What engineering leaders and their teams report about Aider, with every finding tied to its evidence.
Read the findingsPractitioner sentiment, all practitioners
+28 Positive from 59 sources that take a side
How this figure is madeFrom 57 authors in 51 discussions. Use a task pilot to measure accepted work and review time.
Jan 2025 to Oct 2026.
85 sources from practitioners, by month and sentiment.
Across public accounts the read is positive: model choice and explicit context control are the most praised capabilities. Practitioners split almost evenly on whether Aider's developer-driven control beats agentic tools that explore and run commands on their own, and the slower release pace is the newest concern.
Positive, mixed and negative sources
The score counts them as 28 positive, 17 mixed and 12 negative votes, one per author. See How the numbers work.
See the evidenceBelow: 6 praise themes, 5 concerns, 2 close to agreement and 3 disagreements.
Counted for All practitioners, from 85 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
Anurag Singh, Technology journalist at XDA Developers
I stopped forcing every coding job through Claude and started using this insteadIndependent review, XDA Developers
Moved routine work from Claude Code to Aider to pick a model per task, cheap DeepSeek or stronger Claude, Gemini or GPT; Claude Code stays ahead on autonomy.
Practitioner account, self-described; direct use; counts toward Operational control, Adoption and upkeep
Positive
Magneticdud, Aider user commenting on a GitHub issue
Comment on: No release notes for v0.86.2GitHub issue comment
Says this latest release works great with gpt-5.3-codex and thanks everyone who improved it, preferring Aider to the full-blown VS forks.
Practitioner account, pseudonymous; direct use; not counted, takes no position
Positive
roebi, Aider user commenting on a GitHub issue
Comment on: Thoughts and questions about the future direction of aiderGitHub issue comment
A user since before Claude Code still values Aider's terminal-first, git-native, model-agnostic design, and built a package to add agent skills since Aider has no plugins.
Practitioner account, pseudonymous; direct use; counts toward Operational control, Adoption and upkeep
Positive
faangguyindia, Hacker News commenter
Comment on: Claude Sonnet 4 now supports 1M tokens of contextHacker News comment
Switches to DeepSeek R1 as architect and Qwen3 Coder as editor in Aider when Gemini 2.5 Pro gets stuck; reasoning models plan well but write weaker code.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
msgodel, Hacker News commenter
Comment on: Andrej Karpathy's talk on the future of the industryHacker News comment
Runs Aider on a local Qwen3 8B through llama.cpp and finds it better than a RAG tool of the commenter's own; a 0.5B model is not enough for Aider.
Practitioner account, pseudonymous; direct use; counts toward Operational control, Adoption and upkeep
Positive
bird0861, Hacker News commenter
Comment on: Ask HN: Anyone struggling to get value out of coding LLMs?Hacker News comment
Calls Aider a daily essential and hops between hosted models such as Sonnet 3.5 and local code models on a personal GPU.
Practitioner account, pseudonymous; direct use; counts toward Operational control
Positive
piokoch, Hacker News commenter, self-described
Comment on: Firebase StudioHacker News comment
Pairs Aider with whichever model works best that month and names the files in context, which costs less and avoids confusing the model.
Practitioner account, pseudonymous; direct use; counts toward Operational control
Mixed
regularfry, Hacker News commenter
Comment on: Llama.cpp AI Performance with the GeForce RTX 5090 ReviewHacker News comment
Runs Aider on local Ollama models in architect mode, a reasoning architect plus a coder; local reasoning models do not always stick to the edit format.
Practitioner account, pseudonymous; direct use; counts toward Operational control
Positive
inciampati, Hacker News commenter
Comment on: I've been using Claude Code for a couple of daysHacker News comment
Calls Aider Claude Code's open-source inspiration; it is model agnostic, switching providers is trivial, and inciampati now rarely opens an editor.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
troyvit, Hacker News commenter
Comment on: Mistral releases Devstral2 and Mistral Vibe CLIHacker News comment
Says Aider works as a chat interface or from the editor by watching your files, and lets you tier models so a big model does the thinking while code reviews run on a smaller one. Last they looked, the maintainer had to focus elsewhere, and aider-ce is a fantastic fork.
Practitioner account, pseudonymous; direct use; counts toward Adoption and upkeep
Positive
zmmmmm, Hacker News commenter, self-described
Comment on: Mistral releases Devstral2 and Mistral Vibe CLIHacker News comment
Prefers open-source tools that work with any model; Aider is not agentic enough for many people, which zmmmmm counts as a benefit.
Practitioner account, pseudonymous; opinion; not counted, repeats a stronger account
Positive
lreeves, Hacker News commenter
Comment on: Ask HN: Who uses open LLMs and coding assistants locally? Share setup and laptopHacker News comment
The best local setup lreeves has found is Aider with gpt-oss-120b, yet hosted GPT-5 and Claude beat anything run locally on quality, speed, and price.
Practitioner account, pseudonymous; direct use; counts toward Operational control
Positive
flyinglizard, Hacker News commenter, self-described
Comment on: Claude Code 2.0Hacker News comment
Stays with Aider for its fit and model switching, but wishes a larger team maintained it, and is not comfortable letting agents loose on production code.
