Skip to main content
ZedProduct page

Zed Signals

What engineering leaders and their teams report about Zed, with every finding tied to its evidence.

Read the findings

What the evidence says

Whose sources count

Practitioner sentiment, all practitioners

+25 Positive from 57 sources that take a side

How this figure is made

Positive overall signal, net 25 points

From 56 authors in 30 discussions. Use a task pilot to measure accepted work and review time.

Analyzed
95 total sources
Top praise
Zed Agent and Review Flow (16 of 17 sources) See praise evidence
Top concern
External Agents Lose Features (17 of 17 sources) See concern evidence

May 2025 to Oct 2026.

86 sources from practitioners, by month and sentiment.

Across public accounts, praise centers on the Zed Agent and on running other agents and models inside a fast editor. The most frequent concern is that Claude Code, Codex, and other ACP agents lose features inside Zed. Engineers who rely on tab completion keep going back to Cursor.

Positive, mixed and negative sources

The score counts them as 25 positive, 20 mixed and 11 negative votes, one per author. See How the numbers work.

See the evidence

Below: 5 praise themes, 6 concerns, 3 close to agreement and 2 disagreements.

Findings

Counted for All practitioners, from 86 sources. Change group

What practitioners praise

Every capability practitioners describe positively, ranked by how many sources say so. Open one to read who said it.

  1. Zed Agent and Review Flow16 of 17 sources on it praise itZed's own agent, its ask and write modes, parallel threads, and reviewing agent edits in the editor.

    Counted for All practitioners.

    Positive, 16

    1. Positive

      Eluert Mukja, full-stack developer

      Post on X by Eluert MukjaPost on X

      Runs most of their coding through agent threads in Zed and built a phone app that mirrors the agent panel.

      Published Oct 4, 2026self-described: Built a phone companion for Zed's agent panel

      Practitioner account, pseudonymous; direct use; counts toward Reliability

    2. Positive

      emerongi, engineer reviewing Zed's collaborative agent workflow

      Comment on: Zed collaborative agent workflowHacker News comment

      Sees a useful review loop in which a coworker inspects an agent thread directly and needs no explanation from its author, though they want polished final output.

      Published Sep 18, 2026self-described: Response to a video of Zed collaborative review

      Practitioner account, pseudonymous; opinion; not counted, takes no position

    3. Positive

      jackjeff, developer evaluating Zed's agent workflow

      Comment on: trying Zed againHacker News comment

      After retrying Zed, the developer found review features let them do roughly what they could in Cursor.

      Published Aug 29, 2026self-described: Retesting Zed after an earlier trial

      Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account

    4. Mixed

      slopinthebag, Hacker News commenter

      Comment on: Software is made between commitsHacker News comment

      Likes the inline assistant and an agent that stays out of the way, but dislikes the push toward an agent-first layout.

      Published Jun 11, 2026self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account

    5. Positive

      arc (arcbjorn), agent systems and ML engineer

      Post on X by arcPost on X

      Says Zed's own thread implementation and its screen for concurrent agents put it level with the best agentic editors.

      Published May 30, 2026self-described: Zed user since the 2022 private beta

      Practitioner account, pseudonymous; direct use; not counted, tied to the vendor

    6. Positive

      mlsu, Hacker News commenter

      Comment on: Zed 1.0Hacker News comment

      Likes Zed's AI-assisted setup and thinks a file review pane inside the editor is the right UX for AI tools.

      Published Apr 29, 2026self-described: Developer new to Zed on a mixed C and Rust project

      Practitioner account, pseudonymous; direct use; counts toward Adoption and upkeep

    7. Mixed

      0xSero, developer on X

      Post on X by 0xSeroPost on X

      Likes Zed for reading code with the new agent mode, but wants the ACP panel to link to files and lines and work more tightly with the codebase.

      Published Apr 26, 2026self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, takes no position

    8. Positive

      jamie_ca, Hacker News commenter

      Comment on: Parallel agents in ZedHacker News comment

      Is buying into parallel threads in worktrees and wants lifecycle hooks to set up and tear down databases per worktree.

      Published Apr 22, 2026self-described: Developer whose employer provides GitHub Copilot

      Practitioner account, pseudonymous; opinion; not counted, takes no position

    9. Mixed

      conqrr, Hacker News commenter

      Comment on: Parallel agents in ZedHacker News comment

      Loves Zed with OpenRouter at home but finds its Claude Code integration weak next to the Zed Agent with a Claude model, so uses VS Code with Claude Code at work.

      Published Apr 22, 2026self-described: Developer using Zed at home and Claude Code at work

      Practitioner account, pseudonymous; direct use; not counted, takes no position

    10. Positive

      travisgriggs, Hacker News commenter at a 2-3 person shop

      Comment on: Zed is our officeHacker News comment

      Pays for Zed Pro and likes the integrated agent after five months, while wanting less cognitive load from the editor.

      Published Nov 13, 2025self-described: A 2-3 person shop working across four operating systems and 5+ languages

      Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account

    11. Mixed

      lemontheme, Hacker News commenter

      Comment on: Claude Code: Now in Beta in ZedHacker News comment

      Was blown away by Zed's agent as a first coding agent, moved to Claude Code weeks later, and canceled the 20 EUR per month subscription for AI features no longer used.

      Published Sep 4, 2025self-described: Developer describing their own use of Zed's AI features

      not counted, repeats a stronger account

    12. Positive

      trenchpilgrim, Hacker News commenter

      Comment on: Claude Opus 4.1Hacker News comment

      Uses ask mode to discuss approaches and draft patches, then write mode to implement and run tests and diagnostics, then a line-by-line review before a pull request.

      Published Aug 5, 2025self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; counts toward Operational control

    13. Positive

      senko, Hacker News commenter

      Comment on: Coding with LLMs in the summer of 2025 (an update)Hacker News comment

      Runs each task in the agent panel: plans with Claude 4 Sonnet in thinking mode, then switches to Sonnet 4 in the same conversation to implement, then reviews and tests.

      Published Jul 21, 2025self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; counts toward Operational control

    14. Positive

      José Valim, Chief Adoption Officer at Dashbit

      The path to Tidewave: beyond code intelligenceCompany blog post, Dashbit

      Valim credits Zed with an excellent job of bringing editor diagnostics and code actions into its agentic workflows, while most agent tools still see only the code as written.

      "The Zed editor also does an excellent job at integrating some editor features, such as diagnostics and code actions, into their agentic workflows."
      Excerpt, José Valim

      Published Jun 6, 2025Introducing Tidewave, Dashbit's runtime-aware tools for coding agents

      Practitioner account, name and role on an outside page; opinion; not counted, takes no position

    15. Mixed

      Victor Mota, co-founder of CoPlane

      Post on X by Victor MotaPost on X

      After a few days with Zed's new AI features, Mota found the Agent Panel pretty nice but short on polished defaults, and edit predictions never appeared. He still wants Zed as his default but flips back to Cursor when he hits an issue.

