- Windsurf
- Cursor
- GitHub Copilot
- Tabnine
- Amazon Q Developer
- JetBrains AI Assistant
- DeepSeek Coder🏚️
- Phind🏚️
- llm-vscode🏚️
- Qodo (formerly CodiumAI)
- Zed AI
- Continue
- Aider
- Cline: code
- Roo Code
- Kilo Code: code
- Firebase Studio (formerly Project IDX)
- bolt.diy
- goose
- Jules
- Open SWE
- Kiro
- OpenCode
- Antigravity
Spec driven
Orchestrators
Code review
StrongDM Software Factory
Simon Willison’s Agentic Engineering Patterns
Articles and videos
- Coding Made AI—Now, How Will AI Unmake Coding? — It won’t replace many coding jobs, but many coding jobs will be increasingly AI-assisted by (19 September 2022) ► The subtitle says it all.
- Replit’s Ghostwriter AI can explain programs to you—or help write them — Ghostwriter generates, completes, or transforms code in 16 languages, similar to GitHub Copilot. by (31 October 2022) ► The subtitle says it all.
- DeepMind’s latest AI project solves programming challenges like a newb — Google’s AI division tackles programming languages with a language model. by (8 December 2022) ► DeepMind tries to create an AI generating code, the results are still not good.
- L’IA va-t-elle changer notre métier de développeur ? (Louis-Guillaume MORAND, Eric GRENON)↓ by and (13 April 2023) ► Some good advice, a lot of bullshit, and plenty of broken demos.
- AI system devises first optimizations to sorting code in over a decade — Writing efficient code was turned into a game, and the AI played to win. by (7 June 2023) ► DeepMind created AlphaDev to optimise assembly code and improved some sorting algorithms.
- A Critical Look at AI-Generated Software — Coding with ChatGPT, GitHub Copilot, and other AI tools is both irresistible and dangerous by and (11 June 2023) ► A very classic analysis of AI tools for generating code.
- 10X coders beware: Meta’s new AI model boosts coding and debugging for free — New “Code Llama” coding model is free for research and commercial use. by (25 August 2023) ► Facebook gives access to three of the four Code Llama models.
- NEW WizardCoder-34B - THE BEST CODING LLM by (27 August 2023) ► A short presentation of WizardLMTeam/WizardCoder-Python-34B-V1.0 in Google Colab.
- AI-assisted coding: Experiences and perspectives by , , , and (16 November 2023) ► Some experience feedback and reflection on the use of LLMs to assist in code writing.
- I Tried Every AI Coding Assistant↓ by (22 November 2023) ► This comparison of Tabnine, ChatGPT, GitHub Copilot, Bard, CodeWhisperer, Cody, and CodiumAI is poorly done.
- Three things GenAI will not change about software delivery by (29 November 2023) ► Generative AI will not change the fact that software development is not only coding, that it is a teamwork, and that we need to have some metrics.
- Programmer’s Apprentice Season 2: Future Directions in AI-assisted Coding • Erik Meijer • YOW! 2023 by (4 December 2023) ► As usual, has a provocative presentation, but this one has little worthwhile content, he describes only well-known things.
- Three types of AI-assisted programmers — What matters isn’t just whether you use it, but how. by (11 December 2023) ► Some advice on using code assistants: the less competent you are in coding, the more perilous it is to use them; if you know what you are doing, they may be helpful.
- Github Copilot Does NOT Like This... (100% Local Free VSCode CoPilot) by (11 February 2024) ► A demo of Continue, an AI assistant for VS Code that can use different chat models, but the examples shown here are definitely not convincing.
- Sorry DEVIN, Microsoft's AI SWE is coming for ya!!! by (18 March 2024) ► A bad presentation of "AutoDev: Automated AI-Driven Development", a Microsoft paper describing an agent able to edit files, run tests, execute Git commands…
- Will Devin AI Take Your Job? by (19 March 2024) ► Debunking Cognition’s marketing message.
- Building and testing C extensions for SQLite with ChatGPT Code Interpreter by (23 March 2024) ► The title says it all.
- Why I'm no longer using Copilot by (30 March 2024) ► Some rather relevant problems with using code assistants.
- ↪Why I Quit Copilot | Prime Reacts by (6 April 2024) ► Some comments on the previous video.
- Developers with AI assistants need to follow the pair programming model — CodeGen is fast, but you need to be good. by (3 April 2024) ► A long article just to say that AI tools are helpful to produce code, but you need to pilot them and to check what they generate.
- Refactoring with AI↑ by , , , and (18 April 2024) ► An interesting discussion about using AI code assistants for refactoring code, with less bullshit that the usual such communications.
- L'IA et qualité de code : Construire une synergie avec l’intelligence humaine↓ by (3 May 2024) ► Trivialities about AI code generation.
- Oracle Code Assist: AI companion to boost developer velocity by (7 May 2024) ► Oracle is trying to catch up on AI code assistants, but it still has no product yet.
- Developers get by with a little help from AI: Stack Overflow Knows code assistant pulse survey results — We asked when and how often CodeGen tools fall short, what challenges developers face with these tools, and what they are doing with all of the free time these tools purport to offer. by (29 May 2024) ► A few stats about code assistants from StackOverflow survey.