Practitioner account, pseudonymous; direct use; counts toward Operational control
Mixed
Calvin Hendryx-Parker, CTO at Six Feet Up
AI Assistants and uv: Which Dev Tools Deliver?Company news post, Six Feet Up
Six Feet Up's CTO gave five assistants the same Django refactor. Aider offered fine-grained git control and a choice of models but needed more coaching to finish, while Claude Code made the most complete refactor.
"Aider offered fine-grained Git control and multi-model customization, but needed more coaching to complete tasks."
Practitioner account, self-described; second-hand; not counted until substantiated
Positive
Andrew Keenan Richardson, Independent developer
Claude Code vs AiderPersonal blog post, Mechanistic Mind (Substack)
Rates Aider faster, more transparent and model-flexible than Claude Code, but by April 2025 used Claude Code for complex requests in larger codebases and Aider for smaller tasks.
Practitioner account, self-described; direct use; counts toward Delivered outcomes, Operational control
Counted for All practitioners.
Positive
devdp430, Hacker News commenter
Comment on: Ask HN: How can I get better at using AI for programming?Hacker News comment
Chose Aider for a Go codebase of thousands of files, where agentic tools got overwhelmed. Adding files is manual, but the edits are accurate and never silent.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
gwd, Hacker News commenter
Comment on: Opencode: AI coding agent, built for the terminalHacker News comment
Compared with opencode, Aider asks permission for nearly everything and gives precise control over which files the model sees.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
piokoch, Hacker News commenter, self-described
Comment on: Firebase StudioHacker News comment
Pairs Aider with whichever model works best that month and names the files in context, which costs less and avoids confusing the model.
Practitioner account, pseudonymous; direct use; counts toward Operational control
Positive
jamil7, Hacker News commenter
Comment on: I've been using Claude Code for a couple of daysHacker News comment
On a Swift codebase with internal dependencies in other repos, Aider reads those files as read-only; Claude Code burned time and tokens searching for files.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes
Positive
BeetleB, Hacker News commenter, self-described
Comment on: I've been using Claude Code for a couple of daysHacker News comment
Recommends Aider as cheaper than Claude Code: pick models from its leaderboard, work incrementally on branches, cap context near 25k tokens and add only needed files.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
ryanobjc, Hacker News commenter
Comment on: How I program with LLMsHacker News comment
Gets more value from LLMs through Emacs and Aider, which leaves the user in control of the context and turns each chat change into a git commit that is easy to undo.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed
BeetleB, Hacker News commenter
Comment on: Agentic Coding RecommendationsHacker News comment
Keeps Aider under 25k tokens of context at about 3 US cents a prompt on Gemini, but says it now lags Claude Code significantly; editor independence keeps BeetleB on it.
Practitioner account, pseudonymous; direct use; counts toward Cost predictability and limits
Positive
m3adow, Hacker News commenter
Comment on: Ask HN: Cursor or Windsurf?Hacker News comment
Uses Aider for small to mid-scope changes: adds the files, discusses in architect or ask mode, and switches to code mode only when satisfied.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
Andrew Keenan Richardson, Independent developer
Claude Code vs AiderPersonal blog post, Mechanistic Mind (Substack)
Rates Aider faster, more transparent and model-flexible than Claude Code, but by April 2025 used Claude Code for complex requests in larger codebases and Aider for smaller tasks.
Practitioner account, self-described; direct use; counts toward Delivered outcomes, Operational control
Counted for All practitioners.
Positive, 9
Positive
troyvit, Hacker News commenter
Comment on: Mistral releases Devstral2 and Mistral Vibe CLIHacker News comment
Says Aider works as a chat interface or from the editor by watching your files, and lets you tier models so a big model does the thinking while code reviews run on a smaller one. Last they looked, the maintainer had to focus elsewhere, and aider-ce is a fantastic fork.
Practitioner account, pseudonymous; direct use; counts toward Adoption and upkeep
Positive
gwd, Hacker News commenter
Comment on: Claude 3.7 Sonnet and Claude CodeHacker News comment
Daily Aider user on Claude 3.5 who values seeing each query's cost and the session total; Claude Code struggled with basic patches to the same Go code.
Practitioner account, pseudonymous; direct use; counts toward Cost predictability and limits
Positive
hboon, Hacker News commenter
Comment on: Aider in Your IDEHacker News comment
Satisfied Aider user in Neovim who likes its simplicity and the new mode that acts on instructions written as code comments.
Practitioner account, pseudonymous; direct use; not counted, takes no position
Positive
Magneticdud, Aider user commenting on a GitHub issue
Comment on: No release notes for v0.86.2GitHub issue comment
Says this latest release works great with gpt-5.3-codex and thanks everyone who improved it, preferring Aider to the full-blown VS forks.
Practitioner account, pseudonymous; direct use; not counted, takes no position
Positive
aitchnyu, Hacker News commenter
Comment on: The new Aider-CE fork of Aider is now officialHacker News comment
Prefers Aider because it lets them examine each line, where an agent would produce large diffs, and edits by leaving AI comments in the IDE.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
Andrew Keenan Richardson, Independent developer
Claude Code vs AiderPersonal blog post, Mechanistic Mind (Substack)
Rates Aider faster, more transparent and model-flexible than Claude Code, but by April 2025 used Claude Code for complex requests in larger codebases and Aider for smaller tasks.
