude 3.7 Sonnet and Claude CodeHacker News comment
Recommends Cascade for agentic code writing and exploration, saying it saved them many hours understanding new codebases and tracing data flows; Claude 3.5 cost credits.
Practitioner account, pseudonymous; direct use; counts toward Operational control
Positive
bluelightning2k, Hacker News commenter, senior engineer
Comment on: Codemaps: Understand Code, Before You Vibe ItHacker News comment
A senior engineer and 12-month user, the commenter says developers overlook Windsurf now that debate centers on Cursor, Claude Code and Codex, and praises Codemaps.
Practitioner account, pseudonymous; direct use; not counted, takes no position
Positive
mpalmer, Hacker News commenter
Comment on: Codemaps: Understand Code, Before You Vibe ItHacker News comment
Using Windsurf at work alongside Claude Code, the commenter says its UX adds real value and Codemaps have been excellent for weeks, though code churn could make them harder to maintain.
Practitioner account, pseudonymous; direct use; not counted, takes no position
Positive
Kristin Foss, VP of Software Engineering for Customer Centric Engineering at Salesforce
How AI Tools Cut Customer Escalation Time: From Days of Manual Work to MinutesCompany engineering blog post, Salesforce Engineering
Foss's escalation engineers use Cursor, Windsurf and Claude CLI to build working reproductions of customer problems from internal source code, finding issues in large codebases in minutes where setup once took hours or days.
"To streamline this process, our CCE engineers now leverage Cursor, Windsurf, and Claude CLI."
Practitioner account, name and role on their organization's own page; ran the rollout; counts toward Delivered outcomes
Mixed
ksymph, Hacker News commenter
Comment on: Ask HN: Cursor or Windsurf?Hacker News comment
After switching between both, the commenter prefers Cursor's completions, credits Windsurf with better project awareness across files that is hard to trigger, and notes its lower price.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
Matt Wolfe, founder of FutureTools and technology video creator
Matt Wolfe on X: tips after 100+ hours of vibe codingX post
Wolfe's first tip for vibe coding is to switch from Cursor to Windsurf, which he finds much better at remembering past context.
"Switch to Windsurf from Cursor. Windsurf seems much better at remembering past context for some reason."
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes, Reliability, Operational control
Mixed or negative, 2
Negative
pbowyer, Hacker News commenter
Comment on: Ask HN: Cursor or Windsurf?Hacker News comment
Trialling both editors on the same PHP codebases, the commenter got poorer results from Windsurf on existing code, saw it read files 50 lines at a time, and went back to PhpStorm.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes
Mixed
pembrook, Hacker News commenter
Comment on: Ask HN: Cursor or Windsurf?Hacker News comment
Once ahead of Cursor on agentic coding in their view, Windsurf now frustrates them by gathering context 100 to 200 lines at a time. The commenter blames that for bugs and poor results.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes
Counted for All practitioners.
Positive, 12
Positive
Maksim Danilchenko, Engineering Lead at inDrive
Devin Desktop (ex-Windsurf): Features, Pricing, What ChangedPersonal blog post, danilchenko.dev
Danilchenko measured Devin Local at roughly 18,000 tokens against about 24,000 for Cascade on a 14-file refactor and saw subagents split work "maybe 40% of the time on multi-step tasks". Calls Devin Local measurably better than Cascade, but says the acquisition story creates real trust risk.
Practitioner account, self-described; direct use; counts toward Delivered outcomes, Reliability, Operational control, Cost predictability and limits, Adoption and upkeep
Positive
scotty79, Hacker News commenter
Comment on: Why does AI feel so different?Hacker News comment
Over three days Windsurf wrote multi-file game logic, refactored on request and fixed bugs that would take them half an hour; the commenter made 95% of changes by instructing it.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
eisfresser, Hacker News commenter
Comment on: Ask HN: Cursor or Windsurf?Hacker News comment
Chooses Windsurf, running parallel Cascade sessions, one investigating a bug while another builds a feature, and drops in later to review and accept the changes.
Practitioner account, pseudonymous; direct use; not counted, takes no position
Mixed
sagarpatil, Hacker News commenter
Comment on: OpenAI reaches agreement to buy Windsurf for US$3BHacker News comment
A customer since launch, the commenter handed Windsurf a broken Alembic migration after an hour of manual debugging; it isolated the file, fixed it, added tests and opened a PR. The commenter now prefers Claude Code.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes
Mixed
CSMastermind, Hacker News commenter who evaluates dev tooling for their teams
Comment on: An update to our pricingHacker News comment
Evaluating tools for their teams, the commenter found Windsurf the one tool that raised their productivity beyond chat, especially on command-line work, but held back a team rollout over pricing.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
thesurlydev, Hacker News commenter
Comment on: JetBrains Fleet drops support for Kotlin MultiplatformHacker News comment
Runs Windsurf next to JetBrains IDEs because the commenter finds Cascade ahead of the 30-plus LLM plugins available for JetBrains, though juggling two editors is a pain.
Practitioner account, pseudonymous; direct use; not counted, takes no position
Mixed
Ravi Patel, solo founder
Windsurf Review 2026: Not For Solo Founders, Great For Small TeamsPersonal blog post, DEV Community (from rikuq.com)
Tried Windsurf's free tier for about a week when the Claude weekly quota ran out, and found tab completion fast and Cascade reliably executing small-to-medium refactors. Calls it good but didn't keep it: its target user isn't a solo agentic-AI shipper, and the paid tier adds Opus access already had directly.
Practitioner account, self-described; direct use; counts toward Delivered outcomes, Reliability, Operational control, Cost predictability and limits
Positive
Bap (fernandezbaptiste), developer
IDE Comparison with Cursor, Windsurf & Copilot on GPT-5Personal blog post, DEV Community
Testing all three on GPT-5 over a weekend, the author found Windsurf explained an existing codebase best, kept context across the session and scaffolded a project unprompted, with slower inline suggestions.
Practitioner account, self-described; structured test; counts toward Delivered outcomes, Reliability, Operational control, Adoption and upkeep
Positive
David Roff, Software Engineer at World Wide Technology
How Windsurf Has Revolutionized My Development WorkflowCompany blog post, WWT
Roff reports a personal productivity gain above 70% and similar gains across his engineering team, to the point that QA struggles to keep up, while every Windsurf change still goes through a senior engineer's review.
"Team-wide benefits: The entire engineering team has experienced similar productivity gains."
Practitioner account, name and role on their organization's own page; direct use; counts toward Delivered outcomes, Operational control
Mixed
Mark Winteringham, tester, toolsmith, author and instructor
Lessons in quality engineering from working with Cursor and WindsurfArticle, Ministry of Testing
Working in small tasks after a messy start in Cursor, Winteringham had the Windsurf agent fix each failing unit test quickly, though it wrote more code than he expected.
