Issue #004 · April 16, 2026
Weekly signal digest. The reads worth your time, with our take. No noise, no recaps — just what's actually moving the conversation.
tl;drPaper of the month:A benchmark of 56 model organisms with hidden behaviors finds that auditing-tool rankings depend heavily on how the organism was trained — and the investigator agent, not the tools, is the bottleneck.Research highlights:Linear “emotion vectors” in Claude c. Read more →
Why AI is a hype and a hatAnthropic's revenue grew from $1B to $19 in 14 months (Zoom's previous record: from $0.3B to $2.6B in ~12 months during COVID)OpenAI raised $110B last month (previous record Saudi Aramco IPO = $29.4B in 2019, and this is not even an IPO)NVIDIA invested $. Read more →
The IMF published a fresh analysis confirming AI is not uniformly replacing jobs — it's bifurcating the workforce. Workers who learn to orchestrate AI agents are capturing disproportionate value.
Anthropic accidentally shipped a `.map` file with the Claude Code npm package, exposing full source code. Key findings from analysis: - Anti-distillation layer: When enabled, Claude Code injects fake.
More than 30 OpenAI and Google DeepMind employees signed onto a statement supporting Anthropic's lawsuit against the Defense Department after the agency labeled the AI firm a supply-chain risk, according to court filings. Read more →
IMF Blog, January 2026
Research Lab Morning Brief