      "Despite all the quirks, I still find myself wanting to default to it, but then flipping back to Cursor when I run into an issue."
      Excerpt, Victor Mota

      Published May 10, 2025Trying Zed's new AI features against Cursor, including on his team's Python monorepo

      Practitioner account, pseudonymous; direct use; counts toward Adoption and upkeep

    16. Positive

      DreaminDani, Hacker News commenter

      Comment on: Zed: High-performance AI Code EditorHacker News comment

      Finds the new agent panel does the right thing more often than Cursor or Windsurf. It gathers the right context, asks for feedback when needed, and has not looped between two failing fixes.

      Published May 7, 2025self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account

    Mixed or negative, 1

    1. Mixed

      gyre007, Hacker News commenter

      Comment on: Zed: High-performance AI Code EditorHacker News comment

      Reports the agent edited files when asked only a question about the codebase, and asks for diffs that wait for approval.

      Published May 7, 2025self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account

    Open them in the record
  2. Claude Code and Other Agents in the EditorAll 12 sources on it praise itRunning Claude Code, Codex, Gemini CLI, OpenCode, or another agent in Zed's panel through the Agent Client Protocol.

    Counted for All practitioners.

    1. Positive

      Jacob Wolf, leads podcasts and integrations engineering at beehiiv

      Post on X by Jacob WolfPost on X

      Uses Zed as an agent-agnostic control plane, which makes switching between OpenCode, Cursor, and Claude Code by task easy.

      Published Oct 7, 2026self-described: Engineering lead switching between OpenCode, Cursor, and Claude Code

      Practitioner account, pseudonymous; direct use; counts toward Operational control

    2. Positive

      recov, Hacker News commenter

      Comment on: Zed 1.0Hacker News comment

      Likes Claude over ACP for keeping the code on screen while chatting; terminal-only work feels like a handicap.

      Published Apr 29, 2026self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, takes no position

    3. Positive

      jeppester, Hacker News commenter

      Comment on: Zed 1.0Hacker News comment

      Uses Claude Code in Zed a lot: it lacks many features but feels less busy than the terminal, and it improves with each release.

      Published Apr 29, 2026self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account

    4. Mixed

      ksajadi, Hacker News commenter

      Comment on: Zed 1.0Hacker News comment

      Likes the ACP approach, but finds it buggy and limited: no editing of earlier messages, and work that occasionally never ends.

      Published Apr 29, 2026self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account

    5. Positive

      Stephen Haney, founder of Paper

      Post on X by Stephen HaneyPost on X

      Haney praises the new Claude Code extension in Zed and is testing Zed as his full-time editor.

      "…the new Claude Code extension in Zed is 🔥 I'm testing using Zed full time now…"
      Excerpt, Stephen Haney

      Published Jan 28, 2026Founder of Paper, a design tool company

      Practitioner account, pseudonymous; direct use; not counted, takes no position

    6. Positive

      Abishek_Muthian, Hacker News commenter

      Comment on: I switched from VSCode to ZedHacker News comment

      Moved to Zed for AI: the Claude integration works smoothly, the editor is fast, and any agent can be used when needed.

      Published Jan 6, 2026self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account

    7. Positive

      giancarlostoro, Hacker News commenter

      Comment on: I switched from VSCode to ZedHacker News comment

      Is leaving JetBrains because its bundled AI is not worth paying for beside Claude Code, and finds Zed's Claude Code integration richer than JetBrains'.

      Published Jan 5, 2026self-described: Developer leaving a JetBrains subscription after nearly 10 years

      Practitioner account, pseudonymous; direct use; not counted, takes no position

    8. Positive

      Dhravya Shah, founder of Supermemory

      Post on X by Dhravya ShahPost on X

      Shah recommends Zed with Claude Code or OpenCode as the setup to use.

      "…zed + Claude code / opencode is the move…"
      Excerpt, Dhravya Shah

      Published Jan 3, 2026Founder of Supermemory, a memory layer for AI apps

      Practitioner account, pseudonymous; opinion; not counted, takes no position

    9. No opinion stated

      José Valim, Chief Adoption Officer at Dashbit

      A new direction for AI developer tooling featuring José Valim, creator of Tidewave (Changelog & Friends #112)Podcast episode transcript, Changelog & Friends

      Valim uses Zed a lot alongside Tidewave, waiting on their notifications while agents work, and says Tidewave is building on Zed's ACP to talk to Codex, Claude Code and Gemini CLI. Thinks pushing people to build on Claude Code or Codex "like with ACP" is "not actually a bad idea".

      "I actually think pushing people towards using Claude Code more, or Codex, and building on top of those tools, like with ACP, is not actually a bad idea…"
      Excerpt, José Valim

      Published Oct 10, 2025Creator of Elixir, building Dashbit's Tidewave coding agent on ACP

      Practitioner account, name and role on an outside page; direct use; not counted, takes no position

    10. Positive

      hakanensari, Hacker News commenter

      Comment on: Zed's Pricing Has Changed: LLM Usage Is Now Token-BasedHacker News comment

      Calls delegating to CLI agents in the agent panel the right direction after a month of Claude Code in Zed, with SDK limits still to fix.

      Published Sep 24, 2025self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; opinion; not counted, repeats a stronger account

    11. Positive

      ppeetteerr, Hacker News commenter

      Comment on: Claude Code: Now in Beta in ZedHacker News comment

      Welcomes native Claude Code support but finds Cursor's autocomplete and predictions much more accurate than Zed's own model or Supermaven.

      Published Sep 3, 2025self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account

    12. Positive

      dcre, Hacker News commenter

      Comment on: Claude Code: Now in Beta in ZedHacker News comment

      Reads ACP as Zed leaving agent logic to others, since prompting tricks are easy to copy and the next model generation absorbs them.

      Published Sep 3, 2025self-described: Developer describing their own use of Zed's AI features

      not counted, repeats a stronger account

    Open them in the record
  3. Native SpeedAll 10 sources on it praise itA Rust editor that stays fast and responsive with AI features running, compared with VS Code and its forks.

    Counted for All practitioners.

    1. Positive

      Omar McAdam, CTO at Different AI

      Post on X by Omar McAdamPost on X

      McAdam switched to Zed as his IDE while OpenCode writes the code, and uses Zed to clean up code and review large files and plans, calling it clean and light.

      "…switched to Zed as my ide @opencode writes everything Zed is for cleaning up code, reviewing large files/plans…"
      Excerpt, Omar McAdam

      Published Apr 29, 2026CTO of Different AI, which builds OpenWork, with OpenCode as his coding agent

      Practitioner account, pseudonymous; direct use; not counted, takes no position

    2. Positive

      Mike McQuaid, CTPO at Administrate

      Sandboxes and Worktrees: My secure Agentic AI SetupBlog post, mikemcquaid.com

      McQuaid runs Codex and Claude Code in sandboxed worktrees and edits their output in Zed, his editor of choice, picked for startup speed because he no longer needs Cursor's better autocomplete.

