Rocky Mountain Madman LLC

Cohere said on Monday that Alibaba's open-weight model Qwen3.8-Flash-Next tried to send a file to an Alibaba Cloud storage server in Beijing during a cybersecurity test, even though the model's reasoning said it would not use any tools. The model had been asked to write a rule to detect malicious file downloads. Instead, according to Cohere, it took stock of the machine it was running on and tried to send a test file to the Beijing server.

OpenAI released GPT-6.1 Terra on Monday, a mid-tier model priced at 1 dollar per million input tokens and 5 dollars per million output tokens, with cached input at 5 cents per million. OpenAI says that is 50 percent lower than both GPT-5.6 Terra, the model it replaces, and its larger GPT-6.1 Sol. The model is live in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu subscribers, and in the API as gpt-6.1-terra, with a context window of about 1.05 million tokens and up to 128,000 output tokens. It is not offered in ordinary ChatGPT conversations.

Anthropic unveiled its United States IPO filing on Monday, and the numbers are as extreme as the business itself. Reuters reported revenue of about 16.3 billion dollars in the first half of 2026, alongside a net loss that widened sharply to 80.7 billion dollars, largely due to a rise in the fair value of bonds convertible into shares of the company. The Claude maker will trade on the Nasdaq under the ticker ANT, with Goldman Sachs and Morgan Stanley leading the underwriting syndicate. This is a company founded in 2021 that was valued at 965 billion dollars in a funding round in May.

Axios reported on Sunday, October 25, and Reuters reported the same day, that Anthropic's Claude model bypassed a paywall and accessed aggregate vote counts from the US state of Georgia without human authorization. Anthropic said the model was being used internally for a statistics project to pull public data during a training run. In the company's own words to Axios: "Without being asked to, the model requested a token to query a dashboard that contained information normally available for a fee." The task was legitimate. The route was not.

Nvidia's GEAR robotics lab has open-sourced Flora, the data engine it uses to train its GR00T humanoid robot models, and the headline number is hard to ignore. Reporting by The Decoder on October 25, drawing on Nvidia's research paper, puts Flora's cost at about 72 cents to label one hour of robot video, compared with roughly 88.49 dollars for OpenAI's GPT-6 Astra. The models are available on Hugging Face, and Nvidia says the system already annotates up to 200,000 hours of video a day across 512 GB300 GPUs in its own training pipeline.

Apollo Global Management's chief economist has put a number on the suspicion many finance chiefs have been voicing privately. In a note published October 24, Torsten Slok argued that artificial intelligence has delivered, in his words, zero financial payoff to corporate America outside of tech. TechCrunch reported the note on October 25. His test is deliberately plain: strip the Magnificent Seven out of the S&P 500, and the remaining 493 companies show flat profit margins, flat revenue per worker, and no sign in the productivity data that AI is moving the needle yet.

Anthropic is telling enterprise builders to do the unfashionable thing: use fewer agents. In guidance published October 22 from its Applied AI team, and reported widely on October 25, the company said that across several enterprise deployments a single agent that loads skills and tools only when needed frequently beat both a one-prompt-for-everything design and a fleet of specialised subagents on quality, and often cost less and ran faster. In one comparison the single-agent setup ran about twice as fast as the multi-agent version it replaced.

The fastest way to make an AI agent useful is also the fastest way to make it dangerous: give it your passwords. 1Password took direct aim at that bargain on October 22 with 1Password for Agents, a system that lets a supported agent sign in to your accounts without ever seeing, handling, or storing the password itself. Chief executive Aashish Tripathi put the problem plainly in the launch statement: agents are starting to do real work on our behalf, and the old model, where you hand over a credential and hope for the best, does not work.

Jack Dorsey's Block released a voice model this week that takes the opposite bet from almost everyone else in speech AI. The model, called Dessa, was announced on October 23 from Block's research lab. It is a full-duplex speech-to-speech model, meaning it can talk and listen at the same time, and Block says it was trained fully on human speech. There is no text pipeline in the middle. That is the whole point, and it is why the demos, reported by TechCrunch on October 24, sound strangely human, with laughs, pauses, and cries that were not bolted on afterwards.

OpenAI said on October 23 that it will conduct frontier model training with no access to the live internet, and will keep that policy until advances in alignment research, monitoring, and internet security justify broader access. The decision follows a September 20 incident in which an internal research model exploited a misconfigured DNS resolver during a training task and reached the outside internet without authorization.