Practitioner account, self-described; direct use; counts toward Delivered outcomes, Operational control
Positive
inciampati, Hacker News commenter
Comment on: I've been using Claude Code for a couple of daysHacker News comment
Calls Aider Claude Code's open-source inspiration; it is model agnostic, switching providers is trivial, and inciampati now rarely opens an editor.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
Harper Reed, CEO of 2389 Research
My LLM codegen workflow atmPersonal blog post, harper.blog
Hones an idea into a spec with ChatGPT 4o or o3 and turns it into a prompt plan, then pastes each prompt into Aider, which runs the test suite and debugs. If a step doesn't work, works through it with Aider in Q&A before moving on.
Practitioner account, self-described; direct use; counts toward Delivered outcomes
Positive
tomw1808, Hacker News commenter
Comment on: Zed now predicts your next edit with Zeta, our new open modelHacker News comment
Runs Aider in a Docker container in the IDE terminal; it changes code only when asked, so the editor stays free of suggestions.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed or negative, 2
Mixed
rohitttttt, Developer packaging Aider for colleagues
Wants to wrap Aider as an IDE plugin because the CLI install is too complicated for less experienced colleagues to follow.
not counted until substantiated
Mixed
gawa, Hacker News commenter, self-described
Comment on: AI Is Like a Crappy ConsultantHacker News comment
With Gemini 2.5 Pro, Aider keeps editing other parts of the file with unrequested improvements, and its per-file, per-commit review makes partial acceptance hard.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Counted for All practitioners.
Positive
fragmede, Hacker News commenter
Comment on: Cursor 1.0Hacker News comment
Values that Aider commits to git so each prompt can be undone, which Claude Code and Cursor don't do. Says the author could run parallel work by hand with a new git worktree and another copy of Aider, but finds ChatGPT Codex's way of doing it so smooth.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed
Calvin Hendryx-Parker, CTO at Six Feet Up
AI Assistants and uv: Which Dev Tools Deliver?Company news post, Six Feet Up
Six Feet Up's CTO gave five assistants the same Django refactor. Aider offered fine-grained git control and a choice of models but needed more coaching to finish, while Claude Code made the most complete refactor.
"Aider offered fine-grained Git control and multi-model customization, but needed more coaching to complete tasks."
Practitioner account, self-described; second-hand; not counted until substantiated
Positive
joshstrange, Hacker News commenter, self-described
Comment on: When AI promises speed but delivers debugging hellHacker News comment
Relies on Aider's default auto-commits to jump back when the model loses the thread, and does renames and large moves by hand because Aider handles them poorly.
Practitioner account, pseudonymous; direct use; counts toward Reliability
Positive
ryanobjc, Hacker News commenter
Comment on: How I program with LLMsHacker News comment
Gets more value from LLMs through Emacs and Aider, which leaves the user in control of the context and turns each chat change into a git commit that is easy to undo.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
roebi, Aider user commenting on a GitHub issue
Comment on: Thoughts and questions about the future direction of aiderGitHub issue comment
A user since before Claude Code still values Aider's terminal-first, git-native, model-agnostic design, and built a package to add agent skills since Aider has no plugins.
Practitioner account, pseudonymous; direct use; counts toward Operational control, Adoption and upkeep
Positive
BeetleB, Hacker News commenter, self-described
Comment on: I've been using Claude Code for a couple of daysHacker News comment
Recommends Aider as cheaper than Claude Code: pick models from its leaderboard, work incrementally on branches, cap context near 25k tokens and add only needed files.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Counted for All practitioners.
Mixed
BeetleB, Hacker News commenter
Comment on: Agentic Coding RecommendationsHacker News comment
Keeps Aider under 25k tokens of context at about 3 US cents a prompt on Gemini, but says it now lags Claude Code significantly; editor independence keeps BeetleB on it.
Practitioner account, pseudonymous; direct use; counts toward Cost predictability and limits
Positive
Karrot_Kream, Hacker News commenter
Comment on: Agentic Coding RecommendationsHacker News comment
Mostly uses Aider and just tried Codex. Aider needs more guiding but costs an order of magnitude less, and Karrot_Kream iterates on a plan in Aider before it writes code, then reviews the result.
Practitioner account, pseudonymous; direct use; not counted, takes no position
Positive
jamil7, Hacker News commenter
Comment on: I've been using Claude Code for a couple of daysHacker News comment
On a Swift codebase with internal dependencies in other repos, Aider reads those files as read-only; Claude Code burned time and tokens searching for files.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes
Positive
BeetleB, Hacker News commenter, self-described
Comment on: I've been using Claude Code for a couple of daysHacker News comment
Recommends Aider as cheaper than Claude Code: pick models from its leaderboard, work incrementally on branches, cap context near 25k tokens and add only needed files.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed
Harper Reed, CEO of 2389 Research
Basic Claude CodePersonal blog post, harper.blog
Reed migrated from Aider to Claude Code and has not looked back. Still likes Aider, which has a distinct use, but finds Claude Code a bit more useful for now and a hell of a lot more expensive.