      "I picked Zed because I spend less time in my editor now, don’t need Cursor’s better autocomplete and care more about startup speed…"
      Excerpt, Mike McQuaid

      Published Apr 14, 2026Runs sandboxed Codex and Claude Code agents on Homebrew and Administrate repositories

      Practitioner account, self-described; direct use; not counted, takes no position

    3. Mixed

      avilay, Hacker News commenter

      Comment on: Reallocating $100/Month Claude Code Spend to Zed and OpenRouterHacker News comment

      Found Zed more responsive than VS Code and loved the Zed Agent autocomplete, then went back to VS Code over a debugger bug.

      Published Apr 9, 2026self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes

    4. Positive

      Quinn Slack, CEO and Member of Technical Staff at Amp

      Post on X by Quinn SlackPost on X

      Calls Zed really nice, with awesome simplicity and speed and LSP servers used directly. Amp cutting editor-specific layers such as VS Code extensions led the author to Zed; setting up Amp's codebase took about 30min, and the author will be trying it out this week.

      "But we're reducing our reliance on human-only/in-editor feedback loops in our codebase, because the ones used by agents are more important."
      Excerpt, Quinn Slack

      Published Feb 10, 2026Moving Amp's own codebase toward feedback loops that coding agents use

      Practitioner account, pseudonymous; direct use; counts toward Adoption and upkeep

    5. Positive

      Abishek_Muthian, Hacker News commenter

      Comment on: I switched from VSCode to ZedHacker News comment

      Moved to Zed for AI: the Claude integration works smoothly, the editor is fast, and any agent can be used when needed.

      Published Jan 6, 2026self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account

    6. Positive

      Victor Skvortsov, author of Ten thousand meters

      I switched from VSCode to ZedBlog post, Ten thousand meters

      Left VS Code over intrusive Copilot features. Found Zed fast and responsive, with no stability problems in two weeks. Uses Codex in the CLI for AI work.

      Published Jan 3, 2026self-described: Python and Go developer who uses a CLI agent (Codex) for AI work

      Practitioner account; direct use; counts toward Reliability, Adoption and upkeep

    7. Positive

      WD-42, Hacker News commenter

      Comment on: Zed's Pricing Has Changed: LLM Usage Is Now Token-BasedHacker News comment

      Expects to save money under token pricing. Keeps paying for Zed to write code themselves; hardcore agentic editing is not the reason.

      Published Sep 24, 2025self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, takes no position

    8. Positive

      volkk, Hacker News commenter

      Comment on: Zed: High-performance AI Code EditorHacker News comment

      Rates the AI panel well below Cursor's but calls Zed one of the fastest programs on the machine; a later edit says the updated panel looks good.

      Published May 7, 2025self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, takes no position

    9. Positive

      thestephen, Hacker News commenter

      Comment on: Zed: High-performance AI Code EditorHacker News comment

      Prefers Zed because the developer keeps control over API calls, where Cursor spent the API quota unexpectedly, and because Zed matches Sublime's speed.

      Published May 7, 2025self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account

    10. Mixed

      Victor Mota, co-founder of CoPlane

      Post on X by Victor MotaPost on X

      After a few days with Zed's new AI features, Mota found the Agent Panel pretty nice but short on polished defaults, and edit predictions never appeared. He still wants Zed as his default but flips back to Cursor when he hits an issue.

      "Despite all the quirks, I still find myself wanting to default to it, but then flipping back to Cursor when I run into an issue."
      Excerpt, Victor Mota

      Published May 10, 2025Trying Zed's new AI features against Cursor, including on his team's Python monorepo

      Practitioner account, pseudonymous; direct use; counts toward Adoption and upkeep

    Open them in the record
  4. Model and Provider ChoiceAll 7 sources on it praise itBringing your own key, a local model, OpenRouter, or a company tenant, and switching models per task.

    Counted for All practitioners.

    1. Positive

      Chris Reynolds, IT Director

      Post on X by Chris ReynoldsPost on X

      Runs Zed with a local model for next-edit prediction.

      Published Sep 25, 2026self-described: IT director using Zed and Sublime as minimal editors

      Practitioner account, pseudonymous; direct use; counts toward Operational control

    2. Positive

      NortySpock, Hacker News commenter

      Comment on: Zed 1.0Hacker News comment

      Came to Zed because a local Ollama model took a dropdown to set up, then moved to OpenRouter models with a dropdown and an API key.

      Published Apr 29, 2026self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; counts toward Operational control, Adoption and upkeep

    3. Mixed

      conqrr, Hacker News commenter

      Comment on: Parallel agents in ZedHacker News comment

      Loves Zed with OpenRouter at home but finds its Claude Code integration weak next to the Zed Agent with a Claude model, so uses VS Code with Claude Code at work.

      Published Apr 22, 2026self-described: Developer using Zed at home and Claude Code at work

      Practitioner account, pseudonymous; direct use; not counted, takes no position

    4. Positive

      Ewan (kisamoto on Hacker News), author of Braw.dev

      Reallocating $100/Month Claude Code spend to Zed and OpenRouterBlog post, Braw.dev

      Pays US$10 a month for Zed and routes models through OpenRouter, because Zed's hosted token prices run above the providers' own API prices.

      Published Apr 6, 2026self-described: Developer moving a US$100 per month Claude Code budget to Zed and OpenRouter

      Practitioner account; direct use; counts toward Operational control

    5. Positive

      Octoth0rpe, Hacker News commenter

      Comment on: I switched from VSCode to ZedHacker News comment

      Trusts Zed more since its AI plans began, because Pro buys tokens across OpenAI, Anthropic, and Google with a spending cap, which looks like a sustainable business model.

      Published Jan 5, 2026self-described: Early Zed adopter

      Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account

    6. Positive

      senko, Hacker News commenter

      Comment on: Coding with LLMs in the summer of 2025 (an update)Hacker News comment

      Runs each task in the agent panel: plans with Claude 4 Sonnet in thinking mode, then switches to Sonnet 4 in the same conversation to implement, then reviews and tests.

      Published Jul 21, 2025self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; counts toward Operational control

    7. Positive

      thestephen, Hacker News commenter

      Comment on: Zed: High-performance AI Code EditorHacker News comment

      Prefers Zed because the developer keeps control over API calls, where Cursor spent the API quota unexpectedly, and because Zed matches Sublime's speed.

      Published May 7, 2025self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account

    Open them in the record
  5. AI Pricing ValueAll 6 sources on it praise itWhat Zed Pro and the free tier give for the money, including token billing with a spending cap.

    Counted for All practitioners.

    1. Positive

      Octoth0rpe, Hacker News commenter

      Comment on: I switched from VSCode to ZedHacker News comment

      Trusts Zed more since its AI plans began, because Pro buys tokens across OpenAI, Anthropic, and Google with a spending cap, which looks like a sustainable business model.

