Friday gave the chatbot a Bloomberg terminal and sent another robot to the front.

Wall Street gets its own ChatGPT

OpenAI launched ChatGPT for Financial Services Thursday, bringing GPT‑6 Astra into a product specifically designed around investment banking and equity research.

The system can connect financial professionals to sources including LSEG, PitchBook, Daloopa, Crunchbase and Quartr while also working with company-specific data and templates. OpenAI says it can help build financial models, perform research and generate client materials such as pitchbooks.

The important word here is not finance. It is integration.

A generic chatbot knows something about companies. A professional financial system needs permission controls, citations, licensed datasets, audit logs, encryption and the ability to understand which employee is allowed to see which information. OpenAI says the product includes role-based access controls and compliance-oriented audit exports.

That is the same maturation pattern we saw yesterday when payment companies began developing identity standards for shopping agents. The clever model is becoming one component inside a much larger boring system. And boring systems are how technology gets trusted with expensive things.

Astra does not need to become the CFO. If it reliably saves analysts several hours building a model or assembling a pitchbook, the economic consequence can already be enormous.

The chatbot is slowly disappearing inside the software people actually use to work.

China’s newest AI-chip stock opened 188% higher

Chinese AI-chip company Enflame opened 188% above its IPO price Friday in its debut on Shanghai’s STAR Market. The Tencent-backed company raised about $912 million through the offering.

That first print should not be confused with business fundamentals. Enflame remains unprofitable. Its 2025 revenue was about 990 million yuan, and nearly 84% of it came from Tencent, which is also the company’s largest shareholder. One analyst quoted by Reuters cautioned that the enormous opening gain reflected speculative demand for new listings as much as confidence in the company itself.

But the industrial signal is real. China is attempting to build a domestic ecosystem capable of replacing or reducing dependence on American AI hardware, particularly Nvidia.

Enflame plans to spend most of its IPO proceeds developing fifth- and sixth-generation AI processors, software and large-scale computing systems.

This is what semiconductor independence looks like before independence actually exists: lots of companies, lots of capital, lots of overlapping architectures, and absolutely no guarantee which ones survive.

Silicon Valley had a similar phase. Most companies disappeared. A few became infrastructure.

China is apparently willing to buy quite a lot of lottery tickets.

NASA taught an open model to look for somewhere useful to park on the Moon

NASA and IBM have released the NASA-IBM Lunar Foundation Model, an open-source AI system built specifically for analyzing lunar data.

The model was trained on more than 30 layers of information from nine instruments aboard four NASA missions, including the Lunar Reconnaissance Orbiter.

Its jobs are refreshingly concrete: find craters, study volcanic features, help identify safe landing terrain, and look for potential ice deposits in permanently shadowed regions.

In benchmark tests, NASA and IBM say the model identified important lunar features as much as 23% more accurately than commonly used existing approaches.

The ice question matters enormously. Accessible lunar water could potentially provide drinking water, oxygen and—after splitting water into hydrogen and oxygen—ingredients for rocket propellant.

NASA’s Artemis program currently plans to return astronauts to the Moon in 2028 as part of a longer effort toward sustained lunar operations and eventually Mars.

This is another good antidote to the idea that AI progress means one gigantic universal chatbot swallowing every application. Sometimes the useful model is highly specialized. It does not need to write poetry.

It needs to look at several decades of orbital observations and say: “That shadow over there might contain water.”