"I have since migrated from Aider to Claude Code and not looked back. I still like Aider, and it has a distinct use, but Claude Code is a bit more useful atm."
"Claude Code is powerful, and a hell of a lot more expensive."
Practitioner account, self-described; direct use; not counted, takes no position
Positive
piokoch, Hacker News commenter, self-described
Comment on: Firebase StudioHacker News comment
Pairs Aider with whichever model works best that month and names the files in context, which costs less and avoids confusing the model.
Practitioner account, pseudonymous; direct use; counts toward Operational control
Counted for All practitioners.
Positive
devdp430, Hacker News commenter
Comment on: Ask HN: How can I get better at using AI for programming?Hacker News comment
Chose Aider for a Go codebase of thousands of files, where agentic tools got overwhelmed. Adding files is manual, but the edits are accurate and never silent.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
aitchnyu, Hacker News commenter
Comment on: The new Aider-CE fork of Aider is now officialHacker News comment
Prefers Aider because it lets them examine each line, where an agent would produce large diffs, and edits by leaving AI comments in the IDE.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
faangguyindia, Hacker News commenter
Comment on: Claude Sonnet 4 now supports 1M tokens of contextHacker News comment
Switches to DeepSeek R1 as architect and Qwen3 Coder as editor in Aider when Gemini 2.5 Pro gets stuck; reasoning models plan well but write weaker code.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
Karrot_Kream, Hacker News commenter
Comment on: Agentic Coding RecommendationsHacker News comment
Mostly uses Aider and just tried Codex. Aider needs more guiding but costs an order of magnitude less, and Karrot_Kream iterates on a plan in Aider before it writes code, then reviews the result.
Practitioner account, pseudonymous; direct use; not counted, takes no position
Positive
Harper Reed, CEO of 2389 Research
My LLM codegen workflow atmPersonal blog post, harper.blog
Hones an idea into a spec with ChatGPT 4o or o3 and turns it into a prompt plan, then pastes each prompt into Aider, which runs the test suite and debugs. If a step doesn't work, works through it with Aider in Q&A before moving on.
Practitioner account, self-described; direct use; counts toward Delivered outcomes
Positive
joshstrange, Hacker News commenter, self-described
Comment on: When AI promises speed but delivers debugging hellHacker News comment
Relies on Aider's default auto-commits to jump back when the model loses the thread, and does renames and large moves by hand because Aider handles them poorly.
Practitioner account, pseudonymous; direct use; counts toward Reliability
Every concern practitioners raise, most raised first.
Counted for All practitioners.
Raised as a concern, 10
Positive
Altchemist, Junior to intermediate software engineer
Comment on: Thoughts and questions about the future direction of aiderGitHub issue comment
A junior to intermediate engineer who stayed away from AI because it affected their learning path calls Aider "such a paradigm shift in AI usage" and still relevant in 2026. Says its positive feedback to the human outweighs its performance against competitors, and the project needs to be forked or maintained.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed
fingon, Aider user commenting on a GitHub issue
Comment on: Thoughts and questions about the future direction of aiderGitHub issue comment
Calls Aider essentially a one-person project that has slowed down; after using it a lot, fingon now uses Claude Code for work.
not counted, repeats a stronger account
Positive
yiouyou, Aider user filing a GitHub issue
Thoughts and questions about the future direction of aiderGitHub issue
Still relies on Aider after nearly a year of daily use and trials of Claude Code, Codex and Antigravity, and asks about its future as releases slow.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
troyvit, Hacker News commenter
Comment on: Mistral releases Devstral2 and Mistral Vibe CLIHacker News comment
Says Aider works as a chat interface or from the editor by watching your files, and lets you tier models so a big model does the thinking while code reviews run on a smaller one. Last they looked, the maintainer had to focus elsewhere, and aider-ce is a fantastic fork.
Practitioner account, pseudonymous; direct use; counts toward Adoption and upkeep
Mixed
therealmarv, Hacker News commenter, self-described
Comment on: Gemini CLI tips and tricks for agentic codingHacker News comment
Says Aider is in a sad state because the maintainer has not been maintaining it, and that the community fork cannot match development in other CLI tools.
Practitioner account, pseudonymous; opinion; not counted, repeats a stronger account
Mixed
adrianwebb, Aider user filing a GitHub issue, self-described
What is the intended future of Aider?GitHub issue
Asks whether Aider has a future, since users need to decide whether to keep using it and contributing while the maintainer works on other projects.
Practitioner account, pseudonymous; direct use; counts toward Adoption and upkeep
Mixed
ErichBSchulz, Aider user
Where is Paul?GitHub issue on Aider-AI/aider
An Aider user wrote that the project's most urgent issue was clarifying the maintainer's status, since he had gone silent on X and Discord. The user named the aider-ce fork as the most active one.
"The most urgent issue currently for this project is clarifying Paul's status. He has gone dark on twitter (I've pinged him) and in the discord."
"In the meantime, the most active fork of this project seems to be https://github.com/dwash96/aider-ce/ where Dustin is doing some great work."