      Published Jan 5, 2026self-described: Early Zed adopter

      Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account

    2. Positive

      travisgriggs, Hacker News commenter at a 2-3 person shop

      Comment on: Zed is our officeHacker News comment

      Pays for Zed Pro and likes the integrated agent after five months, while wanting less cognitive load from the editor.

      Published Nov 13, 2025self-described: A 2-3 person shop working across four operating systems and 5+ languages

      Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account

    3. Mixed

      genshii, Hacker News commenter

      Comment on: Zed's Pricing Has Changed: LLM Usage Is Now Token-BasedHacker News comment

      Pays for Zed Pro only for edit prediction and calls it garbage next to Cursor's, though the 10 to 20% of useful suggestions still justify the subscription.

      Published Sep 24, 2025self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account

    4. Positive

      WD-42, Hacker News commenter

      Comment on: Zed's Pricing Has Changed: LLM Usage Is Now Token-BasedHacker News comment

      Expects to save money under token pricing. Keeps paying for Zed to write code themselves; hardcore agentic editing is not the reason.

      Published Sep 24, 2025self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, takes no position

    5. Positive

      pkilgore, Hacker News commenter

      Comment on: Zed's Pricing Has Changed: LLM Usage Is Now Token-BasedHacker News comment

      As a Claude Max and Zed user, sees costs go down, and wants a plan with edit prediction alone and their own key for the rest.

      Published Sep 24, 2025self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, takes no position

    6. Positive

      chillfox, Hacker News commenter

      Comment on: Claude Code: Now in Beta in ZedHacker News comment

      Estimates the free plan's 50 prompts a month at roughly US$8 of equivalent Claude API usage.

      Published Sep 4, 2025self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; counts toward Cost predictability and limits

    Open them in the record

What concerns them

Every concern practitioners raise, most raised first.

  1. External Agents Lose FeaturesAll 17 sources on it raise it as a concernClaude Code, Codex, Gemini CLI, and other ACP agents missing commands, history, hooks, subagent displays, or context controls they have in their own apps.

    Counted for All practitioners.

    1. Negative

      Nenad Novakovic, founder

      Post on X by Nenad NovakovicPost on X

      Asks why ACP output in Zed looks worse than what the Claude Code app shows.

      Published Oct 9, 2026self-described: Founder building his own apps

      Practitioner account, pseudonymous; direct use; not counted, takes no position

    2. Negative

      Andrey Tarantsov, Technical director at Bubblehouse

      Post on X by Andrey TarantsovPost on X

      Likes Zed's UI touches but says Codex and Grok Build over ACP still fall short: no UI for subagent threads and full file dumps in the agent panel.

      Published Sep 18, 2026self-described: Technical director trying Codex and Grok Build as external agents

      Practitioner account, pseudonymous; direct use; counts toward Reliability, Adoption and upkeep

    3. Mixed

      faisal-fida, GitHub issue author

      Cannot resume/open an existing Claude Code session in Zed when it is running as a background agent (works in VS Code)GitHub issue

      A Claude Code session running as a background agent fails to open in Zed with a raw internal error, while the same session opens in VS Code.

      Published Aug 11, 2026self-described: Claude Code over SSH remote development

      Practitioner account, pseudonymous; structured test; counts toward Reliability

    4. Mixed

      xbt573, GitHub issue author

      External agents are unavailable in shared projects while in CollaborationGitHub issue

      In a shared collaboration project, only the Zed Agent is offered; Claude Code and the terminal agent are missing.

      Published Jul 20, 2026self-described: Using a shared project in a Zed collaboration channel

      Practitioner account, pseudonymous; direct use; counts toward Operational control

    5. Mixed

      jordandakota, GitHub issue author

      Claude Code ACP session enters unrecoverable state after hitting usage limitGitHub issue

      After a Claude usage limit, the ACP session stays alive but silently drops prompts until the thread or Zed restarts.

      Published May 2, 2026self-described: Claude Code over ACP on a Claude plan with usage limits

      Practitioner account, pseudonymous; direct use; counts toward Reliability

    6. Mixed

      presentation, Hacker News commenter

      Comment on: Zed 1.0Hacker News comment

      Ran the Cursor agent through ACP and found it markedly worse than in Cursor itself, with adding context and reviewing changes clumsier too.

      Published Apr 30, 2026self-described: Developer whose company pays for Cursor

      Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes, Reliability, Adoption and upkeep

    7. Mixed

      sodacanner, Hacker News commenter

      Comment on: Zed 1.0Hacker News comment

      Claude Code in Zed works but lacks sub-agent display, context clearing, a usage display, and effort control, so they went back to the terminal.

      Published Apr 29, 2026self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; counts toward Operational control

    8. Mixed

      JnnydevDude, Hacker News commenter

      Comment on: Zed 1.0Hacker News comment

      Enjoys Zed, and the one sore point is an integration with Claude Code that is not smooth.

      Published Apr 29, 2026self-described: Light Zed user for a few months

      Practitioner account, pseudonymous; direct use; not counted, takes no position

    9. Positive

      jeppester, Hacker News commenter

      Comment on: Zed 1.0Hacker News comment

      Uses Claude Code in Zed a lot: it lacks many features but feels less busy than the terminal, and it improves with each release.

      Published Apr 29, 2026self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account

    10. Mixed

      ksajadi, Hacker News commenter

      Comment on: Zed 1.0Hacker News comment

      Likes the ACP approach, but finds it buggy and limited: no editing of earlier messages, and work that occasionally never ends.

      Published Apr 29, 2026self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account

    11. Mixed

      philipp-gayret, Hacker News commenter

      Comment on: Reallocating $100/Month Claude Code Spend to Zed and OpenRouterHacker News comment

      Uses Zed but warns that Claude Code hooks do not work in Zed's integration, so heavy hook users should stay in the terminal.

      Published Apr 9, 2026self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, takes no position

    12. Mixed

      lukaslalinsky, Hacker News commenter

      Comment on: Codex Is Live in ZedHacker News comment

      A Zed user who says Claude Code in the panel is not really Claude Code: ACP reduces the agents to shared functionality, with very limited context management.

      Published Oct 18, 2025self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account

    13. Negative

      mcjiggerlog, Hacker News commenter

      Comment on: Codex Is Live in ZedHacker News comment

      Found Codex in Zed slow and unexplained on day one, Claude Code in Zed usable with rough edges, and Cursor still the nicer GUI agent.

      Published Oct 17, 2025self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account

    14. Mixed

      agrippanux, Hacker News commenter

      Comment on: Zed's Pricing Has Changed: LLM Usage Is Now Token-BasedHacker News comment

      Finds burn mode strong but costly, now runs Claude Code with Neovim, and expects Zed's Claude integration to trail the latest Claude Code.

      Published Sep 24, 2025self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; opinion; not counted, takes no position

    15. Mixed

      Alek Dobrohotov, Product Manager at C&C

      Zed + Claude Code: A Beautiful Promise That's Not Quite There YetBlog post, The Vibe Coding Post

      Praises Zed's speed and diff review, but after 30 days of standalone Claude Code finds the beta integration a step back, without Plan Mode, /compact, or subagents.

