<< Go back to HN Digests
Warning : This document is still a draft
Warning : This document is still a draft
2026 W10 - 03-08 - GPT-5.4 is out!
First time here? Check the introduction here
What you Missed this Week
This week, GPT-5.4 is out! It looks like OpenAI is catching up to Opus from Anthropic, but not yet.
Another release this week is from Google. Not an AI model but a set of CLI tools to interact with all their services. Useful to integrate in an agent.
My top preferences of this week go to Tropes (the DON’Ts to provide to the LLMs) and Error 406 (the reason your LLM PR was rejected). For LLM practitioners, these highlight the common mistakes that LLMs can make.
Articles
Model Training / Efficiency / Hardware
- 1014 pt [Link / HN] GPT-5.4
- 569 pt [Link / HN] Show HN: I built a sub-500ms latency voice agent from scratch
- 469 pt [Link / HN] How to run Qwen 3.5 locally
- 416 pt [Link / HN] Qwen3.5 Fine-Tuning Guide
- 395 pt [Link / HN] GPT‑5.3 Instant
- 271 pt [Link / HN] A CPU that runs entirely on GPU
- 185 pt [Link / HN] NanoGPT Slowrun: Language Modeling with Limited Data, Infinite Compute
- 184 pt [Link / HN] Autoresearch: Agents researching on single-GPU nanochat training automatically
- 178 pt [Link / HN] Sarvam 105B, the first competitive Indian open source LLM
- 169 pt [Link / HN] Show HN: Timber – Ollama for classical ML models, 336x faster than Python
- 87 pt [Link / HN] Phi-4-reasoning-vision and the lessons of training a multimodal reasoning model
Coding Agents
- 626 pt [Link / HN] Hardening Firefox with Anthropic’s Red Team
- 341 pt [Link / HN] LLM Writing Tropes.md
- 188 pt [Link / HN] Parallel coding agents with tmux and Markdown specs
- 89 pt [Link / HN] Sem – Semantic version control. Entity-level diffs on top of Git
- 76 pt [Link / HN] Claude Code LSP
Learning
- 833 pt [Link / HN] Claude’s Cycles [pdf]
- 541 pt [Link / HN] Agentic Engineering Patterns
- 118 pt [Link / HN] SWE-CI: Evaluating Agent Capabilities in Maintaining Codebases via CI
- 97 pt [Link / HN] Neural Boids
- 57 pt [Link / HN] Language Model Contains Personality Subnetworks
Personal Blogs / Opinions
Tool for Agents
- 947 pt [Link / HN] Google Workspace CLI
- 622 pt [Link / HN] Agent Safehouse – macOS-native sandboxing for local agents
- 412 pt [Link / HN] If AI writes code, should the session be part of the commit?
- 322 pt [Link / HN] Show HN: Jido 2.0, Elixir Agent Framework
- 218 pt [Link / HN] A tool that removes censorship from open-weight LLMs
- 145 pt [Link / HN] Show HN: PageAgent, A GUI agent that lives inside your web app
- 100 pt [Link / HN] Show HN: Claude-replay – A video-like player for Claude Code sessions
- 92 pt [Link / HN] Show HN: P0 – Yes, AI can ship complex features into real codebases
Optimists
- 1047 pt [HN] Tell HN: I’m 60 years old. Claude Code has re-ignited a passion
- 449 pt [Link / HN] LLMs work best when the user defines their acceptance criteria first
- 304 pt [Link / HN] A standard protocol to handle and discard low-effort, AI-Generated pull requests
- 290 pt [Link / HN] Files are the interface humans and agents interact with
- 399 pt [Link / HN] Relicensing with AI-Assisted Rewrite
- 269 pt [Link / HN] Anthropic, please make a new Slack
- 255 pt [Link / HN] We should revisit literate programming in the agent era
- 219 pt [Link / HN] We might all be AI engineers now
- 204 pt [Link / HN] My spicy take on vibe coding for PMs
- 199 pt [Link / HN] A case for Go as the best language for AI agents
- 162 pt [Link / HN] You need to rewrite your CLI for AI agents
- 96 pt [Link / HN] We automated everything except knowing what’s going on
Pessimists
- 664 pt [Link / HN] The L in “LLM” Stands for Lying
- 381 pt [Link / HN] The changing goalposts of AGI and timelines
- 377 pt [Link / HN] Anthropic Cowork feature creates 10GB VM bundle on macOS without warning
- 304 pt [Link / HN] When AI writes the software, who verifies it?
- 142 pt [Link / HN] Claude Code wiped our production database with a Terraform command
- 94 pt [Link / HN] Will Claude Code ruin our team?
Politics / Strategy
- 801 pt [Link / HN] Dario Amodei calls OpenAI’s messaging around military deal ‘straight up lies’
- 782 pt [Link / HN] Something is afoot in the land of Qwen
- 627 pt [Link / HN] Where things stand with the Department of War
- 430 pt [Link / HN] Pentagon formally labels Anthropic supply-chain risk
News
-
1100 pt [Link / HN] “Microslop” filtered in the official Microsoft Copilot Discord server
- 630 pt [Link / HN] A GitHub Issue Title Compromised 4k Developer Machines
- 605 pt [Link / HN] Ars Technica fires reporter after AI controversy involving fabricated quotes
- 566 pt [Link / HN] US economy unexpectedly sheds 92k jobs in February
- 468 pt [Link / HN] Uploading Pirated Books via BitTorrent Qualifies as Fair Use, Meta Argues
- 362 pt [Link / HN] India’s top court angry after junior judge cites fake AI-generated orders
- 280 pt [Link / HN] OpenClaw surpasses React to become the most-starred software project on GitHub
- 225 pt [Link / HN] Jensen Huang says Nvidia is pulling back from OpenAI and Anthropic
- 193 pt [Link / HN] AI-generated art can’t be copyrighted after Supreme Court declines review
- 185 pt [Link / HN] US tech firms pledge at White House to bear costs of energy for datacenters
- 162 pt [Link / HN] AMD will bring its “Ryzen AI” processors to standard desktop PCs for first time
- 160 pt [Link / HN] Claude struggles to cope with ChatGPT exodus
- 159 pt [Link / HN] Training students to prove they’re not robots is pushing them to use more AI
- 109 pt [Link / HN] Palantir and Anthropic AI helped the US hit 1k Iran targets in 24 hours
- 81 pt [Link / HN] 1.5 Million Users Leave ChatGPT
Labor Impact
- 329 pt [Link / HN] Labor market impacts of AI: A new measure and early evidence
- 225 pt [Link / HN] BMW Group to deploy humanoid robots in production in Germany for the first time
- 170 pt [Link / HN] Oracle may slash up to 30k jobs to fund AI data-centers as US banks retreat
- 101 pt [Link / HN] I don’t know if my job will still exist in ten years
- 80 pt [Link / HN] You are going to get priced out of the best AI coding tools (2025)
- 73 pt [Link / HN] Why developers using AI are working longer hours
- 61 pt [Link / HN] AI doesn’t replace white collar work
- 59 pt [Link / HN] The Case of the Disappearing Secretary
Unclassified
>> You can subscribe to my mailing list here for a monthly update. <<