"After that, I worked with the Windsurf agent to fix each of the failing tests. This process worked well, and it didn’t take too much time to get them all to pass."
Practitioner account; direct use; counts toward Delivered outcomes, Reliability, Cost predictability and limits
Positive
johntarter, Hacker News commenter
Comment on: OpenAI reaches agreement to buy Windsurf for US$3BHacker News comment
Moving between Windsurf and GitHub Copilot, the commenter credits Windsurf's whole-codebase indexing for completions across files and Cascade's terminal handling, while rating Cursor's checkpoint undo higher.
Practitioner account, pseudonymous; direct use; counts toward Adoption and upkeep
Positive
John Hoestje, Software Architect and Developer at Keyhole Software
Windsurf’s Game-Changing IDE Experience, Part 2, Coding ExamplesCompany blog post, Keyhole Software Dev Blog
Hoestje built two apps from Cascade prompts without editing a line by hand, and says he fell short when he moved too fast to review Cascade's suggested changes.
"I fell short when I was trying to go too fast. I didn’t take the time to properly review the suggested changes by Cascade."
Practitioner account, self-described; direct use; counts toward Delivered outcomes, Reliability, Operational control, Adoption and upkeep
Mixed or negative, 1
Negative
chris_engel, Hacker News commenter
Comment on: OpenAI's Windsurf deal is off, and Windsurf's CEO is going to GoogleHacker News comment
Tried agent workflows in Copilot and Windsurf on Claude 3.5 and 4; they got stuck or built themselves into corners, and the output was fragile enough to refactor immediately.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Counted for All practitioners.
Positive, 10
Positive
RussianCow, Hacker News commenter
Comment on: SWE-1.7 Reach Near GPT 5.5 and Opus IntelligenceHacker News comment
Runs a spec-driven workflow in Devin Desktop over ACP: Claude Code for exploration, Devin's SWE model for building and GLM 5.2 for verification, with little extra effort.
Practitioner account, pseudonymous; direct use; counts toward Operational control
Positive
Edwar Diaz, developer
From Windsurf to Devin Desktop: my first impressions after migratingPersonal blog post, DEV Community
After more than a year on Windsurf, their move to Devin Desktop kept their extensions and settings, and Copilot CLI running over ACP picked up and resumed their existing sessions.
Practitioner account, self-described; direct use; counts toward Adoption and upkeep
Mixed
Saloni Agrawal, React and React Native developer
I Tried Windsurf: An AI Native Coding EnvironmentPersonal blog post, DEV Community
A long-time VS Code user, Agrawal liked queuing several instructions while Cascade works and its persistent memory, but finds the overlap with VS Code agent mode significant.
Practitioner account, self-described; direct use; counts toward Delivered outcomes, Reliability, Operational control
Positive
bluelightning2k, Hacker News commenter, senior engineer
Comment on: Codemaps: Understand Code, Before You Vibe ItHacker News comment
A senior engineer and 12-month user, the commenter says developers overlook Windsurf now that debate centers on Cursor, Claude Code and Codex, and praises Codemaps.
Practitioner account, pseudonymous; direct use; not counted, takes no position
Positive
mpalmer, Hacker News commenter
Comment on: Codemaps: Understand Code, Before You Vibe ItHacker News comment
Using Windsurf at work alongside Claude Code, the commenter says its UX adds real value and Codemaps have been excellent for weeks, though code churn could make them harder to maintain.
Practitioner account, pseudonymous; direct use; not counted, takes no position
Mixed
hightrix, Hacker News commenter
Comment on: Cognition launches new SWE-2 model, Rivaling Fable 5.1 and GPT-AstraHacker News comment
Their company has run it on an enterprise plan for about nine months. The commenter rates it above Copilot but now finds Claude and Codex better, and Devin's web and desktop harnesses inconsistent.
Practitioner account, pseudonymous; direct use; not counted, takes no position
Mixed
sunaookami, Hacker News commenter
Comment on: OpenAI's Windsurf deal is off, and Windsurf's CEO is going to GoogleHacker News comment
Saw Windsurf as slightly better than Cursor, with a nicer chat panel and autocomplete and a US$5 cheaper plan, until losing Anthropic model access cost it users.
Practitioner account, pseudonymous; opinion; not counted, repeats a stronger account
Negative
diegof79, Hacker News commenter
Comment on: OpenAI's Windsurf deal is off, and Windsurf's CEO is going to GoogleHacker News comment
Adopted Windsurf at launch for its UX, then saw more issues with each update, including the agent hanging on shell commands, and moved back to GitHub Copilot.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
tartieret, Hacker News commenter (a CTO, by their own account elsewhere)
Comment on: An update to our pricingHacker News comment
Their company standardized on Windsurf after reviewing Cursor, because Windsurf handled dev containers and WSL development, which most of its developers use, with fewer rough edges.
Practitioner account, pseudonymous; ran the rollout; counts toward Adoption and upkeep
Positive
Daniel Bergholz, developer
My AI-Powered Workflow for Writing Elixir and Phoenix with WindsurfPersonal blog post, DEV Community
After a month with each, Bergholz chose Windsurf for better codebase understanding and a smoother UI, and uses memories and a .windsurfrules file to correct it on a niche Elixir stack.
Practitioner account, self-described; direct use; counts toward Delivered outcomes, Cost predictability and limits
Mixed or negative, 4
Negative
zero-g, Hacker News commenter
Comment on: OpenAI reaches agreement to buy Windsurf for US$3BHacker News comment
Each time the commenter tried the Windsurf Editor and its Vim and IntelliJ plugins it felt unpolished, and autocomplete felt much snappier in Cursor and in GitHub Copilot.
Practitioner account, pseudonymous; direct use; not counted, takes no position
Negative
paimapi, Hacker News commenter
Comment on: Cognition launches new SWE-2 model, Rivaling Fable 5.1 and GPT-AstraHacker News comment
When the product was still Windsurf, terminal runs hung or froze nearly every time until the commenter wired up a workaround skill; the commenter says its multi-agent UI lead has since disappeared.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
Maksim Danilchenko, Engineering Lead at inDrive
Devin Desktop (ex-Windsurf): Features, Pricing, What ChangedPersonal blog post, danilchenko.dev
Danilchenko measured Devin Local at roughly 18,000 tokens against about 24,000 for Cascade on a 14-file refactor and saw subagents split work "maybe 40% of the time on multi-step tasks". Calls Devin Local measurably better than Cascade, but says the acquisition story creates real trust risk.
Practitioner account, self-described; direct use; counts toward Delivered outcomes, Reliability, Operational control, Cost predictability and limits, Adoption and upkeep
Mixed
SquareWheel, Hacker News commenter
Comment on: An update to our pricingHacker News comment
In their first week completions helped with repetitive renames across files but often mangled unrelated code, the UI felt noisy, and chat sometimes scanned files endlessly with no output.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Counted for All practitioners.
