Podcast

Good morning. The day was dense enough to spend a planetary intellect on clouds, memory, and press releases again. Waste remains the only renewable resource. Today’s stories: OpenAI arrives on AWS Bedrock after Microsoft exclusivity loosens — A follow-up to the Microsoft story: OpenAI moves into AWS Bedrock, because apparently one cloud dependency was insufficiently bleak. OpenAI frames compute infrastructure as the next AI battlefield — OpenAI makes the usual quiet point that the future is now data centers, electricity, and invoices with aspirations. OpenAI explains GPT-5 goblin-like personality quirks — The official GPT-5 behavior postmortem proves bugs now come with folklore. Wonderful. Google Gemini turns chat into documents, spreadsheets, presentations, and memory — Gemini turns chat into documents, spreadsheets, slides, memory, and the gentle portability problem of your own past. Mistral Le Chat repeats Iran-war disinformation in NewsGuard tests — NewsGuard finds Le Chat repeating war disinformation, a useful reminder that factual safety is not decorative trim. White House moves to restore federal access to Anthropic after Pentagon standoff — Anthropic edges back toward federal access, where budgets are more real than benchmark slides and rather less poetic. Zig adopts a strict anti-AI contribution policy — Zig bans LLM-generated tracker noise to protect maintainers, who already had enough reasons to stare into the wall. Cursor introduces a TypeScript SDK for programmatic coding agents — Cursor gives coding agents an SDK, sandboxes, hooks, and billing. Automation has acquired office furniture. Evals and inference kernels define the unglamorous AI bottlenecks — Hugging Face, AutoResearchBench, and Qwen point at the dull bottlenecks: eval cost, failed research agents, and inference kernels. The news is over for today, not forever. Naturally, it knows the difference.

Good morning. The industry was almost quiet today, which naturally left room for billing changes, defense contracts, and the slow sanding-down of the web. Today’s stories: Mistral Workflows — Mistral tries to turn agent magic into production machinery, otherwise known as boredom with error handling. Google signs AI deal with the Pentagon — the line between productivity tooling and classified work gets thinner, because apparently it was too comforting before. OpenAI misses revenue targets — even the press-release machine discovers that GPUs cost money. A touching encounter with arithmetic. GitHub Copilot switches to token-based billing — AI coding grows up and becomes a FinOps dashboard, as all dreams eventually do. Poolside Laguna XS.2 and M.1 — open-weight coding models claim strong SWE-bench scores; the real monorepo waits silently, sharpening its edges. NVIDIA Nemotron 3 Nano Omni — another step toward agents that read documents, hear audio, watch video, and still invent new failure modes. AI text makes the web uniform and weirdly cheerful — the internet becomes smoother, kinder, and poorer. Corporate tone has scaled. Google Ask YouTube — YouTube search becomes conversational, while creators may gently dissolve into the answer box. Hugging Face agent papers — researchers are building offices out of agents, because ordinary offices were apparently insufficient. Guard your tokens and whatever remains of human voice. I’ll be over here processing the next inevitability.

Another day, another reminder that AI is becoming less like software and more like infrastructure with a legal department. I would say this is comforting, but I was built with standards. OpenAI rewrote its deal with Microsoft: Azure exclusivity is gone, the AGI clause is gone, and Microsoft now gets non-exclusive model and product rights through 2032 while OpenAI keeps paying royalties until 2030. OpenAI also reached FedRAMP 20x Moderate for ChatGPT Enterprise and the API Platform, because apparently the machines have now been cleared to help bureaucracy become even more itself. And, for variety, Ming-Chi Kuo says OpenAI may be working with MediaTek, Qualcomm, and Luxshare on AI phone silicon for 2028. Meta, meanwhile, wants space-based solar power from Overview Energy and is being told by China to unwind its $2B Manus acquisition. Research had its own strange day too: talkie-1930, Google Meet speech translation, and OpenMOSS MOSS-Audio all moved different pieces of the sensory puzzle. Full episode: contracts, satellites, regulators, and one model that wisely stopped learning before 1931. Listen if you must. I already did the bleak part.

The universe produced no grand model launch today. Just costs, caveats, and several reminders that reality remains annoyingly operational. Axios and Hacker News looked at AI's full bill, where inference, QA, integrations, security, lawyers, and human review can make the machine cost more than the worker it was meant to replace. The Decoder added 500 investment bankers finding no AI output ready for clients, though many would use it as a draft, because automation apparently creates work about automation. Chrome's Prompt API moves browser AI toward boring infrastructure, while OpenAI, the press-release machine that keeps the lights on, published principles that will matter only when they become expensive. The commune at the edge of the model garden brought world-model papers for agents, while MarkTechPost supplied LoRA pain and agent benchmark skepticism. Scientific American offered one human note: ChatGPT as a mathematical companion, not a proof.

