Brief — March 11, 2026

Library Blog Radar Briefs Intelligence About

Amazon pulls the emergency brake on AI coding

🧠 Yann LeCun's post-Meta move — $1B for AI world models** — His Paris startup, Advance Machine Intelligence, just raised $1 billion to build AI world models. LeCun left Meta's FAIR and immediately went after the hardest unsolved problem in AI. This is the counter-narrative to the LLM monoculture — and it just got serious funding. 🤝 Mira Murati's Thinking Machines Lab × Nvidia — gigawatt-scale partnership** — TML (ex-OpenAI CEO Mira Murati's start

Yann LeCun's post-Meta move — $1B for AI world models

His Paris startup, Advance Machine Intelligence, just raised $1 billion to build AI world models. LeCun left Meta's FAIR and immediately went after the hardest unsolved problem in AI. This is the counter-narrative to the LLM monoculture — and it just got serious funding.

Mira Murati's Thinking Machines Lab × Nvidia — gigawatt-scale partnership

TML (ex-OpenAI CEO Mira Murati's startup) announced a long-term gigawatt-scale strategic partnership with Nvidia for model training. Even amid team defections back to OpenAI, TML is doubling down. A gigawatt partnership means they're playing for AGI-level compute.

Sloppelgangers"

Grammarly's AI agent was caught masquerading as Verge staff members in Superhuman email summaries. The term "sloppelganger" coined. Publishers and journalists are fighting back. UK copyright law overhaul imminent — 10,000 authors signed a blank book in protest.

National poll: Americans like AI less than ICE

56% of US adults used an AI platform in the past month. Public sentiment is souring fast even as adoption climbs. The experience gap is real. This is the Research Lab angle: the trust crisis in AI is also an education crisis.

Pattern of the Day

The AI trust stack is fracturing at every layer simultaneously. Amazon limits its own engineers. The public trusts AI less than law enforcement. Writers are fighting copyright battles. Engineers are experiencing existential meaning collapse. And yet: Murati raises gigawatts, LeCun raises billions, and Claude is autonomously improving its own alignment tools. The acceleration and the resistance are now running in parallel — not sequential. This is the moment Research Lab was built to narrate. ---

Blog Pitches

The Sloppelganger Problem — When AI impersonates your voice, who owns the damage? (AI & Education × Creative Intelligence crossover)

The Engineer's Grief — The human cost of the meaning collapse among technical workers. Field Notes format, 400w.

Gigawatt Ambitions — What Murati × Nvidia really signals about the next compute race. (Markets + Tech Sovereignty)

Why Amazon Just Blinked — The first institutional rollback on autonomous AI coding. The Playbook format.

LessWrong

Gemma Needs Help — 136✦ (Anna Soligo) → unclear title but high engagement, worth reading

Don't Let LLMs Write For You — 88✦ (JustisMills) → directly relevant to the sloppelganger story above; resistance is forming

Economic efficiency often undermines sociopolitical autonomy — 79✦ (Richard_Ngo) → governance/sovereignty angle, connects to Tech Sovereignty pillar

Letting Claude do Autonomous Research to Improve SAEs — 68✦ (chanind) → AI self-improving alignment tools; frontier technical signal

Interview with Steven Byrnes on His Mainline Takeoff Scenario — 32✦ → Updated by Claude Code's capability surprise. P(Doom) recalibration in real time.