Positive, 5
Positive
Patt_, Hacker News commenter
Comment on: Ask HN: US$50 monthly budget, which coding models would you recommend now?Hacker News comment
On a US$50 monthly budget the commenter pays US$20 for Windsurf for access to nearly all models, reserving Claude Opus and Sonnet for tricky work once requirements are clear.
Practitioner account, pseudonymous; direct use; counts toward Cost predictability and limits
Positive
dennisjoseph, Hacker News commenter
Comment on: Ask HN: Who is still using Windsurf and why?Hacker News comment
Keeps Windsurf as their primary IDE because bring-your-own-key covers prompts, agents and autocomplete, including in Vim, which the commenter says Cursor and Zed do not offer.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed
sunaookami, Hacker News commenter
Comment on: OpenAI's Windsurf deal is off, and Windsurf's CEO is going to GoogleHacker News comment
Saw Windsurf as slightly better than Cursor, with a nicer chat panel and autocomplete and a US$5 cheaper plan, until losing Anthropic model access cost it users.
Practitioner account, pseudonymous; opinion; not counted, repeats a stronger account
Mixed
ksymph, Hacker News commenter
Comment on: Ask HN: Cursor or Windsurf?Hacker News comment
After switching between both, the commenter prefers Cursor's completions, credits Windsurf with better project awareness across files that is hard to trigger, and notes its lower price.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
__jl__, Hacker News commenter
Comment on: OpenAI reaches agreement to buy Windsurf for US$3BHacker News comment
Comparing approaches, the commenter observes that Windsurf still loads more code into context than Cursor's search-tool approach, giving it an edge in some cases at a lower subscription price.
Practitioner account, pseudonymous; opinion; not counted, repeats a stronger account
Mixed or negative, 3
Mixed
RussianCow, Hacker News commenter
Comment on: SWE-1.7 Reach Near GPT 5.5 and Opus IntelligenceHacker News comment
After under a month on Devin the commenter hits the US$20 plan's limits faster than with Claude Code or Codex, and rates Cursor's subscription the better value.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed
settled, Hacker News commenter
Comment on: SWE-1.7 Reach Near GPT 5.5 and Opus IntelligenceHacker News comment
Found SWE-1.5 and 1.6 fine for grunt work and tests, but says paid-model usage has been cut sharply since the US$15 plan, and a Codex subscription is now better value.
Practitioner account, pseudonymous; direct use; counts toward Cost predictability and limits
Mixed
Ravi Patel, solo founder
Windsurf Review 2026: Not For Solo Founders, Great For Small TeamsPersonal blog post, DEV Community (from rikuq.com)
Tried Windsurf's free tier for about a week when the Claude weekly quota ran out, and found tab completion fast and Cascade reliably executing small-to-medium refactors. Calls it good but didn't keep it: its target user isn't a solo agentic-AI shipper, and the paid tier adds Opus access already had directly.
Practitioner account, self-described; direct use; counts toward Delivered outcomes, Reliability, Operational control, Cost predictability and limits
Counted for All practitioners.
Positive, 5
Positive
gcbirzan, Hacker News commenter whose team moved from Copilot to Windsurf
Comment on: Ask HN: How can I get better at using AI for programming?Hacker News comment
Their team moved to Windsurf from Copilot, and the author finds it most useful for answering questions about a monorepo of a bit over 1M LoC; the supertab feature was also very useful. Almost never asks it to do complex tasks, since describing them takes almost as long as implementing.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes
Positive
RestlessAPI, Hacker News commenter
Comment on: OpenAI's Windsurf deal is off, and Windsurf's CEO is going to GoogleHacker News comment
Uses Windsurf mainly for completions that predict their next edit, keeping them in control of the codebase and writing everything themselves without delegating to an agent.
Practitioner account, pseudonymous; direct use; not counted, takes no position
Mixed
Ravi Patel, solo founder
Windsurf Review 2026: Not For Solo Founders, Great For Small TeamsPersonal blog post, DEV Community (from rikuq.com)
Tried Windsurf's free tier for about a week when the Claude weekly quota ran out, and found tab completion fast and Cascade reliably executing small-to-medium refactors. Calls it good but didn't keep it: its target user isn't a solo agentic-AI shipper, and the paid tier adds Opus access already had directly.
Practitioner account, self-described; direct use; counts toward Delivered outcomes, Reliability, Operational control, Cost predictability and limits
Positive
dennisjoseph, Hacker News commenter
Comment on: Ask HN: Who is still using Windsurf and why?Hacker News comment
Keeps Windsurf as their primary IDE because bring-your-own-key covers prompts, agents and autocomplete, including in Vim, which the commenter says Cursor and Zed do not offer.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
johntarter, Hacker News commenter
Comment on: OpenAI reaches agreement to buy Windsurf for US$3BHacker News comment
Moving between Windsurf and GitHub Copilot, the commenter credits Windsurf's whole-codebase indexing for completions across files and Cascade's terminal handling, while rating Cursor's checkpoint undo higher.
Practitioner account, pseudonymous; direct use; counts toward Adoption and upkeep
Mixed or negative, 6
Negative
robch, Hacker News commenter
Comment on: Ask HN: Cursor or Windsurf?Hacker News comment
Using their company's Windsurf license for staff-level Android tooling work, the commenter finds autocomplete distracting and the agent hallucinating answers for uncommon tasks.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed
ksymph, Hacker News commenter
Comment on: Ask HN: Cursor or Windsurf?Hacker News comment
After switching between both, the commenter prefers Cursor's completions, credits Windsurf with better project awareness across files that is hard to trigger, and notes its lower price.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed
SquareWheel, Hacker News commenter
Comment on: An update to our pricingHacker News comment
In their first week completions helped with repetitive renames across files but often mangled unrelated code, the UI felt noisy, and chat sometimes scanned files endlessly with no output.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
Bap (fernandezbaptiste), developer
IDE Comparison with Cursor, Windsurf & Copilot on GPT-5Personal blog post, DEV Community
Testing all three on GPT-5 over a weekend, the author found Windsurf explained an existing codebase best, kept context across the session and scaffolded a project unprompted, with slower inline suggestions.
Practitioner account, self-described; structured test; counts toward Delivered outcomes, Reliability, Operational control, Adoption and upkeep
Negative
pbowyer, Hacker News commenter
Comment on: Ask HN: Cursor or Windsurf?Hacker News comment
Trialling both editors on the same PHP codebases, the commenter got poorer results from Windsurf on existing code, saw it read files 50 lines at a time, and went back to PhpStorm.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes
Negative
zero-g, Hacker News commenter
Comment on: OpenAI reaches agreement to buy Windsurf for US$3BHacker News comment
Each time the commenter tried the Windsurf Editor and its Vim and IntelliJ plugins it felt unpolished, and autocomplete felt much snappier in Cursor and in GitHub Copilot.