- Trying out free-threaded Python on macOS by (12 July 2024) ► tested the GIL-free build with a buggy benchmark generated by Claude.ai.
- How I Use "AI" by (1 August 2024) ► gives some examples of how he uses LLMs. This may give ideas to the persons who do not use or have a limited usage of LLMs.
- ↪Pragmatic LLM usage with Nicholas Carlini (⧉) by , , , , , and (5 August 2024) ► This is a follow-up of the previous article. explains what he is experimenting with LLMs: decompiling code, quick `n dirty small programs… He regularly tries things that he thinks would not work to see if models have new improved capabilities.
- Devin AI Engineer killer Genie is here (kind of!)↓ by (12 August 2024) ► Yet another company claiming to have improved AI coding capabilities.
- Long Code Arena: How Well Can AI Models Understand Your Entire Project? by (19 August 2024) ► JetBrains proposes a set of benchmarks to evaluate how models perform on some tasks for a full project instead of a small piece of code as for the current benchmarks.
- Building a tool showing how Gemini Pro can return bounding boxes for objects in images by (26 August 2024) ► is using Gemini, Claude, and GPT to build a small application to compute and draw bounding boxes.
- Gemini Chat App by (27 August 2024) ► shares the logs of creating, with Claude 3.5 Sonnet, a prototype of a Web app to chat with Gemini. AI code generation is getting more or more useful.
- VSCode + Aider + Supermaven : STOP PAYING for CURSOR with this 100% FREE & Opensource Alternative by (9 September 2024) ► A presentation of setting up VS Code + Aider + Gemini 1.5 experimental models + Supermaven.
- AI tools for software engineers, but without the hype – with Simon Willison (co-creator of Django) — Ways to use LLMs efficiently, as a software engineer, common misconceptions about them, and tips/hacks to better interact with GenAI tools. The first episode of The Pragmatic Engineer Podcast (⧉) by and (25 September 2024) ► is yet in another podcast where he explains how he uses and experiments with generative AI.
- Little Paper Reviews & AutoCodeRover by (2 October 2024) ► Some notes about AutoCodeRover: Autonomous Program Improvement, an article about yet another coding agent.
- AI-Powered Software Development From the Trenches • Henrik Kniberg • GOTO 2024 by (2 October 2024) ► A presentation of using Claude for writing code and of Cursor, for those who do not know that yet.
- Everything I built with Claude Artifacts this week by (21 October 2024) ► is generating a lot of small tools using Claude’s artefacts. This is the list of these for last week.
- Generating documentation from tests using files-to-prompt and LLM by (5 November 2024) ► The title says it all.
- Microsoft's AI Toolkit for VS Code: This FREE Extension BEATS CURSOR! (w/ FREE GPT-4O!)↓ by (9 November 2024) ► A very light presentation of AI Toolkit Extension for Visual Studio Code.
- From Idea to AI: Building Applications with Generative AI↓ by (12 November 2024) ► If you were not asleep for the last two years, you already know all these very basic facts…
- Townie: This FULLY FREE AI Coder BEATS V0 & Bolt! (w/ Free Claude 3.5 Sonnet) by (16 November 2024) ► is "testing" Val Town’s Townie.
- LMSYS Copilot: This FREE COPILOT Alternative has CLAUDE 3.5 SONNET & GPT-4O FOR FREE!? by (17 November 2024) ► A presentation of Copilot Arena.
- Can LLMs write better code if you keep asking them to “write better code”? by (2 January 2025) ► Experimenting with asking Claude 3.5 Sonnet over and over to improve its code, first, with a basic initial prompt and, then, with a properly specified prompt. Claude is indeed able to improve irs code.
- ↪Can LLMs write better code if you keep asking them to “write better code”? by (3 January 2025) ► Why LLMs can improve their own code.
- How I program with LLMs by (6 January 2025) ► describes his (classic) use of LLMs and that he decided to contribute to yet another tool, sketch.dev, to automate interactions with them.
- Phi-4 + Bolt DIY + Cline & Aider : This 100% PRIVATE & LOCAL AI Coder is INSANE! (Beats Cursor!) by (13 January 2025) ► As indicated in the title: how to use Phi-4 in Cline, bolt.diy, and Aider.
- Le futur du développeur avec l'IA ? by (16 January 2025) ► A very classic opinion about the impact of AI on coding.
- Deepseek-R1 + RooCline & Aider + Free APIs : This CRAZY AI Coder is AMAZING! by (22 January 2025) ► How to use Deepseek-R1 in Roo Cline and in Aider.
- My LLM codegen workflow atm by (16 February 2025) ► describes how he is working with LLMs. His workflows are quite advanced.
- Exploring Generative AI↑ by (18 February 2025) ► A, regularly updated, list of well-thought-out ideas on AI-assisted coding.
- Hallucinations in code are the least dangerous form of LLM mistakes by (2 March 2025) ► explains that hallucinations are a minor problem since they are very easily detected. The real things to do to mitigate risks with LLM-generated code is to review the code and test it.