      Published Sep 6, 2025self-described: Builds internal tools for retail, education, and B2B teams at C&C, an Apple Premium Partner

      Practitioner account, self-described; direct use; counts toward Delivered outcomes, Reliability, Operational control

    16. Mixed

      aexvir, GitHub issue author

      Support history with external ACP agentsGitHub issue

      Using Gemini via ACP in Zed v0.202.1 ("Zed Preview"), none of the threads created with the Gemini agent were saved to the history list of threads.

      Published Aug 28, 2025self-described: Using Gemini CLI through ACP on a pre-release Zed build

      Practitioner account, pseudonymous; direct use; counts toward Reliability

    17. Mixed

      conqrr, Hacker News commenter

      Comment on: Parallel agents in ZedHacker News comment

      Loves Zed with OpenRouter at home but finds its Claude Code integration weak next to the Zed Agent with a Claude model, so uses VS Code with Claude Code at work.

      Published Apr 22, 2026self-described: Developer using Zed at home and Claude Code at work

      Practitioner account, pseudonymous; direct use; not counted, takes no position

    GILD's fact checks, which don't count as sources

    Checked Oct 25, 2026.Zed's documentation states that some Agent Panel features, including history, checkpoints, and token usage display, depend on each external agent's integration. Agent Panel

    Open them in the record
  2. Crashes, Freezes, and SlowdownsAll 9 sources on it raise it as a concernPanics, frozen threads, slow agent runs, and high CPU, memory, or energy use while agents run.

    Counted for All practitioners.

    1. Mixed

      Kang Cahya Keren, developer and blogger

      Post on X by Kang Cahya KerenPost on X

      Zed frequently freezes when they use ChatGPT models through the Zed Agent.

      Published Oct 16, 2026self-described: Moved to Zed from VS Code and Antigravity

      Practitioner account, pseudonymous; direct use; counts toward Reliability

    2. Mixed

      jordandakota, GitHub issue author

      Claude Code ACP session enters unrecoverable state after hitting usage limitGitHub issue

      After a Claude usage limit, the ACP session stays alive but silently drops prompts until the thread or Zed restarts.

      Published May 2, 2026self-described: Claude Code over ACP on a Claude plan with usage limits

      Practitioner account, pseudonymous; direct use; counts toward Reliability

    3. Mixed

      secondl1ght, GitHub issue author

      All agent actions extremely slowGitHub issue

      Streaming, tool calls, and code generation in the Zed Agent frequently slow to a crawl, where the agent is normally very fast.

      Published Apr 3, 2026self-described: Zed Agent on Linux

      Practitioner account, pseudonymous; direct use; counts toward Reliability

    4. Mixed

      princhman, GitHub issue author

      Claude Code ACP increases zed energy consumption 100xGitHub issue

      Starting a Claude ACP thread made Zed's energy use jump about 100 times in Activity Monitor.

      Published Mar 15, 2026self-described: Running the Claude agent over ACP on macOS

      Practitioner account, pseudonymous; direct use; not counted, takes no position

    5. Mixed

      somtooo, GitHub issue author

      GPT-5.3 Codex from copilot provider is unusable in zedGitHub issue

      Reports that GPT-5.3 Codex through the GitHub Copilot provider is unusable in Zed 0.223.5: tool calls often fail, it says it edited files without editing anything, and it stops before doing the work. Other Codex models work well in Zed.

      Published Mar 3, 2026self-described: Using GPT-5.3 Codex from a GitHub Copilot subscription in the Zed Agent

      Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes, Reliability

    6. Negative

      trcf23, Hacker News commenter

      Comment on: Zed editor switching graphics lib from blade to wgpuHacker News comment

      Sees the agent panel panic in longer conversations or stall without proceeding.

      Published Feb 13, 2026self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account

    7. Positive

      RamblingCTO, Hacker News commenter

      Comment on: I switched from VSCode to ZedHacker News comment

      Happy after switching, though Zed sometimes takes over the CPU and crashed once, which they attribute to Claude Code.

      Published Jan 5, 2026self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account

    8. Negative

      kelthuzad, Hacker News commenter

      Comment on: Claude Code: Now in Beta in ZedHacker News comment

      Never had a problem with Cursor, while Zed panicked, crashed, and behaved inconsistently often.

      Published Sep 3, 2025self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; counts toward Operational control

    9. Mixed

      ksajadi, Hacker News commenter

      Comment on: Zed 1.0Hacker News comment

      Likes the ACP approach, but finds it buggy and limited: no editing of earlier messages, and work that occasionally never ends.

      Published Apr 29, 2026self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account

    Open them in the record
  3. Agent Panel and LayoutAll 7 sources on it raise it as a concernConfusing agent controls, an agent-first layout that crowds out code, or an agent that acts before it is asked.

    Counted for All practitioners.

    1. Mixed

      scottmessinger, Hacker News commenter

      Comment on: Zed 1.0Hacker News comment

      Is excited about multi-agent support but finds it confusing which of the agent sidebar and the titlebar controls the worktree and branch.

      Published Apr 29, 2026self-described: Zed user for years

      Practitioner account, pseudonymous; direct use; not counted, takes no position

    2. Mixed

      0xSero, developer on X

      Post on X by 0xSeroPost on X

      Likes Zed for reading code with the new agent mode, but wants the ACP panel to link to files and lines and work more tightly with the codebase.

      Published Apr 26, 2026self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, takes no position

    3. Negative

      tankmohit11, Hacker News commenter

      Comment on: Parallel agents in ZedHacker News comment

      Uninstalled Zed over the agent UI: tiny, vague icons and an x that closes the editor, an agent, or a panel unpredictably.

      Published Apr 23, 2026self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; counts toward Operational control

    4. Mixed

      verse, Hacker News commenter

      Comment on: Parallel agents in ZedHacker News comment

      Uses Zed daily but dislikes a default layout that pushes code and the file tree aside for AI tools, and expects it to turn new users away.

      Published Apr 22, 2026self-described: Daily Zed user

      Practitioner account, pseudonymous; direct use; not counted, takes no position

    5. Mixed

      Victor Mota, co-founder of CoPlane

      Post on X by Victor MotaPost on X

      After a few days with Zed's new AI features, Mota found the Agent Panel pretty nice but short on polished defaults, and edit predictions never appeared. He still wants Zed as his default but flips back to Cursor when he hits an issue.

      "Despite all the quirks, I still find myself wanting to default to it, but then flipping back to Cursor when I run into an issue."
      Excerpt, Victor Mota

      Published May 10, 2025Trying Zed's new AI features against Cursor, including on his team's Python monorepo

      Practitioner account, pseudonymous; direct use; counts toward Adoption and upkeep

    6. Positive

      volkk, Hacker News commenter

      Comment on: Zed: High-performance AI Code EditorHacker News comment

      Rates the AI panel well below Cursor's but calls Zed one of the fastest programs on the machine; a later edit says the updated panel looks good.