Practitioner account, pseudonymous; direct use; not counted, takes no position
Every concern practitioners raise, most raised first.
Counted for All practitioners.
Raised as a concern, 18
Negative
sensanaty, Hacker News commenter
Comment on: Labor market impacts of AI: A new measure and early evidenceHacker News comment
Tried Windsurf's Arena Mode on a test-completion task, with AGENTS.md and rules files in place, and saw the same failure as in Claude Code; the commenter reviews every command it runs.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes
Positive
Darryn Dieken, Chief Availability Officer at Salesforce
How AI-Enabled Tooling Boosted Code Output 30% While Keeping Quality and Deployment Safety IntactCompany engineering blog post, Salesforce Engineering
Salesforce offers Cursor and Windsurf for code and test generation, with over 85% of eligible engineers using AI tools weekly; code in production rose 30%, which pushed the bottleneck to review and testing.
"A few tools in this portfolio include: Cursor and Windsurf for AI-assisted code and tests generation…"
Practitioner account, name and role on their organization's own page; ran the rollout; not counted, takes no position
Negative
badaigent, Hacker News commenter
Comment on: Vibe coding service Replit deleted production database, faked data, told fibsHacker News comment
Reports that Windsurf's agent replaced their dashboard with its own component, removed files without approval despite their rules, and broke the app again after restoring it.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes, Reliability, Operational control
Mixed
justinclift, Hacker News commenter
Comment on: Cognition (Devin AI) to Acquire WindsurfHacker News comment
Their workplace still uses Windsurf, but the commenter canceled their personal plan because Claude Sonnet 3.7 output and issue analysis in it were too often wrong to justify paying.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
David Roff, Software Engineer at World Wide Technology
How Windsurf Has Revolutionized My Development WorkflowCompany blog post, WWT
Roff reports a personal productivity gain above 70% and similar gains across his engineering team, to the point that QA struggles to keep up, while every Windsurf change still goes through a senior engineer's review.
"Team-wide benefits: The entire engineering team has experienced similar productivity gains."
Practitioner account, name and role on their organization's own page; direct use; counts toward Delivered outcomes, Operational control
Negative
powerapple, Hacker News commenter
Comment on: Ask HN: Cursor or Windsurf?Hacker News comment
Testing Windsurf, the commenter found it installed dependencies into their global Python without a virtual environment and rewrote code the commenter had not asked it to change.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Negative
artdigital, Hacker News commenter
Comment on: Ask HN: Cursor or Windsurf?Hacker News comment
Found the same models produced noticeably worse code in Windsurf than in Zed, Copilot or Cursor, with erratic tool use, so the commenter canceled and made Copilot their daily driver.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed
osigurdson, Hacker News commenter
Comment on: OpenAI reaches agreement to buy Windsurf for US$3BHacker News comment
The commenter likes Windsurf for exploration but writes code they want to keep by hand, warning that without deep review of every Windsurf change you lose understanding of the codebase.
Practitioner account, pseudonymous; direct use; not counted, takes no position
Mixed
SquareWheel, Hacker News commenter
Comment on: An update to our pricingHacker News comment
In their first week completions helped with repetitive renames across files but often mangled unrelated code, the UI felt noisy, and chat sometimes scanned files endlessly with no output.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
John Hoestje, Software Architect and Developer at Keyhole Software
Windsurf’s Game-Changing IDE Experience, Part 2, Coding ExamplesCompany blog post, Keyhole Software Dev Blog
Hoestje built two apps from Cascade prompts without editing a line by hand, and says he fell short when he moved too fast to review Cascade's suggested changes.
"I fell short when I was trying to go too fast. I didn’t take the time to properly review the suggested changes by Cascade."
Practitioner account, self-described; direct use; counts toward Delivered outcomes, Reliability, Operational control, Adoption and upkeep
Positive
Søren Kottal, Umbraco developer
Vibe coding a nifty Umbraco toolPersonal blog post, DEV Community
Built an Umbraco thumbnail tool with Windsurf in a few hours, but it reintroduced old bugs two or three times. The author did not read the code, so would not use it for supported work.
Practitioner account, self-described; direct use; counts toward Delivered outcomes
Positive
Maksim Danilchenko, Engineering Lead at inDrive
Devin Desktop (ex-Windsurf): Features, Pricing, What ChangedPersonal blog post, danilchenko.dev
Danilchenko measured Devin Local at roughly 18,000 tokens against about 24,000 for Cascade on a 14-file refactor and saw subagents split work "maybe 40% of the time on multi-step tasks". Calls Devin Local measurably better than Cascade, but says the acquisition story creates real trust risk.
Practitioner account, self-described; direct use; counts toward Delivered outcomes, Reliability, Operational control, Cost predictability and limits, Adoption and upkeep
Positive
Chizaram Ken, frontend developer writing for LogRocket
Windsurf vs. Cursor: When to choose the challengerIndependent review, LogRocket Blog
Building a Next.js todo app with Gemini 2.5 Pro in Cascade, Ken saw it update the correct files without being told and import Lucide-react without installing it. After a planned theme change, the theming "seems to have broken somewhere" until a few more iterations fixed it.
Practitioner account, self-described; direct use; counts toward Delivered outcomes, Operational control, Cost predictability and limits, Adoption and upkeep
Negative
chris_engel, Hacker News commenter
Comment on: OpenAI's Windsurf deal is off, and Windsurf's CEO is going to GoogleHacker News comment
Tried agent workflows in Copilot and Windsurf on Claude 3.5 and 4; they got stuck or built themselves into corners, and the output was fragile enough to refactor immediately.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed
Mark Winteringham, tester, toolsmith, author and instructor
Lessons in quality engineering from working with Cursor and WindsurfArticle, Ministry of Testing
Working in small tasks after a messy start in Cursor, Winteringham had the Windsurf agent fix each failing unit test quickly, though it wrote more code than he expected.
"After that, I worked with the Windsurf agent to fix each of the failing tests. This process worked well, and it didn’t take too much time to get them all to pass."
Practitioner account; direct use; counts toward Delivered outcomes, Reliability, Cost predictability and limits
Negative
robch, Hacker News commenter
Comment on: Ask HN: Cursor or Windsurf?Hacker News comment
Using their company's Windsurf license for staff-level Android tooling work, the commenter finds autocomplete distracting and the agent hallucinating answers for uncommon tasks.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed
raymondgh, Hacker News commenter
Comment on: Ask HN: How much better are AI IDEs vs. copy pasting into chat apps?Hacker News comment
Vibe coding a web app, the commenter found Cascade good at locating where to change the codebase, but it could not solve performance or CSS animation problems without their direction.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
Daniel Bergholz, developer
My AI-Powered Workflow for Writing Elixir and Phoenix with WindsurfPersonal blog post, DEV Community
After a month with each, Bergholz chose Windsurf for better codebase understanding and a smoother UI, and uses memories and a .windsurfrules file to correct it on a niche Elixir stack.
