Jensen Huang has stepped down as Nvidia CEO amid spy allegations and a classified memo leak, sending shockwaves through AI markets already reeling from a broad sell-off. The All-In crew dissects who wanted him out, what it means for "the Bank of Nvidia," and China's strategic positioning. This is the biggest leadership story in tech in years.
Harvard mathematician Levent Alpöge says he's proved the Riemann Hypothesis working alongside Claude, one of the most consequential unsolved problems in mathematics carrying a $1M prize. A separate human-only proof attempt is also circulating — making this a rare and surreal head-to-head. Both claims are unverified.
A new study finds majority-Black neighborhoods are 2x more likely to host data centers — a figure that rises to 4.2x in majority-Asian neighborhoods. The AI infrastructure boom is following old redlining fault lines, concentrating noise, heat, and power strain in communities with the least political leverage.
PostgreSQL's annual release slipped from September to November this year, with the team citing a flood of AI-generated bug reports and patches that needed sorting. It's a preview of what open source maintainers will increasingly face: AI as both accelerant and noise machine.
The FBI ran a double-sided sting — impersonating both ransomware victims and the threat actors targeting them — to catch an alleged fraud scheme in real time. The operation reveals how law enforcement is adopting adversarial AI-era tactics to infiltrate criminal networks from multiple angles simultaneously.
An AI system discovered a bug in Bitcoin's network: repeating a single ID 50,000 times can stall nodes entirely. The find underscores a new reality — AI is now doing security-critical infrastructure audits that human researchers miss.
The account flagged Levent Alpöge's claim that he proved the Riemann Hypothesis collaborating with Claude — a $1M Millennium Prize problem unsolved since 1859. It matters not just if the proof holds, but because it marks the first credible public claim of a frontier AI co-authoring a major mathematical breakthrough. The math community is now on watch.
Mollick floated a sharp hypothesis: the slowdown in open-weights model progress may trace directly back to OpenAI and Anthropic cracking down on distillation from their frontier models. If closed labs are the primary training signal for open models, cutting that pipe quietly throttles the whole ecosystem — no announcement required.
Jensen Huang's exit put a spotlight on Epoch AI's ongoing investigation into how advanced Nvidia chips are reaching China despite export controls. Their research suggests the leakage is structural, not incidental — meaning no single CEO departure fixes the underlying enforcement gaps.
DAIR.AI flagged a provocative finding: for long-horizon agent tasks, an append-only text log plus grep beats sophisticated summarization and structured memory systems. Sometimes the dumbest architecture wins — and that has real implications for how agentic systems should be designed at scale.
The week's signal is coming from multiple directions at once, but it all points at the same underlying tension: AI is now operating at the boundary of civilization-scale systems. A mathematician co-proves one of history's hardest problems with Claude. An AI audits Bitcoin's core network and finds a hole humans missed. PostgreSQL ships late because maintainers are drowning in AI-generated contributions. These aren't demos. They're infrastructure events.