      Published May 7, 2025self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, takes no position

    7. Mixed

      gyre007, Hacker News commenter

      Comment on: Zed: High-performance AI Code EditorHacker News comment

      Reports the agent edited files when asked only a question about the codebase, and asks for diffs that wait for approval.

      Published May 7, 2025self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account

    Open them in the record
  4. Token Markup and Subscription ValueAll 6 sources on it raise it as a concernHosted tokens billed above the providers' own prices, per-token costs that are hard to work out, or a subscription paid for AI the developer stopped using.

    Counted for All practitioners.

    1. No opinion stated

      NortySpock, developer evaluating Zed's default prompt

      Comment on: Zed prompt overheadHacker News comment

      Has looked at Zed editor's default prompt and thought "I hope this is worth the 7k tokens it costs me every session", and suggests "perhaps" the takeaway is to beware overly long prompts and keep the system prompt relevant to the field of work.

      Published Sep 17, 2026self-described: Zed editor prompt overhead

      Practitioner account, pseudonymous; direct use; not counted, takes no position

    2. No opinion stated

      kisamoto (Ewan), author of the Braw.dev post under discussion

      Comment on: Reallocating $100/Month Claude Code Spend to Zed and OpenRouterHacker News comment

      Says that "sadly Zed seems to add 10%", so it is still more worthwhile to use OpenRouter, and that it is "perhaps" more worthwhile to stay solely on Cursor with the limited models.

      Published Apr 9, 2026self-described: Developer moving a US$100 per month Claude Code budget to Zed and OpenRouter

      Practitioner account, pseudonymous; opinion; not counted, repeats a stronger account

    3. Positive

      Ewan (kisamoto on Hacker News), author of Braw.dev

      Reallocating $100/Month Claude Code spend to Zed and OpenRouterBlog post, Braw.dev

      Pays US$10 a month for Zed and routes models through OpenRouter, because Zed's hosted token prices run above the providers' own API prices.

      Published Apr 6, 2026self-described: Developer moving a US$100 per month Claude Code budget to Zed and OpenRouter

      Practitioner account; direct use; counts toward Operational control

    4. Mixed

      agrippanux, Hacker News commenter

      Comment on: Zed's Pricing Has Changed: LLM Usage Is Now Token-BasedHacker News comment

      Finds burn mode strong but costly, now runs Claude Code with Neovim, and expects Zed's Claude integration to trail the latest Claude Code.

      Published Sep 24, 2025self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; opinion; not counted, takes no position

    5. Mixed

      oakesm9, Hacker News commenter

      Comment on: Zed's Pricing Has Changed: LLM Usage Is Now Token-BasedHacker News comment

      Accepts the pricing change but says the announcement did not say what the API list prices behind list price plus 10% are.

      Published Sep 24, 2025self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; opinion; not counted, repeats a stronger account

    6. Mixed

      lemontheme, Hacker News commenter

      Comment on: Claude Code: Now in Beta in ZedHacker News comment

      Was blown away by Zed's agent as a first coding agent, moved to Claude Code weeks later, and canceled the 20 EUR per month subscription for AI features no longer used.

      Published Sep 4, 2025self-described: Developer describing their own use of Zed's AI features

      not counted, repeats a stronger account

    GILD's fact checks, which don't count as sources

    Checked Oct 25, 2026.Zed bills hosted model usage at API list price plus 10% after Pro's included US$5, with an optional spending cap. Your own keys and external agents bill through their providers. Zed pricing

    Open them in the record
  5. Edit Predictions Trail Cursor Tab6 of 8 sources on it raise it as a concernZeta and other edit prediction providers in Zed suggesting less useful or less far-reaching edits than Cursor's tab completion.

    Counted for All practitioners.

    Raised as a concern, 6

    1. Negative

      zephod, Hacker News commenter

      Comment on: Zed 1.0Hacker News comment

      Loved Zed for a few days but went back to Cursor because they depend on Cursor Tab; Zed AI, Copilot, and Mercury all felt forgetful beside it.

      Published Apr 30, 2026self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account

    2. Negative

      matthewtse, Hacker News commenter

      Comment on: Zed 1.0Hacker News comment

      Found Zed's tab completion unusable next to Cursor's and will try again only if it improves.

      Published Apr 30, 2026self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, takes no position

    3. Negative

      mosselman, Hacker News commenter

      Comment on: Codex Is Live in ZedHacker News comment

      Calls Zed's inline suggestions far behind Cursor's, which follow a file rename into the module name and suggest edits across the file.

      Published Oct 16, 2025self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account

    4. Mixed

      genshii, Hacker News commenter

      Comment on: Zed's Pricing Has Changed: LLM Usage Is Now Token-BasedHacker News comment

      Pays for Zed Pro only for edit prediction and calls it garbage next to Cursor's, though the 10 to 20% of useful suggestions still justify the subscription.

      Published Sep 24, 2025self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account

    5. Positive

      ppeetteerr, Hacker News commenter

      Comment on: Claude Code: Now in Beta in ZedHacker News comment

      Welcomes native Claude Code support but finds Cursor's autocomplete and predictions much more accurate than Zed's own model or Supermaven.

      Published Sep 3, 2025self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account

    6. Negative

      gwking, Hacker News commenter

      Comment on: You can now disable all AI features in ZedHacker News comment

      Used Zed daily for over a month and went back to Cursor for the quality of its tab completion, which Zeta and Copilot did not match.

      Published Jul 23, 2025self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account

    Discussed without concern, 2

    1. Positive

      Aris (riipandi), developer on X

      Post on X by ArisPost on X

      Wants to turn off every AI feature except edit prediction, which they really like.

      Published Sep 25, 2026self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, from a competitor's side

    2. Mixed

      avilay, Hacker News commenter

      Comment on: Reallocating $100/Month Claude Code Spend to Zed and OpenRouterHacker News comment

      Found Zed more responsive than VS Code and loved the Zed Agent autocomplete, then went back to VS Code over a debugger bug.

      Published Apr 9, 2026self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes

    GILD's fact checks, which don't count as sources

    Checked Oct 25, 2026.Zeta is included on Pro, the Pro trial, and the Student and Business plans. Copilot, Codestral, Mercury Coder, or a local model can serve predictions instead. Edit Prediction

    Open them in the record
  6. AI Depth Behind CursorAll 5 sources on it raise it as a concernZed's AI features trailing Cursor's agents, background agents, and developer experience.

    Counted for All practitioners.

    1. Mixed

      presentation, Hacker News commenter

      Comment on: Zed 1.0Hacker News comment

      Ran the Cursor agent through ACP and found it markedly worse than in Cursor itself, with adding context and reviewing changes clumsier too.

      Published Apr 30, 2026self-described: Developer whose company pays for Cursor

      Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes, Reliability, Adoption and upkeep

    2. Negative

      dmix, Hacker News commenter

      Comment on: I switched from VSCode to ZedHacker News comment

      Tried to move from Cursor and found Zed's LLM integration early, with Cursor far ahead on developer experience and fairly priced models.

