download MP3 · two-host conversation
download MP3 · single-anchor news read
Meta’s AI unit is in open revolt after 8,000 layoffs and a soul-crushing workload, while US tech shed 38k jobs in May—AI the top cited reason. Meanwhile, Mistral eyes €20B and Prometheus raises $12B to build an "artificial general engineer," signaling where capital is flowing.
Meta laid off 8,000 employees in May, with AI explicitly cited as a driver. The cuts are part of a broader pattern: US tech shed 38,242 jobs in May, with AI the most common reason given by employers.
Inside Meta’s remaining AI unit—6,500 strong—engineers describe a "soul-crushing gulag," signaling deep morale issues even among those retained. A growing tribe of jobless techies is now stuck in Silicon Valley’s new reality, per the LA Times.
Beyond Meta, Webflow conducted abrupt layoffs in late May, and Cloudflare’s CEO publicly declared AI has made entire categories of workers obsolete. CEOs are increasingly using AI as a cover story for layoffs, per MIT researchers.
Capital is flooding into AI-native infrastructure and physical-world automation. Jeff Bezos’s Prometheus raised $12B at a $41B valuation to build an "artificial general engineer" for heavy industry and drug design. Mistral is rumored to be raising €3B at a €20B valuation, nearly doubling its prior round.
Startups are also mobilizing. Theker raised $85M to build reconfigurable factory robots, while Expanse (YC P26) and Hyper (YC P26) launched to unlock GPU capacity and power agentic development, respectively. InsForge, an open-source Heroku for coding agents, signals demand for tooling around autonomous workflows.
On the labor side, HackerNews’s June "Who is Hiring?" thread remains active, and roles like Agent Engineer, Eval Engineer, and AI Operations are recurring themes in AI dev stack discussions.
For developers, Claude Code remains a benchmark for agentic coding, though open alternatives like NousCoder-14B and Kimi K2.7 Code (up to 12x cheaper) are gaining traction. Graph databases for agent memory are emerging as a practical pattern.
For non-engineers, Anthropic’s Cowork (Claude Desktop agent) and Google’s Gemini-SQL2 (text-to-SQL leader at 80% accuracy on BIRD) highlight demand for no-code AI tools and structured data skills. Microsoft’s SkillOpt shows how optimized Markdown prompts can boost agent performance.
If you’re in a redundant role, pivot to agentic workflows (Claude Code, Cowork), eval/benchmarking, or AI-native infra—or upskill in graph databases and text-to-SQL. Capital is flowing to physical-world automation; follow it.