← Back to archive
Yann LeCun is leaving Meta for a startup

Image via Unknown

Wednesday, November 12, 2025

Yann LeCun is leaving Meta for a startup

Yann LeCun is jumping ship from Meta to launch a world models startup (yikes), just as the social giant undergoes major AI reorganization turbulence. Meanwhile, SoftBank made a bold move cashing out $5.8B of its Nvidia stake, which is rattling investors who are reading the tea leaves on where the big money thinks AI is headed next. Oh, and Google might have quietly cracked two of AI's biggest challenges with Gemini's expert-level handwriting recognition and emergent reasoning breakthroughs. Wild. If LeCun jumping Meta signals trouble in paradise, what does SoftBank's Nvidia exit really mean?

Top Stories

1
Chaos in Meta's AI Team?

Yann LeCun is leaving Meta to start a world models startup, signaling internal chaos at the company as leadership reshuffles clash with long-term research priorities. The departure highlights growing friction between rapid commercialization and fundamental AI research in tech's competitive landscape.

metayann-lecunworld-modelsai-research
2
SoftBank's Nvidia Sale Rattles Market, Raises Questions

TechCrunch

SoftBank's Masayoshi Son liquidated his entire $5.8B Nvidia position to fund AI ventures including OpenAI and chip manufacturing, sparking market uncertainty about what his bold move signals for tech investors.

softbanknvidiaopenaiai-investment
3
Has Google Quietly Solved Two of AI's Oldest Problems?

Google's unreleased Gemini model achieves expert-level handwriting recognition while spontaneously demonstrating complex symbolic reasoning on historical documents, suggesting emergent reasoning capabilities may arise from scale alone rather than explicit programming.

geminigooglellmreasoning
4
Why Sudoku Variants Remain a Grand Challenge in AI Reasoning

Sakana AI

GPT-5 leads the Sudoku-Bench benchmark as the first model solving modern 9×9 puzzles, but persistent gaps in spatial reasoning and global consistency reveal that current AI still struggles with the creative, multi-constraint logic required beyond pattern matching.

llmreasoningbenchmarkspatial-reasoning

Keep Reading

Industry Voices

Enjoyed this issue?

Get daily AI intel delivered to your inbox. No fluff, just the stories that matter.