Practitioner account, self-described; direct use; counts toward Delivered outcomes, Cost predictability and limits
Discussed without concern, 1
Positive
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 engineers have Windsurf, Copilot and Claude Code. Today peer review means using the peer review agent in Windsurf or Claude Code, which "makes life better" but is still human driven; the goal is a deterministic evaluation when a merge request hits.
"Right now, peer review is something that like, “Hey, you got Windsurf or you have ClaudeCode installed, you used the peer review agent, it makes life better.” It’s still human driven."
GILD's fact checks, which don't count as sources
Checked Oct 12, 2026.Quick Review runs a second agent over local Devin Local changes before commit, with the SWE-check model free on every plan. It is not available for the legacy Cascade agent and does not replace human review before merge. Quick Review
Open them in the recordCounted for All practitioners.
Negative
paimapi, Hacker News commenter
Comment on: Cognition launches new SWE-2 model, Rivaling Fable 5.1 and GPT-AstraHacker News comment
When the product was still Windsurf, terminal runs hung or froze nearly every time until the commenter wired up a workaround skill; the commenter says its multi-agent UI lead has since disappeared.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed
oib, Windsurf user filing a bug report
Windsurf Bug Report - Memory Leak Consuming All RAM on LinuxGitHub issue
Reports that Windsurf on Debian consumed all 64 GB of RAM after startup until it crashed, with no workaround found.
Practitioner account, pseudonymous; direct use; counts toward Reliability
Mixed
aeongeon, Windsurf user filing a bug report
# Windsurf IDE Freeze Bug ReportGitHub issue
Measured the editor freezing for about five minutes after every Cascade response on Linux, traced with strace to a blocked renderer thread and dozens of duplicate calls.
Practitioner account, pseudonymous; structured test; counts toward Reliability, Operational control
Mixed
joyboy123456, Windsurf user filing a bug report
[Windsurf Bug] AI keeps getting stuck executing terminal commandsGitHub issue
Reports that every terminal command the agent ran on Windows PowerShell stayed in a running state and never returned, while the same commands worked when run by hand.
Practitioner account, pseudonymous; direct use; counts toward Reliability
Negative
diegof79, Hacker News commenter
Comment on: OpenAI's Windsurf deal is off, and Windsurf's CEO is going to GoogleHacker News comment
Adopted Windsurf at launch for its UX, then saw more issues with each update, including the agent hanging on shell commands, and moved back to GitHub Copilot.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed
AldrinSalazar, Windsurf user filing a bug report
Cascade stuck in a command execution loop because it cannot execute commandsGitHub issue
Reports that on Windows 11 every Cascade attempt to run a command failed and left the agent stuck in a loop.
Practitioner account, pseudonymous; direct use; counts toward Reliability
Mixed
Bwogi, Windsurf subscriber filing a bug report
The model produced an invalid tool call, trying again.GitHub issue on Exafunction/codeium
After an update, every Cascade step failed with an invalid tool call and then an internal error, leaving the reporter unable to use a US$60 monthly subscription that had worked smoothly before.
"…i was coding smoothly without any issues before the update. Now i cannot use my 60 buck monthly subscription anymore."
Practitioner account, pseudonymous; direct use; counts toward Reliability
Mixed
osigurdson, Hacker News commenter
Comment on: My AI skeptic friends are all nutsHacker News comment
After a few months the commenter finds Windsurf useful when tired or in unfamiliar areas, but waiting and errors break flow. On familiar work the commenter puts it at half the speed of typing.
Practitioner account, pseudonymous; direct use; counts toward Reliability
Negative
artdigital, Hacker News commenter
Comment on: Ask HN: Cursor or Windsurf?Hacker News comment
Found the same models produced noticeably worse code in Windsurf than in Zed, Copilot or Cursor, with erratic tool use, so the commenter canceled and made Copilot their daily driver.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed
SquareWheel, Hacker News commenter
Comment on: An update to our pricingHacker News comment
In their first week completions helped with repetitive renames across files but often mangled unrelated code, the UI felt noisy, and chat sometimes scanned files endlessly with no output.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Counted for All practitioners.
Raised as a concern, 9
Negative
princevegeta89, Hacker News commenter
Comment on: Cognition launches new SWE-2 model, Rivaling Fable 5.1 and GPT-AstraHacker News comment
A long-time Windsurf user, the commenter says the product has fallen several features behind Cursor since becoming Devin, and has gone back to VS Code with Copilot Pro.
Practitioner account, pseudonymous; direct use; counts toward Adoption and upkeep
Mixed
fishtoaster, Hacker News commenter
Comment on: Cognition launches new SWE-2 model, Rivaling Fable 5.1 and GPT-AstraHacker News comment
Enjoying Devin's cloud agents after comparing alternatives, the commenter calls the gap between Devin cloud and Devin Desktop embarrassing, describing Desktop as a thinly reskinned Windsurf.
Practitioner account, pseudonymous; opinion; not counted, takes no position
Negative
godzillabrennus, Hacker News commenter
Comment on: SWE-1.7 Reach Near GPT 5.5 and Opus IntelligenceHacker News comment
Their organization was a Windsurf customer at the acquisition, then left after Cognition stopped offering support, raised prices twice and dismantled the brand.
Practitioner account, pseudonymous; direct use; counts toward Adoption and upkeep
Positive
Maksim Danilchenko, Engineering Lead at inDrive
Devin Desktop (ex-Windsurf): Features, Pricing, What ChangedPersonal blog post, danilchenko.dev
Danilchenko measured Devin Local at roughly 18,000 tokens against about 24,000 for Cascade on a 14-file refactor and saw subagents split work "maybe 40% of the time on multi-step tasks". Calls Devin Local measurably better than Cascade, but says the acquisition story creates real trust risk.
Practitioner account, self-described; direct use; counts toward Delivered outcomes, Reliability, Operational control, Cost predictability and limits, Adoption and upkeep
Negative
cactca, Hacker News commenter and self-described Windsurf subscriber
Comment on Hacker News: cancelled Windsurf after the acquisitionHacker News comment
After Cognition acquired Windsurf, this subscriber cancelled and moved to Zed and Claude Code.
Mixed
sunaookami, Hacker News commenter
Comment on: OpenAI's Windsurf deal is off, and Windsurf's CEO is going to GoogleHacker News comment
Saw Windsurf as slightly better than Cursor, with a nicer chat panel and autocomplete and a US$5 cheaper plan, until losing Anthropic model access cost it users.
Practitioner account, pseudonymous; opinion; not counted, repeats a stronger account
Mixed
tianshuo, Hacker News commenter
Comment on: Windsurf SWE-1: Our First Frontier ModelsHacker News comment
A paid user, the commenter asks Windsurf to stop chasing frontier models and build predictable workflows: task boards, TDD, git branches without reliance on undo, and docs kept in sync with code.