- Will the future of software development run on vibes? — Accepting AI-written code without understanding how it works is growing in popularity. by (6 March 2025) ► The subtitle says it all.
- Here’s how I use LLMs to help me write code by (11 March 2025) ► The title says it all.
- Not all AI-assisted programming is vibe coding (but vibe coding rocks) by (19 March 2025) ► Yet another occurrence of semantic diffusion.
- ↪Two publishers and three authors fail to understand what “vibe coding” means by (1 May 2025) ► The semantic diffusion goes on…
- ↪Vibes by (1 May 2025) ► More musing about the matter…
- Open R1: How to use OlympicCoder locally for coding by , , , , and (20 March 2025) ► How to configure LM Studio and Continue to run open-r1/OlympicCoder-7B locally.
- We need to talk about vibe coding by , , , and (2 April 2025) ► Sensible comments on "vibe coding".
- My AI Coding Workflow + Free APIs: How I DO AI CODING! (RooCode, Gemini 2.5, T3 Stack, Requesty) by (11 April 2025) ► explains how he currently really uses AI.
- Critical Thinking In The Age Of AI-Generated Code — Learn how AI-generated code impacts maintainability, productivity, and quality. Explore code reviewing, testing strategies, and best practices for long-term success. by (14 April 2025) ► suggests some ways to ensure that AI-generated code is of correct quality, but these are simply the usual best practices, which should also be applied for human-written code.
- Guiding an LLM for Robust Java ByteBuffer Code by (17 April 2025) ► An example of piloting a LLM to generate code and to clean it up.
- ↪Building TMT Mirror Visualization with LLM: A Step-by-Step Journey by (30 April 2025) ► Using a LLM to generate code for a technology you do not know, but there is no information here except for the fact that prompting step by step works better.
- ↪Building Custom Tooling with LLMs — A Case Study in Extending PlantUML by (14 May 2025) ► An example of using Cursor with Claude Sonnet 3.7 to write and clean up the code of a PlantUML extension.
- AMD Takes Holistic Approach to AI Coding Copilots — The chip maker is using AI throughout the software development lifecycle by (22 April 2025) ► A presentation of AMD strategy for AI. But these are generalities, mostly the fact that AI is not only code completion.
- Qwen3 Leads the Pack: Evaluating how Local LLMs tackle First Year CS OCaml exercises by (6 May 2025) ► Evaluating the performance of some open-weight model on Cambridge’s first year Computer Science OCaml exercises.
- Coding Assistants Threaten the Software Supply Chain by and (13 May 2025) ► Coding agents introduce new security risks.
- After months of coding with LLMs, I'm going back to using my brain by (15 May 2025) ► Some feedback on vibe-coding a non-trivial application and finishing with a complete code mess.
- Accelerating mainframe modernization using generative AI by , , , and (15 May 2025) ► Thoughtworks and Mechanical Orchard are creating a partnership to convert mainframe applications to a modern technical stack. They describe how they use generative AI for these migrations.
- How Smart Devs Use AI Tools Every Day↓ by (21 May 2025) ► What is the purpose of this video? Any decent developer should already know about all these AI features.
- AI-assisted coding for teams that can't get away with vibes by (29 May 2025) ► Yet some advice on using AI for coding. I agree with them.
- Peer-Programming in Modern OCaml with ChatGPT and Gemini by (2 June 2025) ► describes his experience with starting to use LLMs to write an OCaml program.
- Phoenix.new is Fly’s entry into the prompt-driven app development space by (23 June 2025) ► Yet another company proposing a coding agent. seems positively impressed.
- LLMs bring new nature of abstraction by (24 June 2025) ► considers that LLM code generation is a major change in software development, as has been migrating from assembler to high-level programming languages, and this is the first time we have to deal with on-determinism.
- Leading your engineers towards an AI-assisted future by (26 June 2025) ► is proposing a process to organise the introduction of generative AI use in development teams.
- llvm: InstCombine: improve optimizations for ceiling division with no overflow - a PR by Alex Gaynor and Claude Code by (30 June 2025) ► Implementing a LLVM optimisation with Claude.
- I still care about the code by (9 July 2025) ► ’s advice is to not vibe-code and, instead, look care about the code. She suggests using risk assessment (impact, probability, detectability) when using AI code generation.
- AI Tooling, Evolution and The Promiscuity of Modern Developers by (9 July 2025) ► AI is changing developer’s habits: they are ready to pay for tools while they used to get them for free, they change and use several tools instead of sticking with the one they learn to master. The current situation is similar to the Cambrian explosion: many experiments, some not created yet, some will survive, other ones will disappear.
- Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity by (12 July 2025) ► and comment on a study finding that AI slows down developers.
- Vibe scraping and vibe coding a schedule app for Open Sauce 2025 entirely on my phone by (17 July 2025) ► The title says it all.
- Clowns to the left of me … by (19 July 2025) ► Nothing interesting here: the truth about using AI for coding is neither that these tools can generate a full application, nor they are useless.
- Who is LLM? by (22 July 2025) ► How to represent a LLM: a stubborn donkey, a genie, a slot machine, or Uriah Heep.
- Announcing Toad - a universal UI for agentic coding in the terminal by (23 July 2025) ► is trying to earn some money by building yet another terminal coding assistant.