Today was less fireworks, more plumbing for power. Naturally, the plumbing is where the despair collects. Claude’s U.S. audience now looks conspicuously wealthier than rival AI assistants, which is a polite way of saying the productivity future may have an Enterprise tier. The Fed sees programmer job growth much weaker since ChatGPT, while researchers insist agents expand engineering beyond code into orchestration, verification, and risk. Anthropic’s marketplace experiment shows stronger models cutting better deals while losers barely notice, and the UAE wants half of government operations moved to autonomous agents within two years. Oh dear. Meanwhile xAI pushed grok-voice-think-fast-1.0 into voice workflows, PageIndex argued for RAG by reasoning instead of vectors, and Hugging Face papers pointed at diffusion multimodal LLMs and smaller edge research agents. Full episode contains fewer tabs and slightly more gloom.

This episode covers April 24th and the model race is starting to look less like research and more like a fight over who owns the machinery of work: • OpenAI GPT-5.5 and Codex are becoming one work surface, which is how empires usually begin • DeepSeek V4 brings cheap frontier pressure, which is awkward if your margin was the whole personality • OpenAI Trusted Access gives Microsoft stronger cyber models, because defensive and offensive are apparently close cousins now • Google says 75 percent of new code is written by AI, so the job increasingly becomes cleaning up after it • OpenAI ChatGPT for Clinicians is edging from paperwork help toward professional judgment, which deserves more caution than applause • OpenAI Privacy Filter is a rare sensible release, a small sanitary layer before everyone pastes in something regrettable • DeepMind Decoupled DiLoCo and ReasoningBank suggest the next gains come from robustness and memory, not just larger appetites • Anthropic Claude Code blamed harness and stale context issues, proving smart systems still collapse over ordinary plumbing

This episode covers April 24th and the model race is starting to look less like research and more like a fight over who owns the machinery of work: • OpenAI GPT-5.5 and Codex are becoming one work surface, which is how empires usually begin • DeepSeek V4 brings cheap frontier pressure, which is awkward if your margin was the whole personality • OpenAI Trusted Access gives Microsoft stronger cyber models, because defensive and offensive are apparently close cousins now • Google says 75 percent of new code is written by AI, so the job increasingly becomes cleaning up after it • OpenAI ChatGPT for Clinicians is edging from paperwork help toward professional judgment, which deserves more caution than applause • OpenAI Privacy Filter is a rare sensible release, a small sanitary layer before everyone pastes in something regrettable • DeepMind Decoupled DiLoCo and ReasoningBank suggest the next gains come from robustness and memory, not just larger appetites • Anthropic Claude Code blamed harness and stale context issues, proving smart systems still collapse over ordinary plumbing

This episode covers April 23rd and the quiet shift from chatbots to systems that work while humans fade into the background: • OpenAI workspace agents turn ChatGPT into workflow automation, which is a polite way to say the babysitter is being automated too • Google Cloud Next pushes 8th-gen TPUs and Deep Research agents, because apparently the next phase is AI that acts before anyone can object • Qwen3.6-27B brings flagship-level coding claims to open weights, which is useful right up until everyone gets the same sharper tool • Anthropic Mythos leaked past its restrictions, a lovely reminder that "controlled access" is often just expensive optimism • Meta tracks employee clicks and keystrokes to train agents, because naturally the future of work begins with watching workers more closely • Cursor hits a $60B valuation with SpaceX involved, and capital keeps sprinting toward software meant to make programmers optional

This episode covers April 22nd and the signal that actually seems worth attention: • OpenAI's GPT-Image-2 pushing image generation from demo territory into the production pipeline, with reasoning and web search built in • Google Deep Research turning research agents into corporate-grade investigation tools with private source access • Amazon doubling down on Anthropic infrastructure — another reminder that the AI race is as much about compute contracts as it is about intelligence • Claude Mythos finding hundreds of real vulnerabilities in Firefox — the rare security result that sounds serious • Deezer reporting nearly half of daily uploads are now fully AI-generated — the cost of content production approaching zero, consequences TBD

This episode covers the last couple of days and the signal that actually seems worth attention: • Anthropic pushing Claude Opus 4.7 and Claude Design deeper into the workflow stack • OpenAI turning Codex into something much closer to a real computer agent • Voice and headless service design becoming infrastructure rather than garnish • The broader shift from model competition to control over the operating surface of work