Practitioner account, pseudonymous; direct use; not counted, takes no position
Mixed
hightrix, Hacker News commenter
Comment on: Cognition launches new SWE-2 model, Rivaling Fable 5.1 and GPT-AstraHacker News comment
Their company has run it on an enterprise plan for about nine months. The commenter rates it above Copilot but now finds Claude and Codex better, and Devin's web and desktop harnesses inconsistent.
Practitioner account, pseudonymous; direct use; not counted, takes no position
No opinion stated
Maxwell Zeff, senior reporter, TechCrunch
TechCrunch: Windsurf says Anthropic is limiting its direct access to Claude AI modelsNews report
Anthropic cut Windsurf's direct Claude capacity with little notice in June 2025, and a Windsurf customer since late 2024 switched to Cursor to keep using Claude 4.
"Windsurf CEO Varun Mohan said in a post on X Tuesday that Anthropic gave Windsurf little notice for this change, and the startup now has to find other third-party compute providers to run several of Anthropic’s most popular AI models on its platform."
Discussed without concern, 3
Mixed
bluelightning2k, Hacker News commenter
Comment on: Cognition launches new SWE-2 model, Rivaling Fable 5.1 and GPT-AstraHacker News comment
Still uses Devin Desktop, formerly Windsurf, as their editor but has not used its agent for a while because Codex now handles that work for them.
Practitioner account, pseudonymous; direct use; not counted, takes no position
Positive
fallinditch, Hacker News commenter
Comment on: SWE-1.7 Reach Near GPT 5.5 and Opus IntelligenceHacker News comment
Uses Cognition's SWE-1.6 model for grunt work alongside Opus for planning and harder problems, and is pleased with Cognition's progress since it took over Windsurf.
Practitioner account, pseudonymous; direct use; counts toward Adoption and upkeep
Positive
Edwar Diaz, developer
From Windsurf to Devin Desktop: my first impressions after migratingPersonal blog post, DEV Community
After more than a year on Windsurf, their move to Devin Desktop kept their extensions and settings, and Copilot CLI running over ACP picked up and resumed their existing sessions.
Practitioner account, self-described; direct use; counts toward Adoption and upkeep
GILD's fact checks, which don't count as sources
Checked Oct 12, 2026.Cognition's FAQ states that plans and pricing did not change with the June 2, 2026 rename. Existing settings carry over, and the Windsurf JetBrains plugin is in maintenance mode. Devin Desktop FAQ
Open them in the recordCounted for All practitioners.
Raised as a concern, 9
Mixed
PolCPP, Hacker News commenter
Comment on: Cognition launches new SWE-2 model, Rivaling Fable 5.1 and GPT-AstraHacker News comment
Grandfathered on old pricing, the commenter used Windsurf as their main editor until the pricing model changed and now uses it only to spend their weekly tokens on frontier models.
Practitioner account, pseudonymous; direct use; not counted, takes no position
Mixed
RussianCow, Hacker News commenter
Comment on: SWE-1.7 Reach Near GPT 5.5 and Opus IntelligenceHacker News comment
After under a month on Devin the commenter hits the US$20 plan's limits faster than with Claude Code or Codex, and rates Cursor's subscription the better value.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed
settled, Hacker News commenter
Comment on: SWE-1.7 Reach Near GPT 5.5 and Opus IntelligenceHacker News comment
Found SWE-1.5 and 1.6 fine for grunt work and tests, but says paid-model usage has been cut sharply since the US$15 plan, and a Codex subscription is now better value.
Practitioner account, pseudonymous; direct use; counts toward Cost predictability and limits
Negative
visarga, Hacker News commenter
Comment on: Tell HN: Anthropic no longer allowing Claude Code subscriptions to use OpenClawHacker News comment
Left Windsurf, and later Cursor, after each cut their usage quota within weeks or months of their paying for a plan; the commenter never returned to Windsurf.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed
CSMastermind, Hacker News commenter who evaluates dev tooling for their company
Comment on: Codemaps: Understand Code, Before You Vibe ItHacker News comment
Still rates Windsurf ahead of Cursor but has moved to Codex for its cloud environments; pricing kept them from rolling Windsurf out across their company.
Practitioner account, pseudonymous; direct use; not counted, takes no position
Mixed
CSMastermind, Hacker News commenter who evaluates dev tooling for their teams
Comment on: An update to our pricingHacker News comment
Evaluating tools for their teams, the commenter found Windsurf the one tool that raised their productivity beyond chat, especially on command-line work, but held back a team rollout over pricing.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Negative
godzillabrennus, Hacker News commenter
Comment on: SWE-1.7 Reach Near GPT 5.5 and Opus IntelligenceHacker News comment
Their organization was a Windsurf customer at the acquisition, then left after Cognition stopped offering support, raised prices twice and dismantled the brand.
Practitioner account, pseudonymous; direct use; counts toward Adoption and upkeep
Positive
whywhywhywhy, Hacker News commenter
Comment on: Ask HN: Cursor or Windsurf?Hacker News comment
Uses Cursor until hitting a wall, then switches to Windsurf, which the commenter says solves issues Cursor fails on after 40 minutes of retries, though its plans are limited.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes
Negative
ramesh31, Hacker News commenter
Comment on: An update to our pricingHacker News comment
A daily full-time Windsurf user, the commenter calls it inferior to open-source agents such as Cline and Claude Code that use provider keys directly, without flow credits.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Discussed without concern, 4
Positive
Maksim Danilchenko, Engineering Lead at inDrive
Devin Desktop (ex-Windsurf): Features, Pricing, What ChangedPersonal blog post, danilchenko.dev
Danilchenko measured Devin Local at roughly 18,000 tokens against about 24,000 for Cascade on a 14-file refactor and saw subagents split work "maybe 40% of the time on multi-step tasks". Calls Devin Local measurably better than Cascade, but says the acquisition story creates real trust risk.
Practitioner account, self-described; direct use; counts toward Delivered outcomes, Reliability, Operational control, Cost predictability and limits, Adoption and upkeep
Mixed
zazibar, Hacker News commenter at a company with over 400 engineers
Comment on: Anthropic downgraded cache TTL on March 6thHacker News comment
Their 400-engineer company canceled all IDE subscriptions, Windsurf included, and moved everyone to Claude Code as a cost-saving measure with no migration plan. Token spend then forced limits.
Practitioner account, pseudonymous; second-hand; not counted, takes no position
Positive
Patt_, Hacker News commenter
Comment on: Ask HN: US$50 monthly budget, which coding models would you recommend now?Hacker News comment
On a US$50 monthly budget the commenter pays US$20 for Windsurf for access to nearly all models, reserving Claude Opus and Sonnet for tricky work once requirements are clear.