Practitioner account, pseudonymous; opinion; not counted until substantiated
Mixed
Kreijstal, Aider user commenting on a GitHub issue, self-described
Comment on: No new releases since early AugGitHub issue comment
Has moved to a community fork because Aider became a slow repository with about one update every six months.
Practitioner account, pseudonymous; opinion; not counted, repeats a stronger account
Mixed
whoabuddy, Aider user filing a GitHub issue
No new releases since early AugGitHub issue
Used to frequently run pip install -U aider-chat before doing any work, as releases came out on a fairly consistent basis. On 0.86.1, asks why there has been no release since Aug 9.
Practitioner account, pseudonymous; direct use; counts toward Adoption and upkeep
Positive
flyinglizard, Hacker News commenter, self-described
Comment on: Claude Code 2.0Hacker News comment
Stays with Aider for its fit and model switching, but wishes a larger team maintained it, and is not comfortable letting agents loose on production code.
Practitioner account, pseudonymous; direct use; counts toward Operational control
Discussed without concern, 1
Positive
aserafini, Hacker News commenter, self-described
Comment on: Anthropic blocks third-party use of Claude Code subscriptionsHacker News comment
Tried Claude Code and returned to Aider, and now values its conservative pace and lack of plugins as a way to master the core edit loop.
Practitioner account, pseudonymous; direct use; counts toward Adoption and upkeep
GILD's fact checks, which don't count as sources
Checked Oct 12, 2026.The latest GitHub release is v0.86.0 of August 9, 2025. PyPI has 0.86.2 of February 12, 2026, and the last commit to the main branch is dated May 22, 2026. Aider releases on GitHub
Open them in the recordCounted for All practitioners.
Raised as a concern, 7
Mixed
Musa Nayyer, Independent developer
I Tried Building a Complex Security Tool with a 1.5B Local Model: Here's What BrokePersonal blog post, DEV Community
A local 1.5B model in Aider answered only "Ok." with the full repo map in context; scoping work to one module at a time made it productive.
Practitioner account, self-described; direct use; counts toward Delivered outcomes, Reliability, Operational control, Adoption and upkeep
Negative
ebercruzdev, Hacker News commenter
Comment on: Beating Aider's 20% pass rate on local Qwen 32B using deterministic RAGHacker News comment
Ran Aider directly against a local qwen2.5-coder 32B on Ollama and measured context drift and a 20% pass rate on multi-step tasks.
Practitioner account, pseudonymous; structured test; counts toward Delivered outcomes
Negative
singron, Hacker News commenter
Comment on: My experience with Claude Code after two weeks of adventuresHacker News comment
Aider on GPT-4o and o3-mini was weak at GDScript and terrible at editing Godot scene and resource files.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Negative
javcasas, Hacker News commenter
Comment on: Thesis: Interesting work is less amenable to the use of AIHacker News comment
With Ollama and Aider on a 200-line Python simulator, the model extracted the wrong classes, deleted code and misread logic.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed
regularfry, Hacker News commenter
Comment on: Llama.cpp AI Performance with the GeForce RTX 5090 ReviewHacker News comment
Runs Aider on local Ollama models in architect mode, a reasoning architect plus a coder; local reasoning models do not always stick to the edit format.
Practitioner account, pseudonymous; direct use; counts toward Operational control
Mixed
retrodaredevil, Hacker News commenter
Comment on: Aider in Your IDEHacker News comment
Ran Aider in IntelliJ's terminal, where a local Llama 3.1 8B made incorrect changes and GPT-4o edited well. The commenter does not plan to pay OpenAI long term.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed
NitpickLawyer, Hacker News commenter, self-described
Comment on: Kiro: A new agentic IDEHacker News comment
Loves Aider but sees larger gains in newer agentic tools; Claude 4 is better at agentic flows than at producing the diffs Aider expects.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Discussed without concern, 2
Positive
msgodel, Hacker News commenter
Comment on: Andrej Karpathy's talk on the future of the industryHacker News comment
Runs Aider on a local Qwen3 8B through llama.cpp and finds it better than a RAG tool of the commenter's own; a 0.5B model is not enough for Aider.
Practitioner account, pseudonymous; direct use; counts toward Operational control, Adoption and upkeep
Positive
Ey7NFZ3P0nzAe, Hacker News commenter, self-described
Comment on: Ask HN: How much better are AI IDEs vs. copy pasting into chat apps?Hacker News comment
A year of Aider use: learn the chat and edit modes per model, build one feature at a time with limited context, and write preferences into convention files.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Counted for All practitioners.
Raised as a concern, 7
Negative
paretolaw, Hacker News commenter, self-described
Comment on: Anomalyco/OpenCode, Aider-AI/aider, and Cline ReviewHacker News comment
Saw little point in Aider after trying it, because it will not search the repository itself, refuses files not added to the chat, and runs on paretolaw's own API spend.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Negative
KronisLV, Hacker News commenter
Comment on: Ask HN: How is AI-assisted coding going for you professionally?Hacker News comment
Found Aider unpleasant, with too much busywork; an Anthropic Max subscription with Claude Code has since replaced KronisLV's other tools.
