Keep pulling the thread on Michael Truohr.
Cursor's top priority for the year is to build the best coding model.
Cursor is pricing its Composer 2.5 model at $0.50 per million input tokens and $2.50 per million output tokens, which is half the cost of Opus 4.7 or GPT-5.5.
Cursor claims its Composer 2.5 model is 10 times more efficient than similarly capable models.
Cursor is training a new model from scratch using XAI's Colossus-2 training cluster.
Cloudflare's review of Anthropic's Mythos model concludes that it is a "real step forward" and a "different kind of tool" compared to previous models.
Cloudflare found that Anthropic's Mythos model can create an exploit chain by synthesizing multiple attack primitives, rather than just detecting single bugs.
According to Cloudflare, Anthropic's Mythos model can generate functional exploits for vulnerabilities, making it more useful than models that only detect potential issues.
Elon Musk lost his lawsuit against OpenAI and Sam Altman.
In the lawsuit against OpenAI, the jury found that Elon Musk's claim of breach of charitable trust was barred by the statute of limitations.
Model-focused labs are expected to develop their own agentic harnesses, while harness-first labs like Cursor and Cognition are expected to develop their own models.
In January, Cursor CEO Michael Truohr told staff it was "wartime" because the company's business model was being eroded by competition from both model labs and other harness providers.
Cursor's Composer 2.5 model scored 69.3% on TerminalBench 2.0, slightly behind Opus 4.7's score of 69.4%.