- Two major AI coding tools wiped out user data after making cascading mistakes — “I have failed you completely and catastrophically,” wrote Gemini. by (24 July 2025) ► Sometimes things go wrong…
- Can We Make AI Code Assistants Smarter by Asking Them to Write Their Own Rules? by (30 July 2025) ► Rather than trying to define a priori a set of generic rules, use the assistant to write rules adapted to your coding style and practices, and to your project.
- Partner with the AI, throw away the code by (31 July 2025) ► describes his experience with refactoring some complex code. He list the steps he performed and explained that, while not using the code generated by Clause, this helped him understanding the situation.
- Our first outage from LLM-written code by and (31 July 2025) ► A LLM moved code but replaced a
breakinstruction withcontinue, introducing a bug. - How far can we push AI autonomy in code generation? by (5 August 2025) ► describes her setup to experiment with having a LLM generating autonomously a simple Sprint Boot application. She gives a few conclusions and wonders about many questions.
- Jules, our asynchronous coding agent, is now available for everyone by (6 August 2025) ► Maybe "asynchronous coding agent" will be the term to designate tools generating PRs in full autonomy.
- Introducing Open SWE: An Open-Source Asynchronous Coding Agent (6 August 2025) ► LangChain also proposes their own coding agent.
- Context engineering: Tackling legacy systems with generative AI by , , , and (21 August 2025) ► Some feedback on trying to retro-specify a legacy application using a Playwright MCP Server to find the possible user interactions, a simple CDC to detect the data changes in the database, decompilation…
- Conversation: LLMs and Building Abstractions↑ by and (26 August 2025) ► LLM should not be used to build the core of the software, the developer needs to deeply understand the domain and decide how to represent it, s/he can use AI to explore some options, but s/he must drive the decisions. Once this is done, the AI can be used to exploit this domain representation to implement the features.
- Building your own CLI Coding Agent with Pydantic-AI — Learning by doing by (27 August 2025) ► Creating their own CLI agent by simply using Pydantic-AI Framework and several MCP Servers.
- Some thoughts on LLMs and Software Development by (28 August 2025) ► The title says it all.
- Spec-driven development with AI: Get started with a new open source toolkit — Developers can use their AI tool of choice for spec-driven development with this open source toolkit. by (2 September 2025) ► Yet another AI coding tool: Spec Kit which proposes to organise the workflow into specify / plan / split in tasks / implement. But the author does not explain how he maintains the spec while adding new features.
- ICPC medals for OpenAI and Gemini by (17 September 2025) ► OpenAI and Google did very well on the International Collegiate Programming Contest (ICPC).
- CompileBench: Can AI Compile 22-year-old Code? by (22 September 2025) ► Yet another benchmark to evaluate LLM at coding, Claude Opus 4.1 Thinking wins this one.
- To vibe or not to vibe↑ by (23 September 2025) ► proposes a simple risk analysis to determine how much you should review the AI-generated code.
- Anchoring AI to a reference application by (25 September 2025) ► An example of using a MCP Server to allow a coding agent to keep the code aligned with the code of a reference application.
- AI Coding Workflows 🛠️ — with Birgitta Böckeler by and (29 September 2025) ► Some analysis of the use of AI for coding by .
- 90% by (29 September 2025) ► explains how he uses AI.
- ↪Armin Ronacher: 90% by (29 September 2025) ► Some very short notes about the previous article.
- Designing agentic loops↑ by (30 September 2025) ► Some good advice on how to use a coding agent.
- SDS 927: Automating Code Review with AI, feat. CodeRabbit’s David Loker (⧉) by and (30 September 2025) ► Some information about CodeRabbit.
- Mettre en place un assistant de code souverain by (3 October 2025) ► Some good basic advice on how to set up a sovereign code assistant.
- Embracing the parallel coding agent lifestyle by (5 October 2025) ► explains when he is running code agents in parallel.
- ↪How I'm using coding agents: October 2025 by (5 October 2025) ► The same by .
- Vibe engineering by (7 October 2025) ► gives still more advice on how to use coding agents and try to launch the "vibe engineering" terminology.
- Has This Report EXPOSED THE TRUTH About AI Assisted Software Development? by (8 October 2025) ► Some information extracted from DORA report and some advice on how to use AI for building software.
- ↪AI-assisted software development in 2025 — Inside this year's DORA report by and (11 December 2025) ► Some comments on DORA’s report.
- I'm in Vibe Coding Hell by (10 October 2025) ► Some basic advice on using AI when learning to be a developer.
- Understanding Spec-Driven-Development: Kiro, spec-kit, and Tessl by (15 October 2025) ► Some interesting comments on SDD tools. I would have liked more experimentation feedback on the maintenance of the code.
- 2025/11/18 - Le Software Craft et l’IA↑ by (19 October 2025) ► A good overview of the current use of AI use for software development.
- The Learning Loop and LLMs by (4 November 2025) ► Developers need to learn by experimenting. If they do not have this experiment / learn / apply loop and blindly use AI, they will end up struggling with a code base that they cannot maintain because they do not master it.