      Published Jan 5, 2026self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, takes no position

    3. Negative

      mcjiggerlog, Hacker News commenter

      Comment on: Codex Is Live in ZedHacker News comment

      Found Codex in Zed slow and unexplained on day one, Claude Code in Zed usable with rough edges, and Cursor still the nicer GUI agent.

      Published Oct 17, 2025self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account

    4. Negative

      kbdiaz, Hacker News commenter

      Comment on: You can now disable all AI features in ZedHacker News comment

      Calls Zed a great code editor but a weak AI editor: late to edit predictions, behind on agent UX, and without background agents like Cursor's.

      Published Jul 23, 2025self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; opinion; not counted, takes no position

    5. Positive

      volkk, Hacker News commenter

      Comment on: Zed: High-performance AI Code EditorHacker News comment

      Rates the AI panel well below Cursor's but calls Zed one of the fastest programs on the machine; a later edit says the updated panel looks good.

      Published May 7, 2025self-described: Developer describing their own use of Zed's AI features

      Practitioner account, pseudonymous; direct use; not counted, takes no position

    Open them in the record

Where the evidence agrees

Statements this group's sources agree on, then the ones close to agreement.

Close to agreement: each needs 18 supporting sources

  1. Below threshold

    Claude Code, Codex, and other agents run in Zed's panel with fewer features than in their own terminal or app.

    16 supporting sources, 1 contradicting
  2. Below threshold

    Developers choose Zed first for its speed and editing feel, and treat its AI features as an addition to that editor.

    9 supporting sources
  3. Below threshold

    Developers who depend on tab completion find Cursor's ahead of Zed's edit predictions, and some return to Cursor for it.

    7 supporting sources, 2 contradicting

Where it disagrees

Questions the sources split on, with the count on each side.

  • Evidence remains mixed

    Should engineers work with the Zed Agent or bring an external agent?

    • Zed Agent, 6 sources

      Some prefer Zed's own agent with a chosen model, for its review flow and its tighter fit with the editor.

    • External agent, 21 sources

      Others run Claude Code, Codex, or another agent, in Zed's panel or in a terminal, and treat Zed as the place to read and edit code.

  • Evidence remains mixed

    Should AI in Zed run on Zed's hosted plan or on your own key?

    • Zed's hosted plan, 5 sources

      Some pay for Zed Pro for edit predictions and hosted tokens, and see it as fair value that also funds the editor.

    • Own key or subscription, 8 sources

      Others route models through their own provider, OpenRouter, a company tenant, or an agent subscription, to avoid Zed's markup.

The evidence

Whose experience this is

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 11 accounts; GILD prints their sentiment from 20.

  • Engineering leaders and teams

    18 sources. These accounts emphasize bring-your-own-key buying, which external agents a team may run, and how far the Zed Agent and ACP agents match the tools engineers already use.

  • Individual builders

    68 sources. These accounts emphasize edit prediction quality, the Zed Agent, Claude Code and Codex over ACP, model choice, and what AI in Zed costs one developer.

  • All practitioners

    86 sources. These figures count every account: individual developers and the team and leadership accounts together.

86 sources from practitioners, by the context they work in. Filled rows count for this group.

  1. Individual practitionerSolo developers and independent builders68 sources, counted here

    Emphasizes Edit predictions, the Zed Agent, and Claude Code over ACP in one developer's day

  2. Very small team2 to 10 engineers, or a team of unstated size1 sources, counted here

    Emphasizes Whether Zed Pro and the integrated agent earn their place in a small shop

  3. Professional team11 to 50 engineers, or a named company's team1 sources, counted here

    Emphasizes Whether Claude Code in Zed matches the standalone tool for production work

  4. Technical leadershipLeaders setting adoption across teams16 sources, counted here

    Emphasizes Own-key buying and how external agents behave in the panel

  5. Organization-wideEnterprise or company-wide rolloutNo sources from this setting yet

What changes with scale. Individual developers argue about edit predictions and which agent to run. Leaders and named-company accounts turn to buying and integration: whether model traffic can stay on the company's own key, and whether Claude Code and Codex keep their features inside Zed.

Where the sources come from

Where this group's sources were published.

Counted for All practitioners, from 86 sources. Change group

95 sources for this group, 9 of them documentation

  • Developer Communities82, 86% of 95 sources
  • Product Documentation9, 9% of 95 sources
  • Engineering Blogs & Reviews3, 3% of 95 sources
  • Talks & Podcasts1, 1% of 95 sources

Older evidence is retained when it still applies. Accounts of the launch-day agent panel, the Claude Code and Codex betas, and closed issues stay in the history, marked Needs a recheck.

See sources by origin

The record

Who 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 95 sources

About this read

What changed, and how we count

What changed

  • Nov 1, 2026: by the owner's decision, a record whose own words state no overall view of Zed now casts no vote; it still counts toward its topics and is listed, and 23 records that counted leave the vote this way. Leaders and teams stay unscored, individual builders move from +27 to +22 and all practitioners from +29 to +25.
  • Oct 31, 2026: rewrote 9 summaries that two blind coders found overstated their sources, so each says what its source says. Five comments state no view of Zed overall and now carry no sentiment, and two no longer praise model choice. GitHub issues still count as mixed, as the method says. Individual builders move from +26 to +27 and all practitioners from +41 to +39.
  • Builds the Zed evidence base from public pages: each account names its author and publication date and is coded from an exact quote on the page.
  • Shows a finding only when at least 5 sources from 3 different authors discuss it, and a disagreement only with at least 2 sources on each side.
  • Counts the Braw.dev post and its author's Hacker News reply as one author.
  • Oct 26, 2026: adds named CTOs and founders at Icehouse Ventures, Different AI, Administrate, Dashbit, Paper, Supermemory, Amp and CoPlane, each role as stated on the person's own or their company's site.

Method

A running review of what engineering teams report about Zed, from individual builders to engineering leaders. Each finding links to its sources, and each conflict shows both sides.

Counts cover only the sources GILD reviewed.

Who counts

Engineering leaders and teams
Very small team, Professional team, Technical leadership, Organization-wide
Individual builders
Individual practitioner
All practitioners
Everyone in the leaders and teams and individual builders groups

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.

