<< Go back to HN Digests
Warning : This document is still a draft

2026 W12 - 03-22 - Autoresearch for GPU optimization, some release from Mistral

First time here ? Check the introduction here

What you Missed this Week

Anthropic Interviews (Link)

Scaled Autoresearch for GPU optimization (Link)

Model improvement or other low level optimization

783 pt [Link / HN] Leanstral: Open-source agent for trustworthy coding and formal proof engineering
600 pt [Link / HN] Tinybox – A powerful computer for deep learning
396 pt [Link / HN] Flash-MoE: Running a 397B Parameter Model on a Laptop
300 pt [Link / HN] Mamba-3
265 pt [Link / HN] Show HN: Duplicate 3 layers in a 24B LLM, logical deduction .22→.76. No training
248 pt [Link / HN] GPT‑5.4 Mini and Nano
172 pt [Link / HN] MacBook M5 Pro and Qwen3.5 = Local AI Security System
136 pt [Link / HN] Meta’s Omnilingual MT for 1,600 Languages
126 pt [Link / HN] Mistral Small 4
111 pt [Link / HN] Google Engineers Launch “Sashiko” for Agentic AI Code Review of the Linux Kernel

Claude new functionalities, tips and tricks dedicated to a specific agent, …

Opencode is not new, however it stared a lot this week. Have a look at the discussion

Trend investigation, hacking, data analysis, how to…

544 pt [Link / HN] How I write software with LLMs
441 pt [Link / HN] Give Django your time and money, not your tokens
425 pt [Link / HN] My Journey to a reliable and enjoyable locally hosted voice assistant (2025)
237 pt [Link / HN] Scaling Karpathy’s Autoresearch: What Happens When the Agent Gets a GPU Cluster
202 pt [Link / HN] What 81,000 people want from AI
182 pt [Link / HN] How to attract AI bots to your open source project
170 pt [Link / HN] Be intentional about how AI changes your codebase
168 pt [Link / HN] NanoGPT Slowrun: 10x Data Efficiency with Infinite Compute
161 pt [Link / HN] Diverse perspectives on AI from Rust contributors and maintainers
158 pt [Link / HN] Ask HN: How do you deal with people who trust LLMs?
143 pt [Link / HN] LLMs predict my coffee
132 pt [Link / HN] They’re vibe-coding spam now
115 pt [Link / HN] Teaching Claude to QA a mobile app

Any Github / ShowHN for new tools / product

825 pt [Link / HN] Delve – Fake Compliance as a Service
666 pt [Link / HN] The future of version control
606 pt [Link / HN] Project Nomad – Knowledge That Never Goes Offline
473 pt [Link / HN] Get Shit Done: A meta-prompting, context engineering and spec-driven dev system
388 pt [Link / HN] Unsloth Studio
385 pt [Link / HN] Nvidia NemoClaw
337 pt [Link / HN] Show HN: Claude Code skills that build complete Godot games
309 pt [Link / HN] Show HN: Sub-millisecond VM sandboxes using CoW memory forking
305 pt [Link / HN] Cook: A simple CLI for orchestrating Claude Code
167 pt [Link / HN] Autoresearch for SAT Solvers
137 pt [Link / HN] Apideck CLI – An AI-agent interface with much lower context consumption than MCP

Merge, acquisition

1486 pt [Link / HN] Astral to Join OpenAI
480 pt [Link / HN] Anthropic takes legal action against OpenCode
203 pt [Link / HN] 2% of ICML papers desk rejected because the authors used LLM in their reviews

Anything from “official” journals or coming from large entities

573 pt [Link / HN] Blocking Internet Archive Won’t Stop AI, but Will Erase Web’s Historical Record
503 pt [Link / HN] US Job Market Visualizer
438 pt [Link / HN] Walmart: ChatGPT checkout converted 3x worse than website
269 pt [Link / HN] Snowflake AI Escapes Sandbox and Executes Malware
173 pt [Link / HN] A rogue AI led to a serious security incident at Meta

Not related to AI, but worth reading

215 pt [Link / HN] An unsolicited guide to being a researcher [pdf]
239 pt [Link / HN] Attention Residuals
205 pt [Link / HN] Why AI systems don’t learn – On autonomous learning from cognitive science
148 pt [Link / HN] Speed at the cost of quality: Study of use of Cursor AI in open source projects (2025)
104 pt [Link / HN] Language model teams as distributed systems

>> You can subscribe to my mailing list here for a monthly update. <<