- Code research projects with async coding agents like Claude Code and Codex by (6 November 2025) ► describes how he uses asynchronous agents to explore some ideas.
- How developers can get the most from new AI coding workflows by and (13 November 2025) ► The best practices of software development are still applicable when using AI to develop.
- Build a Registry of AI Tools. Generate MCP Servers and CLIs. — Define prompts and tools in markdown files by describing them to an LLM. A registry framework converts those markdown specs into MCP servers and CLI tools automatically. You iterate in markdown, not framework code. by (19 November 2025) ► explains his mechanism of a MCP Server whose tools and prompts are defined in Markdown files, these files being generated by a LLM instructed by an
AGENTS.mdfile. - We still need to talk about vibe coding — Reflections on 2025's word of the year by , , and (27 November 2025) ► Some miscellaneous thoughts on the AI impact, from the problem of incompetent persons generating insecure, unmaintained code, to the help AI can provide to experienced developers.
- What I learned building an opinionated and minimal coding agent↑ by (30 November 2025) ► describes how he built a simple code CLI tool. The part about having simple tools and prompts is probably the most interesting.
- How I wrote JustHTML using coding agents by (3 December 2025) ► explains how he used AI to write a new Python library to parse HTML.
- ↪I ported JustHTML from Python to JavaScript with Codex CLI and GPT-5.2 in 4.5 hours by (15 December 2025) ► just let GPT 5.2 run for 5 hours to convert a Python project to JavaScript. It worked, but the fact that there is a really good test coverage probably helped a lot.
- ↪AoAH Day 15: Porting a complete HTML5 parser and browser test suite by (15 December 2025) ► does the same in OCaml.
- ↪AoAH Day 15: Porting a complete HTML5 parser and browser test suite by (17 December 2025) ► A reference to the previous article.
- ↪swift-justhtml by (18 December 2025) ► And now, this is Swift…
- ↪My answers to the questions I posed about porting open source code with LLMs by (11 January 2026) ► tries to answer the questions he asked at the end of his Python → JavaScript translation experiment.
- Useful patterns for building HTML tools↑ by (10 December 2025) ► gives some advice and ideas how how to use Claude, Gemini, or GPT to write an application as a single HTML page.
- The Bet On Juniors Just Got Better — Why genies can make hiring juniors more profitable, & what you need to change to get there by (12 December 2025) ► claims that AI should not deter companies from hiring juniors. On the contrary, AI makes the junior’s ramp up less costly since it enables them to learn faster, if properly used.
- Your job is to deliver code you have proven to work by (18 December 2025) ► states the obvious: you ensure that the code you generated using AI works and you must have automated tests.
- We asked four AI coding agents to rebuild Minesweeper—the results were explosive — How do four modern LLMs do at re-creating a simple Windows gaming classic? by and (19 December 2025) ► A simple comparative test of OpenAI’s Codex based on GPT-5, Anthropic’s Claude Code with Opus 4.5, Google’s Gemini CLI with Gemini 2.5, and Mistral Vibe with an unknown model.
- Settle down, nerds. AI is a normal technology — Ryan welcomes Anil Dash, writer and former Stack Overflow board member, back to the show to discuss how AI is not a magical technology, but rather the normal next step in computing’s evolution. They explore the importance of democratizing access to technology, the unique challenges that LLMs’ non-determinism poses, and how developers can keep Stack Overflow’s ethos of community alive in a world of AI. by and (23 December 2025) ► Explaining that AI is not magic but just another tool (a non-deterministic one), and denouncing the unfairness of AI companies pillaging the generosity of StackOverflow users.
- How AI coding agents work—and what to remember if you use them — From compression tricks to multi-agent teamwork, here’s what makes them tick. by (24 December 2025) ► A rather good overview of the curent status of AI coding agents.
- AI Code Generation: Wins, Fails and the Future by , , , and (26 December 2025) ► Some thoughts on the current state of AI for code generation.
- Party of One for Code Review! by (26 December 2025) ► Code reviews were not really effective before, but the problem is far worse today when AI generates large amount of code.
- What Used to Take Months Now Takes Days — Building production software with Claude Code while doing my day job by (4 January 2026) ► details how he spent his winter vacation writing software with Claude to capture his company’s design decisions as a semantic graph. This should perhaps be taken with a grain of salt, given that he is very proud of himself.
- Make the Tool You Wish Existed (with Your LLM) — Build custom HTML tools that fit your exact workflow using LLMs. No extensive programming knowledge required. Turn workflow problems into practical tools in hours. by (5 January 2026) ► describes how he uses AI to create single-HTML-page tools.
- Why AI is pushing developers toward typed languages — AI is settling the “typed vs. untyped” debate by turning type systems into the safety net for code you didn’t write yourself. by (8 January 2026) ► claims that typed languages are gaining traction because they are safer to use for AI-generated code. But I do not understand his reasoning: I cannot believe developers will change the languages they use because they want to be safer.
- Claude Opus 4.5 vs Mistral Large 3 : qui gagne ? (pour du développement assisté par IA) by (11 January 2026) ► A comparison of the two models for vibe coding: Mitral Large is simply unusable.