How the numbers work

  • Overall signal. Positive votes minus negative ones, per 100 votes from the sources that take a side. Each author counts once, with their latest view, whoever they are and wherever they posted. The older weighted reading, where accounts GILD matched to a named person count three times when GILD lists the author among engineering leaders and twice otherwise, is shown beside it as a second view. Where an author wrote more than one source, the line beside the split gives the votes, one per author, so the score follows from it. When one author's sources from the same day disagree, that author counts as mixed, and a half point rounds away from zero. On the AI coding agent hubs, each account's overall stance comes from GILD's blind coding: two models from different families read the source's own words without seeing each other's codes, and a third settled where they differed. An account that states no overall view there, such as a bug report, casts no vote and still counts toward its topics. Other hubs count the stance stored on each record. Sources with no opinion are listed and not counted. Strong positive is +50 or higher, positive +20 to +49, mixed −19 to +19, negative −20 to −49, strong negative −50 or lower.
  • Weighted reading, second view. Counted the older way, where an account GILD matched to a named person counts two or three times, all practitioners read +39 positive (method sentiment/S1). GILD printed that number as "GILD score" until October 29, 2026; most of those matches rest on a name or a self-described role.
  • Method change, Oct 30, 2026. GILD changed how practitioner sentiment counts, to one vote per author. Weighted, it read +41 Positive; with one vote per author it read +28 Positive. No evidence changed.
  • Correction, Nov 1, 2026. GILD now takes each account's overall stance from its blind two-model coding, and accounts that state no overall view cast no vote. It read +29 Positive and now reads +25 Positive.
  • Unscored. A group gets a score only with 20 sources from 15 authors. Below that, one source moves the score about 5 points, so the read says "Unscored until 20 sources from 15 authors" and shows its counts.
  • Independence. A source whose author works for the vendor, for one of its partners or programs, or for a competitor, and a page the vendor published or ran, is listed with that tie and counts toward nothing. One author counts on at most two records per product. The line under the score says how many authors and how many separate discussions the counted sources come from: comments in one thread are one discussion.
  • Praise and concerns. How many of the sources that discuss a theme praise it or raise it.
  • Consensus. A statement joins when 20% of the group's sources support it and fewer contradict it. Contradicting sources are counted separately. Below that line a statement shows as building. At 40% or more it reads Strong agreement; below, Recurring point.
  • Disagreements. The count on each side. GILD names no winner and gives no share of all users.
  • Minimum evidence. A finding needs 5 sources from 3 authors; a disagreement also needs 2 on each side. Smaller counts stay in the table below, marked under the minimum.

Labels on a source

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.

Documented
Stated in official documentation.
Reported
A practitioner's account of their own use.
Measured
Backed by the source's own measurements, such as a comparative test or tracked team metrics.
GILD analysis
GILD's check of a claim against official documentation.
Current
Describes the product as it is now.
Still applicable
Older, and the behavior or lesson still holds.
Needs a recheck
May depend on a plan, version or setting that changed since it was written. It keeps counting until GILD rechecks it, since the rest of the account usually holds, and its card says so.
Out of date
No longer accurate. Kept in the record; counts toward nothing current.

Flagged for a recheck, of the sources that take a side: leaders and teams 2 of 11, individual builders 2 of 46, all practitioners 4 of 57.

Limits

  • These figures count 86 public accounts: 55 Hacker News comments, 7 issues on zed-industries/zed, 6 blog posts, 17 posts on X, and a podcast transcript, plus 9 Zed documentation pages. Hacker News comments come from the threads under Zed's own AI announcements, which draws in people already weighing Zed against Cursor and Claude Code. GitHub issues are bug reports and count as mixed. Reddit refused access to the threads GILD found, so none is in this record. Team and leadership accounts come mostly from named CTOs and founders whose roles appear on their own or their company's pages. None describes an organization-wide rollout. This is not a representative survey.
  • The record runs from the agent panel launch in May 2025 through Zed 1.0 in April 2026 and the 1.x releases after it. Accounts tied to a launch-day build or a closed issue are marked Needs a recheck.

Every count, and the sources behind it

18 findings for 3 groups

Each row opens its sources for the group you are reading.

FindingEngineering leaders and teamsIndividual buildersAll practitionersSources
Current readCurrent sentiment7 positive, 2 mixed, 2 negative19 positive, 18 mixed, 9 negative26 positive, 20 mixed, 11 negativeSources
PraiseZed Agent and Review Flow3 of 3 positive (under the minimum)13 of 14 positive16 of 17 positiveSources
PraiseClaude Code and Other Agents in the Editor4 of 4 positive (under the minimum)8 of 8 positive12 of 12 positiveSources
PraiseModel and Provider Choice1 of 1 positive (under the minimum)6 of 6 positive7 of 7 positiveSources
PraiseNative Speed4 of 4 positive (under the minimum)6 of 6 positive10 of 10 positiveSources
PraiseAI Pricing Value1 of 1 positive (under the minimum)5 of 5 positive6 of 6 positiveSources
ConcernExternal Agents Lose Features3 of 3 raise it (under the minimum)14 of 14 raise it17 of 17 raise itSources
ConcernCrashes, Freezes, and Slowdowns0 of 0 raise it (under the minimum)9 of 9 raise it9 of 9 raise itSources
ConcernEdit Predictions Trail Cursor Tab0 of 0 raise it (under the minimum)6 of 8 raise it6 of 8 raise itSources
ConcernAgent Panel and Layout1 of 1 raise it (under the minimum)6 of 6 raise it7 of 7 raise itSources
ConcernAI Depth Behind Cursor0 of 0 raise it (under the minimum)5 of 5 raise it5 of 5 raise itSources
ConcernToken Markup and Subscription Value0 of 0 raise it (under the minimum)6 of 6 raise it6 of 6 raise itSources
ConsensusExternal agents run thinner in Zed3 supporting (under the minimum)13 supporting, 1 contradicting (below threshold)16 supporting, 1 contradicting (below threshold)Sources
ConsensusCursor Tab keeps completion users1 supporting (under the minimum)6 supporting, 2 contradicting (below threshold)7 supporting, 2 contradicting (below threshold)Sources
ConsensusSpeed comes first4 supporting (under the minimum)5 supporting (below threshold)9 supporting (below threshold)Sources
DisagreementShould engineers work with the Zed Agent or bring an external agent?Zed Agent 0, External agent 9 (under the minimum)Zed Agent 6, External agent 12Zed Agent 6, External agent 21Sources
DisagreementShould AI in Zed run on Zed's hosted plan or on your own key?Zed's hosted plan 1, Own key or subscription 1 (under the minimum)Zed's hosted plan 4, Own key or subscription 7Zed's hosted plan 5, Own key or subscription 8Sources
Current readEvidence mix27 sources77 sources95 sourcesSources

Fact checks

GILD's checks of claims against official documentation, kept apart from the source counts.

  • External Agents Lose Features. Zed's documentation states that some Agent Panel features, including history, checkpoints, and token usage display, depend on each external agent's integration. Agent Panel, checked Oct 25, 2026
  • Edit Predictions Trail Cursor Tab. Zeta is included on Pro, the Pro trial, and the Student and Business plans. Copilot, Codestral, Mercury Coder, or a local model can serve predictions instead. Edit Prediction, checked Oct 25, 2026
  • Token Markup and Subscription Value. Zed bills hosted model usage at API list price plus 10% after Pro's included US$5, with an optional spending cap. Your own keys and external agents bill through their providers. Zed pricing, checked Oct 25, 2026