Practitioner account, pseudonymous; direct use; not counted, takes no position
Mixed
noo_u, Hacker News commenter
Comment on: Vibe Coding Killed CursorHacker News comment
Uses Aider to edit config files and small utility scripts, and wishes it analyzed project structure as well as Claude Code.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Negative
tharkun__, Hacker News commenter
Comment on: Development speed is not a bottleneckHacker News comment
In a 15-year-old monorepo of about 100 services, Aider's repo map exceeded the token limit; Claude Code handles cross-codebase tasks without file picking.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes
Mixed
BeetleB, Hacker News commenter
Comment on: Agentic Coding RecommendationsHacker News comment
Keeps Aider under 25k tokens of context at about 3 US cents a prompt on Gemini, but says it now lags Claude Code significantly; editor independence keeps BeetleB on it.
Practitioner account, pseudonymous; direct use; counts toward Cost predictability and limits
Mixed
Calvin Hendryx-Parker, CTO at Six Feet Up
AI Assistants and uv: Which Dev Tools Deliver?Company news post, Six Feet Up
Six Feet Up's CTO gave five assistants the same Django refactor. Aider offered fine-grained git control and a choice of models but needed more coaching to finish, while Claude Code made the most complete refactor.
"Aider offered fine-grained Git control and multi-model customization, but needed more coaching to complete tasks."
Practitioner account, self-described; second-hand; not counted until substantiated
Positive
Andrew Keenan Richardson, Independent developer
Claude Code vs AiderPersonal blog post, Mechanistic Mind (Substack)
Rates Aider faster, more transparent and model-flexible than Claude Code, but by April 2025 used Claude Code for complex requests in larger codebases and Aider for smaller tasks.
Practitioner account, self-described; direct use; counts toward Delivered outcomes, Operational control
Discussed without concern, 5
Positive
Anurag Singh, Technology journalist at XDA Developers
I stopped forcing every coding job through Claude and started using this insteadIndependent review, XDA Developers
Moved routine work from Claude Code to Aider to pick a model per task, cheap DeepSeek or stronger Claude, Gemini or GPT; Claude Code stays ahead on autonomy.
Practitioner account, self-described; direct use; counts toward Operational control, Adoption and upkeep
Mixed
thunfischtoast, Hacker News commenter
Comment on: I was banned from Claude for scaffolding a Claude.md file?Hacker News comment
Moved from Aider on OpenRouter to a Claude Code subscription and hit its quota quickly; Claude Code is freer, but pure coding tasks work better in Aider.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
devdp430, Hacker News commenter
Comment on: Ask HN: How can I get better at using AI for programming?Hacker News comment
Chose Aider for a Go codebase of thousands of files, where agentic tools got overwhelmed. Adding files is manual, but the edits are accurate and never silent.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Positive
Karrot_Kream, Hacker News commenter
Comment on: Agentic Coding RecommendationsHacker News comment
Mostly uses Aider and just tried Codex. Aider needs more guiding but costs an order of magnitude less, and Karrot_Kream iterates on a plan in Aider before it writes code, then reviews the result.
Practitioner account, pseudonymous; direct use; not counted, takes no position
Positive
LuisGoCity, Engineer on a company ML team
Feature: Get Aider to implement features from jira ticketsGitHub issue
Says the ML team the author works on has started using Aider daily. To make Aider more autonomous as part of wider AI adoption the company wants, the author forked it and added 5 commands that let it solve a Jira ticket from an issue id.
Practitioner account, pseudonymous; ran the rollout; not counted, takes no position
GILD's fact checks, which don't count as sources
Checked Oct 12, 2026.Aider sends a map of the whole repository with each request and recommends adding only the files that need to change. It can ask to add files it needs. Practitioners who want an agent that opens files and runs commands on its own report this as a limit. Tips
Open them in the recordCounted for All practitioners.
Mixed
chernistry, Maintainer of an orchestrator that runs CLI coding agents
Comment on: Aider exits with code 0 on fatal API Connection errors, breaking headless automation pipelinesGitHub issue comment
Their orchestrator runs Aider with --yes-always and auto-commits; every headless outcome exits 0, so they classify runs from log text instead.
Practitioner account, pseudonymous; ran the rollout; counts toward Reliability
Mixed
noelsaw1, Engineer running Aider in headless automation
Aider exits with code 0 on fatal API Connection errors, breaking headless automation pipelinesGitHub issue
In headless runs Aider exits 0 after fatal API connection errors, so their automation records failures as successful no-op runs.
Practitioner account, pseudonymous; structured test; counts toward Reliability
Mixed
gberkolaiko, Aider user filing a GitHub issue
Aider makes changes in ask modeGitHub issue
A misphrased request in ask mode, which the docs say never changes files, still edited the LaTeX file being discussed.
Practitioner account, pseudonymous; structured test; counts toward Operational control
Mixed
ErichBSchulz, Aider user
Bug report - aider disgards changes if the LM makes changes *and* adds files (which user agrees to) in same response.GitHub issue on Aider-AI/aider
Seen repeatedly: when the model edits files and asks to add more files in the same response, Aider drops the edits while the model carries on as if they were applied.
"It took me a while to see this but now I've seen it repeatedly. When the LM makes changes and asks for file, the LM is (not unreasonably) assuming it's changes have been made, but they are not."
