Loading activity…
@fhuszar
Founder & Supreme Leader of Technical Staff at Reasonable. Professor on leave from @Cambridge_CL. Alum of @Twitter, Magic Pony, @Balderton
AI
ReasonableIO says it tested whether the model could write challenging Verus proofs for TLA+ specifications.
AI
A user credits professors and students with foundational AI work before Nvidia and Google embraced the field.
AI
A user recalls the dataset’s October 2021 release and says all AIs at the time struggled with its problems.
AI
A user said their office floor was tilted five degrees that week and asked readers to guess why.
AI
A user pokes fun at confidence in agent sandboxing with a two-line caption and a video link.
AI
The authors frame the guide around people asking Claude to write TLA+, arguing that it isn't a “silver bullet” for formal verification.
