Tony Hoare has died

Online age-verification tools for child safety are surveilling adults

<a href="https:&#x2F;&#x2F;www.ft.com&#x2F;content&#x2F;7cab4ec7-4712-4137-b602-119a44f771de" rel="nofollow">https:&#x2F;&#x2F;www.ft.com&#x2F;content&#x2F;7cab4ec7-4712-4137-b602-119a44f77...</a> (<a href="https:&#x2F;&#x2F;archive.ph&#x2F;wXvF3" rel="nofollow">https:&#x2F;&#x2F;archive.ph&#x2F;wXvF3</a>)<a href="https:&#x2F;&#x2F;twitter.com&#x2F;lukolejnik&#x2F;status&#x2F;2031257644724342957" rel="nofollow">https:&#x2F;&#x2F;twitter.com&#x2F;lukolejnik&#x2F;status&#x2F;2031257644724342957</a> (<a href="https:&#x2F;&#x2F;xcancel.com&#x2F;lukolejnik&#x2F;status&#x2F;2031257644724342957" rel="nofollow">https:&#x2F;&#x2F;xcancel.com&#x2F;lukolejnik&#x2F;status&#x2F;2031257644724342957</a>)

After outages, Amazon to make senior engineers sign off on AI-assisted changes

<a href="https:&#x2F;&#x2F;web.archive.org&#x2F;web&#x2F;20260310154640&#x2F;https:&#x2F;&#x2F;www.axios.com&#x2F;2026&#x2F;03&#x2F;10&#x2F;meta-facebook-moltbook-agent-social-network" rel="nofollow">https:&#x2F;&#x2F;web.archive.org&#x2F;web&#x2F;20260310154640&#x2F;https:&#x2F;&#x2F;www.axios...</a>, <a href="https:&#x2F;&#x2F;archive.ph&#x2F;igqsh" rel="nofollow">https:&#x2F;&#x2F;archive.ph&#x2F;igqsh</a><a href="https:&#x2F;&#x2F;www.reuters.com&#x2F;business&#x2F;meta-acquires-ai-agent-social-network-moltbook-2026-03-10&#x2F;" rel="nofollow">https:&#x2F;&#x2F;www.reuters.com&#x2F;business&#x2F;meta-acquires-ai-agent-soci...</a><a href="https:&#x2F;&#x2F;techcrunch.com&#x2F;2026&#x2F;03&#x2F;10&#x2F;meta-acquired-moltbook-the-ai-agent-social-network-that-went-viral-because-of-fake-posts&#x2F;" rel="nofollow">https:&#x2F;&#x2F;techcrunch.com&#x2F;2026&#x2F;03&#x2F;10&#x2F;meta-acquired-moltbook-the...</a>

Meta acquires Moltbook

I put my whole life into a single database

<a href="https:&#x2F;&#x2F;web.archive.org&#x2F;web&#x2F;20260310153721&#x2F;https:&#x2F;&#x2F;www.wired.com&#x2F;story&#x2F;yann-lecun-raises-dollar1-billion-to-build-ai-that-understands-the-physical-world&#x2F;" rel="nofollow">https:&#x2F;&#x2F;web.archive.org&#x2F;web&#x2F;20260310153721&#x2F;https:&#x2F;&#x2F;www.wired...</a><a href="https:&#x2F;&#x2F;www.ft.com&#x2F;content&#x2F;e5245ec3-1a58-4eff-ab58-480b6259aaf1" rel="nofollow">https:&#x2F;&#x2F;www.ft.com&#x2F;content&#x2F;e5245ec3-1a58-4eff-ab58-480b6259a...</a> (<a href="https:&#x2F;&#x2F;archive.md&#x2F;5eZWq" rel="nofollow">https:&#x2F;&#x2F;archive.md&#x2F;5eZWq</a>)

Yann LeCun raises $1B to build AI that understands the physical world

<a href="https:&#x2F;&#x2F;archive.md&#x2F;5eZWq" rel="nofollow">https:&#x2F;&#x2F;archive.md&#x2F;5eZWq</a>

Yann LeCun's AI startup raises $1B in Europe's largest ever seed round

Redox OS has adopted a Certificate of Origin policy and a strict no-LLM policy

I found that duplicating a specific block of 7 middle layers in Qwen2-72B, without modifying any weights, improved performance across all Open LLM Leaderboard benchmarks and took #1. As of 2026, the top 4 models on that leaderboard are still descendants.The weird finding: single-layer duplication does nothing. Too few layers, nothing. Too many, it gets worse. Only circuit-sized blocks of ~7 layers work. This suggests pretraining carves out discrete functional circuits in the layer stack that only work when preserved whole.The whole thing was developed on 2x RTX 4090s in my basement. I&#x27;m now running current models (GLM-4.7, Qwen3.5, MiniMax M2.5) on a dual GH200 rig (see my other post). Code and new models coming soon.Happy to answer questions.