<< Go back to HN Digests
Warning : This document is still a draft
Warning : This document is still a draft
2026 W1 - 02-xx
First time here ? Check the introduction here
What you Missed this Week
This week, before Wednesday I would have said “nothing happening.” Yet we got the U.S. Department of War trying to make deals with Anthropic and OpenAI. Anthropic says no, OpenAI says “yes.” Then a “cancel ChatGPT” movement started. At the same time, the Claude app becomes the first app on the U.S. App Store.
A lot of “politically” related blogs. Some about mass surveillance using AI (Link), or deanonymisation using LLM (Link).
- Out of that, past week, I discovered LLM diffusion models. This week, there is again a new article on that from another model provider (Link).
- Fun project: minimal transformer to add two digits (HN).
- If you do not know how a transfomer works,
micro GPTis a 200 lines implementation you can look at (Link) - New edition of “your missing CS semester” (Link), updated with coding agents
- Again another article about AI and economic growth (Link)
Articles
Politics
DoW
- 2909 pt [Link / HN] Statement from Dario Amodei on our discussions with the Department of War
- 1390 pt [Link / HN] OpenAI agrees with Dept. of War to deploy models in their classified network
- 1349 pt [Link / HN] I am directing the Department of War to designate Anthropic a supply-chain risk
- 1162 pt [Link / HN] Statement on the comments from Secretary of War Pete Hegseth
- 798 pt [Link / HN] We do not think Anthropic should be designated as a supply chain risk
- 328 pt [Link / HN] President Trump bans Anthropic from use in government systems
- 203 pt [Link / HN] US Military leaders meet with Anthropic to argue against Claude safeguards
Opinions
Cancel ChatGPT
- 1904 pt [Link / HN] OpenAI – How to delete your account
- 1059 pt [Link / HN] How do I cancel my ChatGPT subscription?
Other News
- 722 pt [Link / HN] Anthropic drops flagship safety pledge
- 558 pt [Link / HN] OpenAI raises $110B on $730B pre-money valuation
- 308 pt [Link / HN] Making Wolfram tech available as a foundation tool for LLM systems
- 288 pt [Link / HN] AI Added ‘Basically Zero’ to US Economic Growth Last Year, Goldman Sachs Says
- 255 pt [Link / HN] Claude becomes number one app on the U.S. App Store
- 225 pt [Link / HN] OpenAI resets spending expectations, from $1.4T to $600B
- 211 pt [Link / HN] ChatGPT Health fails to recognise medical emergencies – study
- 156 pt [Link / HN] Anthropic announces proof of distillation at scale by MiniMax, DeepSeek,Moonshot
- 146 pt [Link / HN] Palantir’s AI Is Playing a Major Role in Tracking Gaza Aid Deliveries
- 137 pt [Link / HN] Vibe coded Lovable-hosted app littered with basic flaws exposed 18K users
- 113 pt [Link / HN] IBM Plunges After Anthropic’s Latest Update Takes on COBOL
Coding Agent
### Claude
- 572 pt [Link / HN] Get free Claude max 20x for open-source maintainers
- 552 pt [Link / HN] Anthropic ditches its core safety promise
- 551 pt [Link / HN] Switch to Claude without starting over
- 543 pt [Link / HN] Claude Code Remote Control
- 424 pt [Link / HN] What does “ 2>&1 “ mean?