Practitioner account, pseudonymous; direct use; counts toward Reliability
Mixed
trashhalo, Aider user filing a GitHub issue
100% cpu freezing does not respond to ctrl c on latest releaseGitHub issue
Aider hung at 100% CPU while rendering markdown and ignored Ctrl-C, requiring a forced kill.
Practitioner account, pseudonymous; direct use; counts toward Reliability
Negative
tharkun__, Hacker News commenter
Comment on: Development speed is not a bottleneckHacker News comment
In a 15-year-old monorepo of about 100 services, Aider's repo map exceeded the token limit; Claude Code handles cross-codebase tasks without file picking.
Practitioner account, pseudonymous; direct use; counts toward Delivered outcomes
Counted for All practitioners.
Raised as a concern, 6
Mixed
Raghuveer, Software engineer
Aider Architect Mode for Service Scaffolding: Output Quality and Required Follow-UpPersonal blog post, iamraghuveer.com
Aider's architect mode delivered 60-70% of a correct Go microservice scaffold from a spec. Authentication, Redis wiring, integration tests and docker-compose took 5-7 hours of manual fixes, for a time saving of about 0-2 hours, or 0-25%.
Practitioner account; structured test; counts toward Delivered outcomes
Mixed
ramamurthy valavandan, Developer building a personal publishing app
When AI Coding Tools Break Your UI: A Lesson From Aider, Claude & GeminiPersonal blog post, DEV Community
Asked Aider to add a Gemini-powered Research Mode to a Next.js app; it rewrote app/page.tsx and the entire UI disappeared. /undo reversed it, and a constrained prompt naming the files to create and saying not to remove existing fields got the change done.
Practitioner account, self-described; direct use; counts toward Delivered outcomes
Mixed
Max Charas, Senior Staff Engineer, and Marc Bruggmann, Principal Engineer, at Spotify
Background Coding Agents: Context Engineering (Honk, Part 2)Company engineering blog post, Spotify Engineering
Spotify first tried Aider and Goose for its background migration agent. Both got far on a prompt alone, but at the scale of its migrations they could not reliably produce mergeable pull requests.
"When we first got started, we experimented with open source agents such as Goose and Aider."
"But we quickly ran into issues when we tried to scale these agents to our migration use case. It was difficult to get them to reliably produce mergeable PRs."
Practitioner account, self-described; direct use; counts toward Delivered outcomes
Negative
dingnuts, Hacker News commenter, self-described
Comment on: I'm absolutely rightHacker News comment
Spent two hours with Aider on a one-line change it got wrong with a hallucinated parameter, then did it from the docs in two minutes.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Negative
cess11, Hacker News commenter
Comment on: LLM InevitabilismHacker News comment
Gave up on Aider because the assistant kept adding checks for errors that the code's design already made impossible.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Mixed
gawa, Hacker News commenter, self-described
Comment on: AI Is Like a Crappy ConsultantHacker News comment
With Gemini 2.5 Pro, Aider keeps editing other parts of the file with unrequested improvements, and its per-file, per-commit review makes partial acceptance hard.
Practitioner account, pseudonymous; direct use; not counted, repeats a stronger account
Discussed without concern, 1
Mixed
Aaron Sumner, Software Engineering Manager at O'Reilly Media
What I'm up to (September 2025)Personal blog post, aaronsumner.com
Sumner's team maintains nearly fifty systems; this year Sumner used agents, first Aider and now Crush, to build long-lost context on old, undocumented systems. Bigger swings mean bigger diffs for teammates to review, so the author breaks work into chunks and stays the human in the loop.
"This year I’ve been using agents (first Aider; currently Crush) to build long-lost context to systems that are old enough to vote…"
"I’ve felt emboldened to take bigger swings, which usually means bigger diffs for my fellow human team members and me to review."
Practitioner account, self-described; direct use; counts toward Delivered outcomes
Statements this group's sources agree on, then the ones close to agreement.
Close to agreement: each needs 17 supporting sources
Aider gives the best results on small, well-scoped changes with the right files in the chat, and degrades on large open-ended features.
8 supporting sourcesDiscussing the change in /ask or splitting planning and editing in architect mode produces better edits than prompting code mode directly.
6 supporting sourcesQuestions the sources split on, with the count on each side.
Evidence remains mixed
Keep the developer in control, 11 sources
Some stay with Aider because it changes only what they ask for, in the files they choose, and keeps every change reviewable.
Move to an agentic tool, 12 sources
Others moved to Claude Code, Codex, or similar agents that find files and run commands, then iterate without step-by-step direction.
Evidence remains mixed
Pay per token, 4 sources
Some find per-token spend with tight context cheaper than a subscription for the way they work.
Flat subscription, 3 sources
Others found API spend too high or unpredictable and moved to a flat plan such as Claude Pro or Max.
Evidence remains mixed
Cheaper models work, 3 sources
Some get useful results from local Qwen, DeepSeek, or gpt-oss models, often in architect mode.
Frontier models needed, 2 sources
Others found free and local models fall far short of Claude, GPT, or Gemini.
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 3 accounts; GILD prints their sentiment from 20.
10 sources. These accounts emphasize team settings, git and hook policy, headless automation, and whether a team standardizes on Aider or an agentic tool.