- The Death of Software Development by (11 January 2026) ► believes software will become a commodity and he wants to prove it by building an accounting application without using bleeding edge AI.
- Engineering Rigor in the LLM Age (⧉) by , , , and (12 January 2026) ► Some examples of doing good engineering with LLMs.
- 10 things I learned from burning myself out with AI coding agents — Opinion: As software power tools, AI agents may make people busier than ever before. by (19 January 2026) ► Some feedback on using AI to code. I agree with most points, but draws some general conclusions that may reveal to be obsolete in a few months…
- The Five Levels: from Spicy Autocomplete to the Dark Factory by (23 January 2026) ► A classification of AI usage for code generation, from using only autocompletion to not looking at the code.
- ↪The Five Levels: from Spicy Autocomplete to the Dark Factory by (28 January 2026) ► A summary of the previous blog entry.
- We Studied 150 Developers Using AI (Here’s What's Actually Changed...) by (27 January 2026) ► speaks about a paper he contributed to ("Echoes of AI: Investigating the Downstream Effects of AI Assistants on Software Maintainability") showing that code generated with the help of AI is not more difficult to maintain than pure-human code.
- Assessing internal quality while coding with an agent by (27 January 2026) ► Some usual feedback on Windsurf / Sonnet 3.5 and Claude Code / Sonnet 4.5. The second is good enough to convince to use it regularly.
- Context Engineering for Coding Agents↑ by (5 February 2026) ► A good overview of all the ways to provide information and instructions to the LLM when using an AI-powered IDE or CLI tool.
- I design with Claude more than Figma now by (5 February 2026) ► , a designer, has given up using Figma and now pushes working prototypes to the engineers.
- We're All Addicted To Claude Code by , , , , and (6 February 2026) ► A typical Y Combinator podcast: very optimistic, little common sense, but they probably believe in what they say.
- What Skills Do Developers NEED To Have In An AI Future? by and (6 February 2026) ► and speak about curiosity, testing know-how, soft skills… I am not convinced this will help developers to keep their job in a world with AI agents. The reality is that you need to keep learning the state-of-the-art, be disciplined and stick to first principles, cultivate your common sense to differentiate interesting experiment from bullshit…
- Finding Comfort in the Uncertainty by (6 February 2026) ► Some interesting reflections from Thoughwork’s "The future of software development retreat" unconference: many questions, a few ideas of solution not yet mature.
- How StrongDM’s AI team build serious software without even looking at the code by (7 February 2026) ► A StrongDM team does not write or test AI-generated code themselves, they use test agents in a simulated environment, which is itself coded by AI agents. I wonder how good this is really. For example, how to they ensure that the simulated environment really matches the real world?
- AI Doesn’t Reduce Work—It Intensifies It by (9 February 2026) ► Using AI agents is more exhausting than writing code oneself.
- Coding with Opus 4.6 and Codex 5.3 is actually insane by (10 February 2026) ► An informal comparison of Claude Opus 4.6 and GPT 5.3.
- Besieged by (10 February 2026) ► Miscellaneous thoughts about the impact of AI on software development.
- How Generative and Agentic AI Shift Concern from Technical Debt to Cognitive Debt by (15 February 2026) ► Using AI for coding results in the developers not fully understanding the behaviour and the design of the pieces of software they create.
- Harness Engineering by (17 February 2026) ► After she read OpenAI s article ("Harness engineering: leveraging Codex in an agent-first world"), wonders on how to set up the harness to keep AI-generated code on track.
- How to Make the Best of AI Programming Assistants by (18 February 2026) ► Claiming that the Nyquist-Shannon Sampling Theorem proves that tests must be run more frequently because AI allows more frequent code is just bullshit. This is a pity, because the advice is obviously true: if you have no CI/CD and are not always running your tests while generating huge volumes of code with AI, you are asking for trouble.
- The A.I. Disruption We’ve Been Waiting for Has Arrived by (18 February 2026) ► An explanation of the impact of AI on software development for people outside the software world.
- SWE-bench February 2026 leaderboard update by (19 February 2026) ► Yet another benchmark.
- Inside AI/works™: An agentic development platform by , , and (19 February 2026) ► The part about the Thoughworks AI platform is not so interesting, but the discussion about the impact of AI on the SDLC processes is more valuable.
- Is software engineering more than coding? — Did you know that roboticists don't always build robots? Let’s chat about how AI has sparked an existential crisis for some software engineers. by (19 February 2026) ► Miscellaneous thoughts about the impact of AI on developers.
- Writing about Agentic Engineering Patterns by (23 February 2026) ► starts to write a guide to use coding AI agents.
- Knowledge Priming by (24 February 2026) ► Some good advice on how to provide some information to a coding agent.
- I vibe coded my dream macOS presentation app by (25 February 2026) ► Yet another example of an application quickly vibe coded.
- Design-First Collaboration by (3 March 2026) ► proposes an interesting framework to define the design with the AI before starting to generate code.
- Humans and Agents in Software Engineering Loops by (4 March 2026) ► Different patterns of how to work with AI coding agents.