- 199 pt [Link / HN] Why XML tags are so fundamental to Claude
- 98 pt [Link / HN] Show HN: Claude-File-Recovery, recover files from your ~/.claude sessions
MCP
- 548 pt [Link / HN] MCP server that reduces Claude Code context consumption by 98%
- 320 pt [Link / HN] Making MCP cheaper via CLI
- 267 pt [Link / HN] WebMCP is available for early preview
Others
AI Mmodels and Training
- 1795 pt [Link / HN] Microgpt
- 349 pt [Link / HN] Mercury 2: Fast reasoning LLM powered by diffusion
- 345 pt [Link / HN] The First Fully General Computer Action Model
- 324 pt [Link / HN] Show HN: Steerling-8B, a language model that can explain any token it generates
- 315 pt [Link / HN] Show HN: Moonshine Open-Weights STT models – higher accuracy than WhisperLargev3
- 244 pt [Link / HN] Smallest transformer that can add two 10-digit numbers
- 236 pt [Link / HN] Unsloth Dynamic 2.0 GGUFs
- 172 pt [Link / HN] Show HN: AI Timeline – 171 LLMs from Transformer (2017) to GPT-5.3 (2026)
Learning
- 468 pt [Link / HN] Decision trees – the unreasonable power of nested decision rules
- 451 pt [Link / HN] The Missing Semester of Your CS Education – Revised for 2026
- 314 pt [Link / HN] Can you reverse engineer our neural network?
- 260 pt [Link / HN] Microgpt explained interactively
- 243 pt [Link / HN] 10-202: Introduction to Modern AI (CMU)
- 149 pt [Link / HN] From Noise to Image – interactive guide to diffusion
Blogs
Tools for Agents
- 605 pt [Link / HN] Pi – A minimal terminal coding harness
- 370 pt [Link / HN] “Car Wash” test with 53 models
- 369 pt [Link / HN] When does MCP make sense vs CLI?
- 337 pt [Link / HN] Don’t trust AI agents
- 220 pt [Link / HN] Show HN: A real-time strategy game that AI agents can play
- 207 pt [Link / HN] Verified Spec-Driven Development (VSDD)
- 205 pt [Link / HN] Show HN: Emdash – Open-source agentic development environment
- 201 pt [Link / HN] Show HN: enveil – hide your .env secrets from prAIng eyes
- 130 pt [Link / HN] Launch HN: Cardboard (YC W26) – Agentic video editor
- 122 pt [Link / HN] Just-bash: Bash for Agents
- 115 pt [Link / HN] Right-sizes LLM models to your system’s RAM, CPU, and GPU
- 84 pt [Link / HN] Show HN: Context Mode – 315 KB of MCP output becomes 5.4 KB in Claude Code
- 85 pt [Link / HN] Show HN: Badge that shows how well your codebase fits in an LLM’s context window
- 81 pt [Link / HN] SynthID: A tool to watermark and identify content generated through AI
- 76 pt [Link / HN] Building secure, scalable agent sandbox infrastructure
Tool made with Agents
- 1105 pt [Link / HN] I’m helping my dog vibe code games
- 536 pt [Link / HN] How we rebuilt Next.js with AI in one week
- 536 pt [Link / HN] I built a demo of what AI chat will look like when it’s “free” and ad-supported
- 290 pt [Link / HN] Show HN: Now I Get It – Translate scientific papers into interactive webpages
- 150 pt [Link / HN] Implementing a Z80 / ZX Spectrum emulator with Claude Code
Analysis
- 655 pt [Link / HN] OpenAI, the US government and Persona built an identity surveillance machine
- 607 pt [Link / HN] What Claude Code chooses
- 350 pt [Link / HN] Large-Scale Online Deanonymization with LLMs
- 270 pt [Link / HN] AIs can’t stop recommending nuclear strikes in war game simulations
- 252 pt [Link / HN] Addressing Antigravity Bans and Reinstating Access
- 226 pt [Link / HN] We gave terabytes of CI logs to an LLM
- 266 pt [Link / HN] LLM=True
- 169 pt [Link / HN] Let’s discuss sandbox isolation
- 143 pt [Link / HN] An autopsy of AI-generated 3D slop
- 127 pt [Link / HN] Learnings from 4 months of Image-Video VAE experiments
Opinion
- 401 pt [Link / HN] Will vibe coding end like the maker movement?
- 322 pt [Link / HN] What AI coding costs you
Pessimists
- 481 pt [Link / HN] How will OpenAI compete?
- 319 pt [Link / HN] The Eternal Promise: A History of Attempts to Eliminate Programmers
- 180 pt [Link / HN] AI is making junior devs useless
- 147 pt [Link / HN] The Future of AI
>> You can subscribe to my mailing list here for a monthly update. <<