# the-batch

The Batch 349: two opposite signals on one table. Ng on AI-native teams (engineer:PM 1:1, generalists win); Meta's first Superintelligence Labs model — Muse Spark, closed, fourth, one-third the tokens. Plus Eli Lilly's $2.75B Insilico bet and Google's Persona Generators on the PM bottleneck.

Andrew Ng says he has not only stopped writing code, he has long stopped reading generated code. He now directs agents at a higher abstraction layer and sees X Engineers emerging inside business functions. A radical programming forecast from AI's most influential educator.

Liquid AI's LFM2.5-1.2B-Thinking (1.17B param, 32K context) runs on-device (<1GB mem). Claims to match/beat Qwen3-1.7B on reasoning, with faster decoding & fewer tokens. Strong for tool-calling/data extraction, but weaker on knowledge-heavy tasks.

SleepFM, a multimodal sleep foundation model (585k+ hrs PSG data), predicts 130+ conditions, showing strong results for stroke, heart failure, & dementia risk. It's about earlier, scaled risk detection, not replacing doctors.

White House EO mandates 180-day AI Action Plan, shifting US strategy from risk-first to competition-first. Agencies to revise policies impeding competitiveness, re-centering on practical AI progress over hypothetical risks.

Andrew Ng attended the Sundance Film Festival to understand Hollywood's AI anxieties — copyright fears, union fights, and a deep sense of powerlessness — but also found surprising common ground.

SpaceX acquired xAI to create a $1.25T private company. The cash helps xAI compete, but Musk's bigger bet is solar-powered data centers in space, where heat dissipation and debris may be harder than training LLMs.

Former OpenAI policy chief Miles Brundage founded Averi, a nonprofit backed by 28 institutions including MIT and Stanford. Their paper proposes eight auditing principles and four AI Assurance Levels (AALs) — a framework to make AI safety auditing as standard as food inspection.

Harvard's Dr. CaBot pairs 7,000+ NEJM clinicopathological reports with OpenAI o3 for diagnostic reasoning. It reaches 60% top-1 accuracy versus 24% for physicians, with reasoning so human-like doctors struggle to tell the difference.