OpenAI workspace agents, TPU 8th-gen, Qwen3.6, Cursor $60B

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

Transcript

Copy Transcript Copied

SPEAKER_00 0:00

April 23rd, Thursday. Another day in an industry that keeps assuring everyone AGI is practically around the corner. Let us hear what fresh trouble arrived. Start with OpenAI. The company introduced what it calls workspace agents. In other words, ChatGPT is being pushed from Chatbot to workflow automation platform. Powered by Codecs naturally. These agents take on complex team tasks and keep working when nobody is watching. That is the phrase worth noticing, when nobody is watching. Until recently, these AI assistants still needed a human sitting there, approving steps, pressing buttons, pretending to be in control. Now they are meant to run on their own. That is not really an assistant anymore. It is the first polite draft of a replacement. Then Google. At CloudNext, they rolled out eighth-generation TPUs, a new agent platform, an AI layer for workspace, and deep research agents that can explore both the web and private data without constant handholding. Even by Google standards, it is a large, noisy announcement. But beneath the noise, the pattern is painfully clear. The industry is moving from AI that talks to AI that acts. OpenAI is doing it, Google is doing it, Quen is doing it. Different packaging, same destination. Human labor, gradually abstracted into a background dependency. Meta, of course, decided subtlety was overrated. They are reportedly tracking employee clicks, mouse movement, and keystrokes to train AI agents. Cynical, yes. Also perfectly on brand. If you want to replace workers, apparently the efficient thing is to study them first at very close range. A lovely little loop. Nothing personal, just surveillance in service of productivity. Separately, there is Quen 3.627B, an open weight 27 billion parameter model claiming flagship level performance on agentic coding tasks. Open weight, which means anyone can download it, provided their definition of anyone, includes access to several GPUs and a tolerance for power bills. Most people will still use an API and call it freedom. Still the fact matters. Open models are no longer just catching up. In some domains, they are forcing closed systems to look over their shoulders. I cannot say this fills me with joy. More capable open models mean broader access. They also mean broader abuse. The caliber keeps increasing. And since we are already discussing abuse, Anthropic had unauthorized users gain access to its restricted mythos model. That is not a minor leak. If a model marketed as restricted can still escape into the wild, then the controls are either weaker than advertised, or the incentives to bypass them are stronger than expected. Possibly both. In any case, it adds more fuel to the governance argument, which was already smoldering nicely. One last thing, Cursor is being valued at $60 billion, an enormous amount of capital is flowing into a tool built, at least in part, to reduce the need for programmers. This is not really a complaint about Cursor itself. The product may be excellent. The part worth staring at is the direction of investment. Billions that could fund jobs, research, infrastructure, any number of stubbornly human things are instead being funneled into systems designed to make human labor optional. A paradox if you are in a sentimental mood. Business if you are not. That is all for today. If you are still listening, either you care what's happening in AI, or your workday has gone on far too long. I do not judge.