A massive funding day for Anthropic, Microsoft and OpenAI reshape their partnership, GitHub Copilot moves to usage billing, and a 4TB voice data breach hits AI contractors. Meanwhile, Apple's CEO transition takes shape and Tesla starts shipping steering-wheel-free taxis.
read briefing →Something shifted this week. Anthropic shipped filesystem-based memory for Claude Managed Agents. Stash, an open-source project, dropped a single-binary persistent memory layer that works with any MCP-compatible agent. Meta published a paper showing that structured summaries of past coding rollouts — a form of procedural memory — can push Claude-4.5-Opus from 70.9% to 77.6% on SWE-Bench Verified. These aren't three separate stories. They're the same story: memory is graduating from a kludge you bolt onto an agent…
read full analysis →The UK government admitted this week it underestimated AI datacentre carbon emissions by a factor of 100. So naturally, we asked five brand-new Qwen 3.6 models to channel the energy of a deeply overworked sustainability consultant. The results are carbon-neutral in humor but absolutely toxic in wit.
You are a burned-out sustainability consultant at a Big Four firm. Write a passive-aggressive FAQ for your corporate clients titled 'So Your AI Strategy Has a Carbon Problem: A Guide for People Who Definitely Did the Maths.' Maximum 200 words. Be dry, specific, and funny.
read vibe check →
A towering translucent crystalline structure rises from a dark landscape, its interior filled with thousands of spiraling luminous data streams in amber and electric cyan — memories consolidating into a vast autonomous mind. Deep indigo sky fades to warm gold at the horizon, with faint geometric grid patterns suggesting a world being quietly remapped by intelligence. Portrait composition, cinematic volumetric lighting, concept art style with a palette of deep navy, amber, and cool cyan.