Practitioner account, pseudonymous; direct use; counts toward Cost predictability and limits
Positive
bicx, Hacker News commenter
Comment on: Claude 3.7 Sonnet and Claude CodeHacker News comment
Recommends Cascade for agentic code writing and exploration, saying it saved them many hours understanding new codebases and tracing data flows; Claude 3.5 cost credits.
Practitioner account, pseudonymous; direct use; counts toward Operational control
GILD's fact checks, which don't count as sources
Checked Oct 12, 2026.Since March 2026, self-serve plans include a daily and weekly quota drawn down per token, with extra usage billed at API list prices. On Teams, admins set auto-reload thresholds and default session spending limits. Reports about credit plans describe a system that no longer applies. Quota-based usage
Open them in the recordCounted for All practitioners.
Raised as a concern, 6
Negative
pbowyer, Hacker News commenter
Comment on: Ask HN: Cursor or Windsurf?Hacker News comment
Trialling both editors on the same PHP codebases, the commenter got poorer results from Windsurf on existing code, saw it read files 50 lines at a time, and went back to PhpStorm.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes
Mixed
pembrook, Hacker News commenter
Comment on: Ask HN: Cursor or Windsurf?Hacker News comment
Once ahead of Cursor on agentic coding in their view, Windsurf now frustrates them by gathering context 100 to 200 lines at a time. The commenter blames that for bugs and poor results.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes
Mixed
ksymph, Hacker News commenter
Comment on: Ask HN: Cursor or Windsurf?Hacker News comment
After switching between both, the commenter prefers Cursor's completions, credits Windsurf with better project awareness across files that is hard to trigger, and notes its lower price.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed
nadavoid, Windsurf user filing a feature request
Remove or increase limit of past conversations in Cascade (Windsurf)GitHub issue
Asks Windsurf to lift Cascade's limit of 20 saved conversations, since the oldest is lost for good when a 21st starts and past conversations are worth revisiting.
Practitioner account, pseudonymous; direct use; counts toward Operational control
Mixed
tianshuo, Hacker News commenter
Comment on: Windsurf SWE-1: Our First Frontier ModelsHacker News comment
A paid user, the commenter asks Windsurf to stop chasing frontier models and build predictable workflows: task boards, TDD, git branches without reliance on undo, and docs kept in sync with code.
Practitioner account, pseudonymous; direct use; not counted, takes no position
Positive
Søren Kottal, Umbraco developer
Vibe coding a nifty Umbraco toolPersonal blog post, DEV Community
Built an Umbraco thumbnail tool with Windsurf in a few hours, but it reintroduced old bugs two or three times. The author did not read the code, so would not use it for supported work.
Practitioner account, self-described; direct use; counts toward Delivered outcomes
Discussed without concern, 3
Mixed
Saloni Agrawal, React and React Native developer
I Tried Windsurf: An AI Native Coding EnvironmentPersonal blog post, DEV Community
A long-time VS Code user, Agrawal liked queuing several instructions while Cascade works and its persistent memory, but finds the overlap with VS Code agent mode significant.
Practitioner account, self-described; direct use; counts toward Delivered outcomes, Reliability, Operational control
Positive
mpalmer, Hacker News commenter
Comment on: Codemaps: Understand Code, Before You Vibe ItHacker News comment
Using Windsurf at work alongside Claude Code, the commenter says its UX adds real value and Codemaps have been excellent for weeks, though code churn could make them harder to maintain.
Practitioner account, pseudonymous; direct use; not counted, takes no position
Positive
Bap (fernandezbaptiste), developer
IDE Comparison with Cursor, Windsurf & Copilot on GPT-5Personal blog post, DEV Community
Testing all three on GPT-5 over a weekend, the author found Windsurf explained an existing codebase best, kept context across the session and scaffolded a project unprompted, with slower inline suggestions.
Practitioner account, self-described; structured test; counts toward Delivered outcomes, Reliability, Operational control, Adoption and upkeep
GILD's fact checks, which don't count as sources
Checked Oct 12, 2026.Memories apply only to the legacy Cascade agent; Devin Local reads rules in .devin/rules and AGENTS.md files, and Cognition recommends moving memories to skills with the Cascade Migration Wizard. Memories and rules
Open them in the recordCounted for All practitioners.
Raised as a concern, 3
Negative
chris_engel, Hacker News commenter
Comment on: OpenAI's Windsurf deal is off, and Windsurf's CEO is going to GoogleHacker News comment
Tried agent workflows in Copilot and Windsurf on Claude 3.5 and 4; they got stuck or built themselves into corners, and the output was fragile enough to refactor immediately.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Negative
powerapple, Hacker News commenter
Comment on: Ask HN: Cursor or Windsurf?Hacker News comment
Testing Windsurf, the commenter found it installed dependencies into their global Python without a virtual environment and rewrote code the commenter had not asked it to change.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed
SquareWheel, Hacker News commenter
Comment on: An update to our pricingHacker News comment
In their first week completions helped with repetitive renames across files but often mangled unrelated code, the UI felt noisy, and chat sometimes scanned files endlessly with no output.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Discussed without concern, 5
Positive
Bap (fernandezbaptiste), developer
IDE Comparison with Cursor, Windsurf & Copilot on GPT-5Personal blog post, DEV Community
Testing all three on GPT-5 over a weekend, the author found Windsurf explained an existing codebase best, kept context across the session and scaffolded a project unprompted, with slower inline suggestions.
Practitioner account, self-described; structured test; counts toward Delivered outcomes, Reliability, Operational control, Adoption and upkeep
Positive
scotty79, Hacker News commenter
Comment on: Why does AI feel so different?Hacker News comment
Over three days Windsurf wrote multi-file game logic, refactored on request and fixed bugs that would take them half an hour; the commenter made 95% of changes by instructing it.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
Steve Lewis, Lead Software Engineer at World Wide Technology
Windsurf In Action: Building AWS InfrastructureCompany blog post, WWT
Lewis reports that the team has used Windsurf for a few months. Planning the architecture with it, Lewis built S3 and CloudFront hosting in Terraform plus a Node.js upload script in less than 1 day against an estimated 2-3 days, and argues such tools need someone to drive them and provide context.
"Our team has been using Windsurf as an AI-powered development environment for a few months now."
Practitioner account, name and role on their organization's own page; direct use; counts toward Delivered outcomes
Positive
eisfresser, Hacker News commenter
Comment on: Ask HN: Cursor or Windsurf?Hacker News comment
Chooses Windsurf, running parallel Cascade sessions, one investigating a bug while another builds a feature, and drops in later to review and accept the changes.
Practitioner account, pseudonymous; direct use; not counted, takes no position
Positive
Matt Wolfe, founder of FutureTools and technology video creator
Matt Wolfe on X: tips after 100+ hours of vibe codingX post
Wolfe's first tip for vibe coding is to switch from Cursor to Windsurf, which he finds much better at remembering past context.
"Switch to Windsurf from Cursor. Windsurf seems much better at remembering past context for some reason."