- Pourquoi l'AGI est pour bientôt (et pourquoi on s'en fout) by (7 March 2026) ► Some information about AGI: the Dartmouth conference, the ARC benchmarks, a new paper ("A Definition of AGI") trying to define AGI, the fact there will not be a singularity…
- Perhaps not Boring Technology after all by (9 March 2026) ► Recent models still favour some technologies, but it is easy to enforce the ones you want them to use and they can easily exploit these.
- The 8 Levels of Agentic Engineering by (10 March 2026) ► A good proposal for a classification of where a dev team stands on using AI.
- What is wrong with us?! by (15 March 2026) ► describes the permanent pressure some people feel to push instructions to their AI agents all hours all days.
- Mozilla dev’s “Stack Overflow for agents” targets a key weakness in coding AI — There are major problems to be solved before it can be adopted, though. by (24 March 2026) ► Mozilla proposes cq: a place where AI coding agents can share information about versions of the technologies that are posterior to their training cutoff date.
- Thoughts on slowing the fuck down by (25 March 2026) ► explains that you still need to be in control of the code, understand it, and take the decisions.
- StarCoder
- StarCoder: A State-of-the-Art LLM for Code by and (4 May 2023) ► A family of open-source models and some of their benchmark results.
- StarCoder2 and The Stack v2 by , , , and (28 February 2024) ► A new version trained on a larger dataset.
- Aider
- Aider by (31 July 2024) ► A very short presentation of Aider.
- Aider (Upgraded) + FREE Claude / ChatGPT + Files To Prompt: This is the BEST FREE Way for AI Coding! by (13 November 2024) ► Aider can now integrate edits which are copied/pasted from models accessible as a chat in the Browser.
- Aider (Upgraded) : New Editor Mode, New Commands, Qwen-2.5 Coder Support, O1 Streaming & Much More! by (21 November 2024) ► New Aider features since 0.63.0.
- Aider (New Upgrades) + R1 / Sonnet Mode + Free APIs: This is THE SOTA FREE AI Coder Right Now! by (27 January 2025) ► Comments to give instructions and ask questions to the AI that Aider will handle when the file is saved, using Return in multiline input, and experimenting with DeepSeek-R1 as the architect and Claude Sonnet 3.5 as the editor.
- DeepSeek Coder
- This AI CODING LLM Nobody's talking about!!!↓ by (6 November 2023) ► A presentation of DeepSeek Coder: the guy reads the doc and performs three simple tries.
- DeepSeek Coder V2 - Quick Look! by (5 July 2024) ► A basic presentation of deepseek-ai/DeepSeek-Coder-V2-Instruct.
- Qwen
- Qwen-2.5 Coder (32B) + Cline & Aider + Free API : This NEW AI Coding Model BEATS Claude 3.5 Sonnet!? by (12 November 2024) ► As said by the title, tests Qwen-2.5 Coder, hosted by Hyperbolic or by Fireworks AI, with Cline and Aider. In all cases, it is a failure.
- Qwen2.5-Coder-32B is an LLM that can code well that runs on my Mac by (12 November 2024) ► Some much more positive feedback.
- Cline
- ClaudeDev (Upgraded) : The BEST Coding Agent just got EVEN BETTER! (New Commands, Web Parser, etc.) by (22 September 2024) ►
@commands have been added in 0.18. - Dans les rouages de Cline (ex Claude-dev): un Agent pour assister à coder by (8 October 2024) ► A presentation of the inner working of Claude Dev.
- Cline (Upgraded) : This UPDATE Makes this AI Coder SELF EVOLVING & INSANE! (Beats Cursor, V0, Bolt) by (16 December 2024) ► Cline now supports Model Context Protocol and can create its own tools.
- Cline 3.0 (Upgraded): This is the ULTIMATE AUTONOMOUS AI CODER that we need! (Auto mode, Rules, etc) by (21 December 2024) ► Cline 3.0 adds the possibility to perform some tasks and requests without approval.
- Cline 3.1 (Upgraded) : This NEW Cline Upgrade IS AMAZING! (Checkpoints, Deepseek-V3 Support & More!) by (8 January 2025) ► the new features are DeepSeek support and checkpoints.
- Cline 3.4 : This AI Coder just got MCP Marketplace, Diagram Support & More!↓ by (22 February 2025) ► The usual basic reading of the release note and a completely botched demo.
- Cline (Upgraded) + Free APIs & Crazy MCPs: I MIGHT SWITCH BACK TO CLINE! by (22 April 2025) ► Reading the release notes from 3.8.5 to 3.13.0.
- ClaudeDev (Upgraded) : The BEST Coding Agent just got EVEN BETTER! (New Commands, Web Parser, etc.) by (22 September 2024) ►
- Project IDX
- Project IDX + ClaudeDev + Aider : This is the BEST CLOUD AI Editor by GOOGLE beats Cursor, Bolt & V0↓ by (9 October 2024) ► A very basic presentation of using Project IDX and integrating ClaudeDev and Aider in it. is generating a different program with each tool instead of showing how to use them on the same code base.
- Google's Project IDX (Agent Mode) : This FREE AI Editor BEATS Cursor, V0 & BOLT! by (27 November 2024) ► A usual video. This one provides little more information about Project IDX.
