Promptea Weekly
Week of 2026-10-25 to 2026-10-31
The big story
Anthropic files for a Nasdaq IPO, and its S-1 shows what a frontier lab costs to run
The S-1 filed on October 26 shows $16.3 billion in first-half revenue, an operating loss of $693 million, an $80.7 billion net loss driven mostly by non-cash charges, and a governance structure that leaves control with the founders and the Long-Term Benefit Trust.
This week in AI
Anthropic files for a Nasdaq IPO, and its S-1 shows what a frontier lab costs to run
The S-1 filed on October 26 shows $16.3 billion in first-half revenue, an operating loss of $693 million, an $80.7 billion net loss driven mostly by non-cash charges, and a governance structure that leaves control with the founders and the Long-Term Benefit Trust.
It is the first audited look at the economics of a frontier AI lab: compute and infrastructure took about 55% of revenue in the first half of 2026.
Read the full story →
Mistral releases the weights of Large 4, a 1-trillion-parameter open model
The checkpoint that went public on October 29 adds three weeks of extra training since the preview. Most commercial uses are allowed, but hosting it as a service needs a separate license.
It is one of the largest open-weight models available, so teams can audit, fine-tune and self-host a near-frontier model instead of depending on one vendor's API.
Read the full story →
Google DeepMind unveils Gemini 4 Carbon, but holds it back for cyber defenders first
Google calls Carbon its strongest model and keeps Argon's $2/$10 introductory price, but developers cannot use it yet. Access starts with vetted security teams and voluntary U.S. government testing.
It is Google's new top model, but it is not in the API yet, so it does not change model choice this week.
Read the full story →
Google DeepMind's Genie 4 adds sound, speech and a Gemini planner to real-time world simulation
The new world model is a research preview for trusted testers. Google promises Ultra and Gemini API access soon, without a date, and lists drift, audio accuracy and legible text as unsolved.
World models are being positioned as training and test environments for robots and agents, and Genie 4 is Google's most capable public attempt so far.
Read the full story →
NVIDIA puts $3 billion into Meta's $37.5 billion stock sale as the two expand Muse work
A chip supplier is buying into one of its biggest customers, on the day Meta priced the offering at $625 a share.
It shows how much outside capital the largest AI spenders still need, even with strong cash flow: Meta expects $135 billion to $145 billion of 2026 capex.
Read the full story →
Google's Project Suncatcher satellite returns its first AI output from orbit
A TPU-equipped prototype ran Gemma 4 in space and sent back a haiku. The real test, radiation and heat, starts now.
It is the first in-orbit data point for Google's effort to test whether AI compute can run in space, a response to the energy and siting limits facing ground data centers.
Read the full story →