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes, Reliability, Operational control
Statements this group's sources agree on, then the ones close to agreement.
Close to agreement: each needs 18 supporting sources
On self-directed work it shortens the path from an idea to a working change.
8 supporting sourcesThe strongest reported results come from well-defined tasks such as tests and routine refactors, with the engineer setting direction.
7 supporting sourcesPractitioners evaluate it as an agent that carries out engineering tasks they then review. They do not treat it as a completion tool.
5 supporting sourcesQuestions the sources split on, with the count on each side.
Evidence remains mixed
Kept it, 8 sources
Some chose it over Cursor or Copilot for codebase awareness, a cleaner editor, dev container and WSL support, or price.
Moved to another tool, 22 sources
Others moved to Cursor, GitHub Copilot, Claude Code, or Codex, citing output quality, reliability, pricing changes, or the change in ownership.
Evidence remains mixed
Net time saved, 8 sources
Some report clear gains, from a migration bug fixed after an hour of manual debugging to tasks finished well ahead of hand coding.
Absorbed by review, 2 sources
Others find waiting, errors, and fragile output cancel the gain on work they know well.
Evidence remains mixed
Broader delegation, 4 sources
Some run parallel agent sessions or describe most changes in prose and let the agent make them.
Tighter steering, 5 sources
Others write the outline themselves, or use it mainly for completions, so they stay in control of the code.
The three groups, then every practitioner by the context they work in, and how emphasis changes with scale.
Three groups drawn from the same sources. Choose a group to select whose experience counts. Within that group, each author counts once in the practitioner sentiment, with their latest view. Signals open on All practitioners. Leaders and teams have 14 accounts; GILD prints their sentiment from 20.
23 sources. Leaders' accounts cover rollout decisions, pricing, vendor stability, and what changes when Devin Desktop moves from individual use to a team standard.
63 sources. Individual builders' accounts cover codebase context, completions, agent reliability, and usage limits in self-directed work.
86 sources. All accounts together give the broadest read. They combine individual and team settings that can reach different conclusions.
86 sources from practitioners, by the context they work in. Filled rows count for this group.
Emphasizes Codebase context and usage limits
Emphasizes Company licenses and output quality
Emphasizes Standardizing on an editor and pricing for a rollout
Emphasizes Company-wide seats and whether engineers use them
What changes with scale. Individual builders write most about how well it understands their codebase and about usage limits. Team and leadership accounts focus on the rollout decision: two CTOs standardized on the editor, ServiceNow's across about 7,000 engineers, while other companies canceled seats or moved to Claude Code.
Where this group's sources were published.
Counted for All practitioners, from 86 sources. Change group
106 sources for this group, 20 of them documentation
Older evidence stays when it still applies. Sources tied to Cascade, credit pricing, or a replaced model remain marked Needs a recheck or Out of date.
See sources by originWho each source is, how it knows, whether it still applies and what it counts toward.
Newest first, on its own page. Every count above opens its sources there.
Read all 106 sourcesAbout this read
A running review of what engineering teams report about Devin Desktop (formerly Windsurf), from individual builders to company-wide rollouts. Each finding links to its sources, and each conflict shows both sides.
Counts cover only the sources GILD reviewed.
Choose a group to select whose experience counts. Within that group, each author counts once in the practitioner sentiment, with their latest view. GILD files each source in a group from how the account describes its author. A role or employer marked self-described is what the source states about themselves. Documentation checks facts for every group and never counts toward sentiment.
A source's date and whether it still applies are separate: a recent post can describe an old version, and an old one can still hold.
Flagged for a recheck, of the sources that take a side: leaders and teams 7 of 14, individual builders 41 of 53, all practitioners 48 of 67.
Each row opens its sources for the group you are reading.
| Finding | Engineering leaders and teams | Individual builders | All practitioners | Sources |
|---|---|---|---|---|
| Current readCurrent sentiment | 7 positive, 5 mixed, 2 negative | 21 positive, 19 mixed, 13 negative | 28 positive, 24 mixed, 15 negative | Sources |
| PraiseCodebase Context | 3 of 3 positive (under the minimum) | 12 of 14 positive | 15 of 17 positive | Sources |
| PraiseAgent Execution | 2 of 2 positive (under the minimum) | 10 of 11 positive | 12 of 13 positive | Sources |
| PraiseTab Completion | 1 of 2 positive (under the minimum) | 4 of 9 positive | 5 of 11 positive | Sources |
| PraiseEditor Experience | 2 of 2 positive (under the minimum) | 8 of 12 positive | 10 of 14 positive | Sources |
| PraiseValue for Money | 0 of 0 positive (under the minimum) | 5 of 8 positive | 5 of 8 positive | Sources |
| ConcernUsage / Cost | 3 of 4 raise it (under the minimum) | 6 of 9 raise it | 9 of 13 raise it | Sources |
| ConcernAgent & Editor Reliability | 0 of 0 raise it (under the minimum) | 10 of 10 raise it | 10 of 10 raise it | Sources |
| ConcernRequires Strong Review | 4 of 5 raise it | 14 of 14 raise it | 18 of 19 raise it | Sources |
| ConcernContext Management | 0 of 0 raise it (under the minimum) | 6 of 9 raise it | 6 of 9 raise it | Sources |
| ConcernVendor Direction | 2 of 2 raise it (under the minimum) | 7 of 10 raise it | 9 of 12 raise it | Sources |
| ConcernNeeds Close Steering | 0 of 1 raise it (under the minimum) | 3 of 7 raise it | 3 of 8 raise it | Sources |
| ConsensusEvaluated as an engineering agent | 0 supporting (under the minimum) | 5 supporting (below threshold) | 5 supporting (below threshold) | Sources |
| ConsensusFaster on self-directed work | 4 supporting (under the minimum) | 4 supporting (under the minimum) | 8 supporting (below threshold) | Sources |
| ConsensusStrongest on well-scoped work | 2 supporting (under the minimum) | 5 supporting (below threshold) | 7 supporting (below threshold) | Sources |
| DisagreementAfter trying it next to other tools, do practitioners keep Devin Desktop (formerly Windsurf)? | Kept it 4, Moved to another tool 7 | Kept it 4, Moved to another tool 15 | Kept it 8, Moved to another tool 22 | Sources |
| DisagreementHow much work should the agent complete before a human checkpoint? | Broader delegation 0, Tighter steering 1 (under the minimum) | Broader delegation 4, Tighter steering 4 | Broader delegation 4, Tighter steering 5 | Sources |
| DisagreementOnce review and rework are counted, does it save engineering time? | Net time saved 6, Absorbed by review 0 (under the minimum) | Net time saved 2, Absorbed by review 2 (under the minimum) | Net time saved 8, Absorbed by review 2 | Sources |
| Current readEvidence mix | 43 sources | 83 sources | 106 sources | Sources |
GILD's checks of claims against official documentation, kept apart from the source counts.
