Boston is the SF of the East Coast

Incredible work but the suite is only ~1300 tasks. In the economy people do probably hundreds of thousands of atomic tasks across multiple variations. Gives you an idea on how much work there is still to be done. Really awesome first step

AI labs are increasingly organizing their research efforts around use cases/product (ex. financial services, deep research) rather than capabilities (ex. computer use, coding). This is the right direction long term from a business POV.

Sonnet 4.5 is extremely good. I've noticed it has particularly good traces and "print messages" that are informative but not overwhelming or slop.

It's funny how almost a decade later, the frontier labs are back at the same spot, building RL gyms

Underrated learning from Y Combinator is setting a high bar for talent by comparing potential hires to batchmates

Seems like a natural next step in organizing tools. As the catalog of tools/MCPs grow, giving models unneeded tools can degrade performance / increase cost. A structured way to only deliver necessary tools at run-time is really intuitive

I wonder what they're cooking.

Introducing Yutori Navigator 31 years ago, the modern web era began with Netscape Navigator. Today, we’re introducing Yutori Navigator — a web agent that autonomously navigates websites on its own cloud browser to complete tasks for you. Navigator achieves pareto-domination

Less formal communication = speed = better business

Update: we benchmarked Opus 4.5 on Navi-Bench. • Overall: 74.2% (Sonnet 4.5) → 78.0% (Opus 4.5) • Breakdown per site • Biggest gain: Apartments 76.7% → 90.0% • small bumps on Craigslist and Resy • slight regression on Google Flights