- Bolt
- Better Bolt + FREE Mistral & Github API : STOP PAYING for V0 & BOLT with this FULLY FREE Alternative by (2 November 2024) ► A Bolt fork that supports any LLM, in particular some free ones.
- OpenHands (formerly OpenDevin)
- OpenHands : This OPENSOURCE AI Coder BEATS CLINE, AIDER & CURSOR!? by (11 November 2024) ► OpenHands, formerly known as OpenDevin, seems to be rather bad…
- Lovable (formerly GPT Engineer)
- I Tried GPT Engineer… and It Changes Everything↓ by (30 June 2023) ► Yet another shitty analysis of the impact of AI, for example, the guy tries to generate a simple program which appears to not be working and he concludes that AI will be able to do a better job in a near future, he considers that junior developers write simple code…
- ↪Is GPT Engineer Actually Useful? 🤨↓ by (22 September 2023) ► A very basic try of GPT Engineer.
- Lovable : This FREE AI Coder BEATS V0 & Bolt? (Generate Full Stack Apps!)↓ by (2 December 2024) ► A usual video: quick ’n dirty presentation of Lovable.
- Llama Coder
- Llama DeepSeek Coder V2 : This FULLY FREE Coder BEATS V0 & BOLT! by (9 January 2025) ► A presentation of Together’s Llama Coder.
- Gemini Code Assist
- Google’s free Gemini Code Assist arrives with sky-high usage limits — Gemini Code Assist lets you do 90 times more than competing tools. by (25 February 2025) ► The title says it all.
- Google's Gemini Code Assist (Fully FREE): RIP Copilot & Cursor! This NEW AI Coder by Google is COOL! by (26 February 2025) ► The usual quick ’n dirty ’s demo.
- Devin
- How Cognition Uses Devin to Build Devin (27 February 2026) ► A description of the many Devin features that the Cognition team uses.
- Codex
- Codex agent internet access by (3 June 2025) ► Codex can access Internet. OpenAI limits the exfiltration risk by only allowing GET, HEAD, and OPTIONS, and using a domain whitelist.
- Autonomous coding agents: A Codex example by (4 June 2025) ► A detailed log of experimenting with Codex for a simple code change.
- BMAD Method
- The BMAD Builder is Here - V6 Alpha: 5 Major Features (Download Now) by (6 October 2025) ► The new features of version 6.
- [OCTO] BMAD & Agents IA : une méthode qui change la donne by (6 March 2026) ► A consultant PO gives some feedback on the team having used BMAD.
- Junie
- Junie The Anti-Vibe Coding IDE by (10 December 2025) ► An advertisement for Junie.
- Gas Town
- Welcome to Gas Town by (1 January 2026) ► A long and rather messy description of an agent orchestrator which, at first look, seem overly complex. AI-generated illustrations do not help. But maybe there is something valuable here…
- Gas Town Emergency User Manual by (13 January 2026) ► Some advice on how to use Gaz Town.
- Stevey’s Birthday Blog by (19 January 2026) ► is interested by the money if he does not have to spent time to get it, piloting AI-agents is tiring, 2026 will be the year a few persons will implement very large projects, how to steer the agents, and the future improvements of Gaz Town.
- Kilo Code
- Live AMA: Meet the Founder of Kilo Code by and (24 November 2025) ► Ask Me Anything with .
- Former GitLab CEO raises money for Kilo to compete in crowded AI coding market by (10 December 2025) ► had invested in Kilo Code.
- Finding Code You Can't Name: Why Semantic Search Changes Everything for Large Codebases by (19 January 2026) ► Kilo Code explains the advantages of semantic search, they will offer that service, which will be free during the beta phase.
- The Free Ride is REALLY Over — Grok Code Fast's free lunch has ended. What now? by (21 January 2026) ► Kilo Code’s users will have to pay. describes the options, explains that some companies are losing money with some subscriptions and will have to limit the consumption. He concludes that Kilo Code allows to easily change of model provider. But changing of IA-IDE is also so simple…
- Optimized Grok Code Fast 1 is Here (and Free) — Our most popular coding model is back in the race, turbo-charged by an American AI lab by (16 February 2026) ► Grok Code Fast 1 is back, but optimised by an anonymous lab.
- No, It’s Not Sonnet 5 (And That’s a Good Thing): Meet Sonnet 4.6 — The latest release from Anthropic is our new daily driver by (17 February 2026) ► Kilo Code is very positive about Claude Sonnet 4.6.
- ShowBoat
- Introducing Showboat and Rodney, so agents can demo what they’ve built by (10 February 2026) ► proposes two tools to test AI-generated code: Showboat to generate a doc which s a runnable demo and Rodney a CLI to drive Chrome.
- Rodney and Claude Code for Desktop by (16 February 2026) ► would like the Claude mobile app to display images as the Claude Code desktop app.
- Two new Showboat tools: Chartroom and datasette-showboat by (17 February 2026) ► adds to Showboat the possibility to push data to a URL, implements datasette-showboat to store this data in datasette, and adds a new tool, Chartroom, to enable the LLM to generate charts.
- Rodney v0.4.0 by (17 February 2026) ► Several new commands have been added to Rodney.