Sakana AI、「The AI Scientist」の第3世代を公開 ——複数エージェントで仮説検証から論文執筆まで、図生成ツール「AutoIllustrate」もオープンソース化 https://www.yayafa.com/2912340/ #AgenticAi #AgenticScientist #agents #AI #ArtificialGeneralIntelligence #ArtificialIntelligence #CodingAgents #GihyoJp #エージェント型AI #人工知能 #技術評論社 #汎用人工知能
#coding-agents — Public Fediverse posts
Live and recent posts from across the Fediverse tagged #coding-agents, aggregated by home.social.
-
Oh, look! Another groundbreaking innovation where AI coding agents are steered by "adaptive workflow rules." 🚀 Because clearly, regular coding wasn't complicated enough without adding a layer of inscrutable AI logic. 🤖🔄 What's next? AI to manage the AI that manages the AI? 🙄
https://github.com/awslabs/aidlc-workflows #AIinnovation #AdaptiveWorkflow #CodingAgents #TechTrends #FutureOfCoding #HackerNews #ngated -
Graphicle, a Rust port of Graphify with memory for coding agents
https://github.com/graphicle-ai/graphicle
Comments: https://news.ycombinator.com/item?id=50088713
#HackerNews #Graphicle #Rust #Graphify #CodingAgents #MemoryManagement
-
Mehr Agenten, mehr geschafft? Die Messungen sagen etwas anderes.
Vals AI hat Einzelagenten gegen Agenten-Teams antreten lassen: Die Teams kosteten das 1,8- bis 5,1-Fache und waren in drei von vier Vergleichen nicht messbar besser. Wo viele Agenten wirklich etwas bringen, wann ein Agent mit Subagenten reicht und warum das dieselben Probleme sind wie in menschlichen Teams:
https://kiberblick.de/artikel/workflow/agenten-teams-kosten-nutzen/
-
JetBrains released the new JetBrains Mellum2.1 model. This open-weight AI significantly enhances coding agents with advanced reinforcement learning features.
#JetBrains #Mellum2 #CodingAgents #ArtificialIntelligence #MachineLearning
https://dailytechnow.com/jetbrains-mellum2-1-model-coding-agents/
-
Living in a Micro-Conway world
Conway’s Law: “organizations which design systems … are constrained to produce designs which are copies of the communication structures of these…https://stephanhagemann.com/posts/2026-10-09-14-17-00-the-micro-conway-effect
-
> Matching decompilation is the process of writing source code (typically C or C++) that compiles to the same exact binary as the original.
Sounds like a perfect task to train a specialised Coding LLM using Reinforcement Learning with Verifiable Rewards (RLVR).
It can also try to choose meaningful function names and comments.
-
AI Agents Are Decompiling Every Video Game
https://www.youtube.com/watch?v=rRQzrQLZRWM
References:
https://universalmodder.org/
https://github.com/rehan-remade/universal-modder
https://web.archive.org/web/20260925153118/https://momo5502.com/posts/2026-08-17-mw2-decompilation/ (deleted post)
https://kotaku.com/zelda-and-banjo-recompiled-devs-go-to-war-with-ai-vibe-coded-slopcomps-2000739936#AI #CodingAgents #Claude #Codex #Decompilation #MatchingDecompilation #VibeCoding #Gaming #Modding #Games
-
4001 comments, 580 files changed, and 318K lines added--officially my new record.
I wish I could share the link, but it's in a private repo. It's an adoption of my coding agent accelerator template repo (https://github.com/franklesniak/copilot-repo-template) into an existing repository.
-
Your AI coding agent may already be integrating your product. 🦾 But what is it learning from?
At PCDD 2026, Developer Advocate Ondřej Chrastina will examine how products and open-source projects can prepare their knowledge for AI coding agents, rather than leaving them to rely on outdated training data.
In his talk, 'AI Best Practices for Your Product', Ondřej compares three very different approaches: CKEditor with agent skills, Markdown documentation and an MCP server; TYPO3 with a slower, interface-first approach; and Drupal with its funded AI initiative and evaluation-backed best practices.