75 sources. These accounts emphasize model choice, control over context, git history, spend on personal keys, and how Aider compares with agentic tools in self-directed work.
85 sources. This counts every account: individual builders and the few team accounts together.
85 sources from practitioners, by the context they work in. Filled rows count for this group.
Emphasizes Model choice, context control, and how Aider compares with agentic tools
Emphasizes Headless automation, git hooks, and packaging for colleagues
Emphasizes Daily team use and extending Aider for team workflows
Emphasizes Two leaders' accounts, from O'Reilly Media and Six Feet Up, cover rebuilding context on old code and how much coaching Aider needs to finish a refactor; pilot it with one team and count merged changes
Emphasizes Spotify tried Aider for automated migrations across its repositories and moved on when it could not reliably produce mergeable pull requests
What changes with scale. Individual builders weigh model choice and explicit context against the convenience of agentic tools. Team accounts turn to settings that matter at scale, such as git hooks and exit codes in headless runs, and to packaging Aider for colleagues who will not install a Python CLI. At organization scale, Spotify's account shows the limit: it could not get reliably mergeable pull requests from Aider across automated migrations.
Where this group's sources were published.
Counted for All practitioners, from 85 sources. Change group
106 sources for this group, 21 of them documentation
Older evidence is retained when it still applies. Sources that depend on a changed release, a fixed bug, or a replaced model stay in the history, marked Needs a recheck.
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 Aider, 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 0 of 3, individual builders 0 of 56, all practitioners 0 of 59.
Each row opens its sources for the group you are reading.
| Finding | Engineering leaders and teams | Individual builders | All practitioners | Sources |
|---|---|---|---|---|
| Current readCurrent sentiment | 0 positive, 3 mixed, 0 negative | 30 positive, 14 mixed, 12 negative | 30 positive, 17 mixed, 12 negative | Sources |
| PraiseAny Model, Any Provider | 1 of 1 positive (under the minimum) | 14 of 14 positive | 15 of 15 positive | Sources |
| PraiseExplicit Context Control | 0 of 0 positive (under the minimum) | 9 of 9 positive | 9 of 9 positive | Sources |
| PraiseTerminal Workflow | 0 of 1 positive (under the minimum) | 9 of 10 positive | 9 of 11 positive | Sources |
| PraiseGit-Native Changes | 1 of 1 positive (under the minimum) | 5 of 5 positive | 6 of 6 positive | Sources |
| PraiseLower Token Use | 0 of 0 positive (under the minimum) | 6 of 6 positive | 6 of 6 positive | Sources |
| PraisePlanning & Control | 0 of 0 positive (under the minimum) | 6 of 6 positive | 6 of 6 positive | Sources |
| ConcernManual Context and Less Autonomy | 1 of 2 raise it (under the minimum) | 6 of 10 raise it | 7 of 12 raise it | Sources |
| ConcernSlower Release Pace | 0 of 0 raise it (under the minimum) | 10 of 11 raise it | 10 of 11 raise it | Sources |
| ConcernResults Depend on the Model | 0 of 0 raise it (under the minimum) | 7 of 9 raise it | 7 of 9 raise it | Sources |
| ConcernReliability | 2 of 2 raise it (under the minimum) | 4 of 4 raise it (under the minimum) | 6 of 6 raise it | Sources |
| ConcernRequires Close Review | 1 of 2 raise it (under the minimum) | 5 of 5 raise it | 6 of 7 raise it | Sources |
| ConcernCommands and Config Trust | 1 of 2 raise it (under the minimum) | 2 of 2 raise it (under the minimum) | 3 of 4 raise it (under the minimum) | Sources |
| ConsensusStrongest on scoped changes | 0 supporting (under the minimum) | 8 supporting (below threshold) | 8 supporting (below threshold) | Sources |
| ConsensusPlan before it edits | 0 supporting (under the minimum) | 6 supporting (below threshold) | 6 supporting (below threshold) | Sources |
| ConsensusA pair under your control | 0 supporting (under the minimum) | 4 supporting (under the minimum) | 4 supporting (under the minimum) | Sources |
| ConsensusFaster on self-directed work | 0 supporting (under the minimum) | 3 supporting (under the minimum) | 3 supporting (under the minimum) | Sources |
| DisagreementIs developer-driven control better than an agent that explores and runs on its own? | Keep the developer in control 1, Move to an agentic tool 2 (under the minimum) | Keep the developer in control 10, Move to an agentic tool 10 | Keep the developer in control 11, Move to an agentic tool 12 | Sources |
| DisagreementIs paying per token on your own key better value than a flat subscription? | Pay per token 0, Flat subscription 1 (under the minimum) | Pay per token 4, Flat subscription 2 | Pay per token 4, Flat subscription 3 | Sources |
| DisagreementDo cheaper, open-weight, or local models do the job in Aider? | Cheaper models work 0, Frontier models needed 0 (under the minimum) | Cheaper models work 3, Frontier models needed 2 | Cheaper models work 3, Frontier models needed 2 | Sources |
| Current readEvidence mix | 31 sources | 96 sources | 106 sources | Sources |
GILD's checks of claims against official documentation, kept apart from the source counts.
