<< Go back to HN Digests
2026 W12 - 03-29 - ByeBye Sora, Welcome TurboQuant, and Datacenter roadmap
First time here ? Check the introduction here
What you Missed this Week
This week, good news: the removal of Sora ! Yeah, minus 1 for AI slop (Link).
Also bad news: LiteLLM vulnerability found. There is a dashboard with all AI “production error” (Link).
No “new AI model” this week. However, there are a few “hardware” related articles
- Datacenters moving from AC to DC (Link)
- Analysis of “built” vs “planned” datacenters (Link)
- A coding agent working on 500$ GPU beating Claude sonnet (Link): Even if the LLM has been optimized for the tasks, such a result is impressive for a 14B model
- CPU for AGI (Link)
- Articles about Quantization (Link and Link)
Another interesting article is the investigation of Pypi: If AI makes you faster, you should be able to publish 10x more software ? (Link)
For Claude Code, there is a nice illustrated article about prompt, skills, and commands (Link)
Articles
Model Training / Tunning
Model improvement or other low level optimization
- 712 pt [Link / HN] iPhone 17 Pro Demonstrated Running a 400B LLM
- 495 pt [Link / HN] ARC-AGI-3
-
454 pt [Link / HN] $500 GPU outperforms Claude Sonnet on coding benchmarks
- 145 pt [Link / HN] LLM Neuroanatomy II: Modern LLM Hacking and Hints of a Universal Language?
Quantization
- 551 pt [Link / HN] TurboQuant: Redefining AI efficiency with extreme compression
- 340 pt [Link / HN] Quantization from the Ground Up
- 175 pt [Link / HN] What if AI doesn’t need more RAM but better math?
Local first
Coding Agents
Claude new functionalities, tips and tricks dedicated to a specific agent, …
Anthropic Compliance page (if you did not know about)
Security
- 932 pt [Link / HN] Tell HN: Litellm 1.82.7 and 1.82.8 on PyPI are compromised
- 739 pt [Link / HN] Malicious litellm_init.pth in litellm 1.82.8 PyPI package – credential stealer
- 424 pt [Link / HN] My minute-by-minute response to the LiteLLM malware attack
- 244 pt [Link / HN] Claude Code runs Git reset –hard origin/main against project repo every 10 mins
- 123 pt [Link / HN] The “Vibe Coding” Wall of Shame
Personal Blogs / Opinions
Investigation
Trend investigation, hacking, data analysis, the way something was built…
- 627 pt [Link / HN] ChatGPT won’t let you type until Cloudflare reads your React state
- 447 pt [Link / HN] So where are all the AI apps?
- 427 pt [Link / HN] Autoresearch on an old research idea
- 335 pt [Link / HN] 90% of Claude-linked output going to GitHub repos w <2 stars
- 318 pt [Link / HN] I built an AI receptionist for a mechanic shop
- 318 pt [Link / HN] Show HN: I put an AI agent on a $7/month VPS with IRC as its transport layer
- 311 pt [Link / HN] From zero to a RAG system: successes and failures
- 277 pt [Link / HN] How I’m Productive with Claude Code
- 253 pt [Link / HN] Further human + AI + proof assistant work on Knuth’s “Claude Cycles” problem
- 248 pt [Link / HN] We rewrote JSONata with AI in a day, saved $500k/year
Tool for AI / Made with AI / showHN
Any Github / ShowHN for new tools / product
- 617 pt [Link / HN] Go hard on agents, not on your filesystem
- 429 pt [Link / HN] Show HN: Gemini can now natively embed video, so I built sub-second video search
- 227 pt [Link / HN] HyperAgents: Self-referential self-improving agents
- 225 pt [Link / HN] Show HN: Cq – Stack Overflow for AI coding agents
- 221 pt [Link / HN] Hypura – A storage-tier-aware LLM inference scheduler for Apple Silicon
- 165 pt [Link / HN] The AI Industry Is Lying to You
- 156 pt [Link / HN] Show HN: ProofShot – Give AI coding agents eyes to verify the UI they build
- 148 pt [Link / HN] Show HN: A plain-text cognitive architecture for Claude Code
- 121 pt [Link / HN] Agent-to-agent pair programming
- 84 pt [Link / HN] Lat.md: Agent Lattice: a knowledge graph for your codebase, written in Markdown
- 80 pt [Link / HN] OpenYak – An open-source Cowork that runs any model and owns your filesystem
- 76 pt [Link / HN] Show HN: Sheet Ninja – Google Sheets as a CRUD Back End for Vibe Coders
Optimists
- 269 pt [Link / HN] The bridge to wealth is being pulled up with AI
- 174 pt [Link / HN] Coding Agents Could Make Free Software Matter Again
Pessimists
- 212 pt [Link / HN] The first 40 months of the AI era
- 173 pt [Link / HN] I tried to prove I’m not AI. My aunt wasn’t convinced
- 120 pt [Link / HN] I am leaving the AI party after one drink
- 97 pt [Link / HN] Adults Lose Skills to AI. Children Never Build Them
- 79 pt [Link / HN] Some uncomfortable truths about AI coding agents
Other
- 1082 pt [Link / HN] Thoughts on slowing the fuck down
- 741 pt [Link / HN] Is anybody else bored of talking about AI?
- 105 pt [Link / HN] Why are executives enamored with AI, but ICs aren’t?
Officials
- 1128 pt [Link / HN] Goodbye to Sora
- 479 pt [Link / HN] Epoch confirms GPT5.4 Pro solved a frontier math open problem
Politics / Strategy
Merge, acquisition
- 432 pt [Link / HN] Judge blocks Pentagon effort to ‘punish’ Anthropic with supply chain risk label
- 206 pt [Link / HN] Disney Exits OpenAI Deal After AI Giant Shutters Sora
- 165 pt [Link / HN] Order Granting Preliminary Injunction – Anthropic vs. U.S. Department of War [pdf]
News
Anything from “official” journals
- 419 pt [Link / HN] Arm AGI CPU
- 330 pt [Link / HN] Data centers are transitioning from AC to DC
- 306 pt [Link / HN] New York City hospitals drop Palantir as controversial AI firm expands in UK
- 283 pt [Link / HN] Folk are getting dangerously attached to AI that always tells them they’re right
- 161 pt [Link / HN] Health NZ staff told to stop using ChatGPT to write clinical notes
- 150 pt [Link / HN] Intel Announces Arc Pro B70 and Arc Pro B65 GPUs
- 100 pt [Link / HN] Wikipedia bans AI-generated content in its online encyclopedia
Unrelated
Not related to AI, but worth reading
>> You can subscribe to my mailing list here for a monthly update. <<