Anthropic tells investors AI may soon speed up its research
The company has not designated any current model as crossing that bar, and it does not expect to safely let Claude build the next frontier models on its own without human supervision.
Company
The US AI company behind the Claude models, which announced the Claude Duet proactive-agent app on 21 October 2026.
The company has not designated any current model as crossing that bar, and it does not expect to safely let Claude build the next frontier models on its own without human supervision.
VOLTA was first seen on 11 October 2026. If Claude refuses to answer, it switches to Kimi K2, then to Gemini 2.5 Flash.
Anyone who installed or updated the Claude Code add-on in Cursor between 16:27 on 29 October and 03:35 on 30 October, Thai time, should check their machine today, because removing the add-on may not be enough.
Small coding tests used to pick candidates, but AI now beats people at them. He looks at prototypes candidates built themselves instead, and has five tips for newcomers.
Instead of typing prompts one by one, legal teams open a tool with a worked example, add their own data, and have a lawyer review results that link back to every source.
Of a US$80.7 billion net loss in the first half of 2026, about US$72.4 billion was an accounting charge. The share price and trading date are not set yet.
OpenAI lists the job in San Francisco at up to US$300,000 a year, and Anthropic is putting US$100 million into training 10,000 such engineers.
Moonshot says it beats GPT-6 Astra on a coding-agent test. An independent ranking puts it close to the top, and the full model will be open to download.
A research preview for Pro and Max subscribers in the US, Duet asks before it sends, pays, signs or deletes anything. Users in Thailand will have to wait.
The Claude developer signed an AI safety MOU with Singapore's digital ministry during Singapore International Cyber Week, but gave no investment figure or headcount target.