◆ DAILY BRIEFING
Sunday, March 22, 2026
-
Engineer METR just quantified what every senior engineer suspected: ~50% of AI-generated PRs that pass SWE-bench automated grading would fail human code review.
Half of AI-generated PRs that pass SWE-bench would fail human code review (METR), Cursor's new model is quietly built on Chinese open-source Kimi K2.5, and multi-agent architectures are driving 5,800x…
Read full briefing → -
Security Claude Code Channels now bridges Telegram and Discord directly to live code execution sessions — protected only by a sender allowlist and pairing code.
AI coding agents now bridge messaging platforms directly to code execution, run scheduled tasks overnight without human oversight, and process proprietary source code through silently-swapped Chinese…
Read full briefing → -
Data Science Multi-agent workflows are driving 1,000–6,000x increases in per-user token consumption — and NVIDIA just valued Groq at $20B to solve it.
The inference era arrived with hard numbers this week: multi-agent workflows drive 1,000–6,000x more tokens per user than chat, SWE-bench overstates coding agent quality by 2x, and Anthropic flipped t…
Read full briefing → -
Product Microsoft pulled Copilot from five Windows 11 apps after 'near-universal' backlash, Xbox's new leader is marketing 'No Soulless AI Slop,' and Alibaba/Tencent lost $66B in 24 hours for shipping AI without monetization clarity — while NVIDIA's own chip-design team proved AI fails entirely without traceability, even internally.
The 'add AI everywhere' era ended this week from both directions: consumers systematically reject it (Microsoft retreated from five apps, Xbox banned 'AI slop,' Hachette pulled a book on suspicion alo…
Read full briefing → -
Leader NVIDIA just paid $20B for inference chip maker Groq and announced 35x throughput gains over its own Blackwell — while real-world token consumption among agentic early adopters has exploded 6,000x in two years.
The AI industry hit a defining inflection this week: NVIDIA paid $20B for Groq and announced 35x inference throughput gains while token demand among early agentic adopters exploded 6,000x — but simult…
Read full briefing → -
Investor Microsoft just retreated on Copilot after 'near-universal' negative user feedback, NVIDIA's own chip-design AI failed until they rebuilt their entire org around it, and three sources independently confirm copilot ROI is hitting a hard ceiling at ~30% task acceleration.
AI's application layer just hit its first structural wall — Microsoft retreated on Copilot after 'near-universal' backlash, copilot ROI is capping at 30%, and consumer cultural hostility is hardening…
Read full briefing →