The goal is to provide a practical, PHP-friendly blueprint that developers can apply to their own products and projects.
🔗 https://www.pluswerk.digital/pcdd
@typo3 #PCDD #Pimcore #PHP #AI #CodingAgents #DeveloperExperience #OpenSource #MCP
-
New post: how I make my 20€/month Claude Pro plan last longer in Claude Code.
The main session runs Opus 5.5 at medium effort and makes the decisions. A Sonnet 5.5 subagent writes the code, and Haiku 5.5 subagents scope the task and open the pull request.
Compared with 24 Opus-only sessions, each PR now costs about 40% less, even if the PRs are bigger. My agent files, CLAUDE.md and the numbers are all in the post.
https://www.andreagrandi.it/posts/claude-code-subagents-to-save-usage/
-
👨💻🚀 Oh, the marvels of modern tech: a tool that lets you run coding agents while you snooze in the corner with your laptop closed. Because who wouldn't want to complicate their life by deploying agents on 'dev boxes' all for the thrill of pretending they have an AI workforce? 😂💤
https://berthd.app/ #codingagents #moderntech #AIworkforce #devops #techhumor #HackerNews #ngated -
Everyone is talking about GraphRAG and graph memory for LLMs:
- GraphRAG: an LLM extracts entities and relationships into an explicit knowledge graph.
- Graph memory: facts are edges in a knowledge graph that also remembers when they were true.
- Wiki memory: Markdown pages are the nodes, links are the edges. Karpathy suggests Obsidian’s graph view to see its shape.
- Ontologies: the schema of the graph—which kinds of things exist and how they are allowed to relate.
My post about putting the agent’s memory in the repository is a very, very tiny, limited version of the same idea. Funny how things that have been around for a long time keep coming back.
LLMs are good at plain text files. LLMs are good at Markdown. Markdown can contain links. So of course we write memory into Markdown files, and of course we link those files to each other.
It’s hypertext. It’s the wiki. It’s the World Wide Web… all over again.
#ai #coding agents #memory -
The goal of my everyday life when I was analog-coding was to remove more lines than add. Mostly I was successful and almost always this was net-beneficial.
Since the advent of AI tools this shifted towards finding patterns and making in-repo instructions that make those easy to follow. -
Your AI Agent Says 'Done.' It's Lying.
Patch written. Tests unfinished. I maintain Planning with Files to keep the next step in project Markdown after a reset.
Supported setups can check unfinished phases before stopping. Test evidence still matters.
I opened two docs fixes and a reproducible recovery-evaluation task. Try the demo, pick a contribution, star if useful:
https://planning-with-files.othman-adi.chatgpt.site -
Let the agents push to main
Coding agents are clearly trained on the last 15 years of software development. You can tell because they are scared of main. Ask one to make a change and it reaches for a branch and a pull request. Pull-request-based development won, trunk-based development lost, and the agents learned the lesson: committing straight to main is not safe. Anecdotally, I think it’s exactly what they should be doing, as much as they can. -
We’ve overloaded CI
I can’t remember working professionally without a continuous integration (CI) server. Over time, there are always efforts underway to make CI faster and, at the same time, to cram more tests and gates (linters, validators, architecture checks, etc.) into its run. I’m starting to think we should question the second part. Not the value of CI itself, but rather the habit of turning every check we can think of into a gate.https://web.brid.gy/r/https://stephanhagemann.com/posts/2026-10-01-14-05-00-weve-overloaded-ci
-
If you insist on using a coding agent, add the KISS principle to its system prompt.
-
From the Leanpub Blog: Leanpub Book LAUNCH 🚀 The Code Is Lying: Engineering with Coding Agents by Alen Joy
#books #leanpublishing #selfpublishing #CodingAgents #AIEngineering #AgenticAI #SoftwareArchitecture #CodeReview
-
NEW! Leanpub Book LAUNCH 🚀 The Code Is Lying: Engineering with Coding Agents by Alen Joy
#books #leanpublishing #selfpublishing #CodingAgents #AIEngineering #AgenticAI #SoftwareArchitecture #CodeReview