Meta's Devansh Tandon: LLM recommenders will be AI's biggest consumer app

Perplexity Research's hint-guided self-distillation pairs rejection-sampling fine-tuning with corrections of bad tool calls from real user sessions. In a live A/B test, tool-call failures fell from 2.24% to 1.77% — a 21.2% relative reduction versus an earlier checkpoint.

Cognition AI is on track for $1 billion in annualized revenue based on September performance, roughly doubling its run rate from about four months prior, per a person familiar with the matter.

Oracle cut 21,000 jobs this year with a $1.8B severance bill, plus 800 more cuts scheduled for November 13 per WARN filings, alongside record AI data center capex commitments.

Pope Leo XIV said advances in artificial intelligence and other technologies must remain at the service of people rather than become instruments of domination and injustice.

Raghav Saboo, Staff ML Engineer and tech lead for Search & Personalization at DoorDash, argues the bottleneck in marketplace discovery is semantic understanding, not ranking. He cites the failure case of a gluten-free pasta query surfacing regular spaghetti because a ranker trained on engagement favors high-selling items.