cofounder and ceo @vogentai @ycombinator, @caltech
San Francisco
Joined August 2024
- Tweets47
- Following81
- Followers223
- Likes606
Jagath Vytheeswaran retweeted
macOS sandboxes. Now on Hyperbrowser.
Give your agents a desktop with browser access, native apps, and Terminal control.
Jagath Vytheeswaran retweeted
For a long time, Yann LeCun and others believed in gradient-based planning, but it didn’t work very well … until now. Here’s how we did it using incredibly simple techniques. But first, an introduction to gradient-based planning: 🧵1/11
Jagath Vytheeswaran retweeted
JetBrains is no longer behind. @firebender_com just launched the first-ever background coding agents for all JetBrains IDEs.
These coding agents are incredibly intelligent, have isolated workspaces, and don’t require any cloud setup.
Jagath Vytheeswaran retweeted
🗺️ Can LLMs navigate San Francisco purely from memory, without Google Maps?
I tested spatial reasoning of the 19 top LLMs. Honestly...they were all pretty bad.
Best in cities: o3 (38%)
Best on highways: grok-4, o3, o4-mini (62-64%)
Seemed a bit fishy, so I asked to receive an email from the Google about this before continuing. Without missing a beat, the scammer sent a Google Forms submission -- sly of them to have an easy way to send a "Google-branded" email (3/6)
highly highly recommend checking them out -- one of the few ai-enabled-legacy-service cos that actually knows what they're doing and is changing customers' lives. join if you want to change an industry w a cracked team
Jagath Vytheeswaran retweeted
better auto-evals -> better voice agents -> better customer experience.
congrats to the Vogent team on the launch of their self-improving voice agent -- and awesome to see this self-improvement powered by Verdict judges ⚖️
Jagath Vytheeswaran retweeted
🔥 We built an AI agent that finds product alternatives in SECONDS right from your terminal! Type a product url, get options. That's it. ⚡️
Want to see how we built it? Read this 🧵
1/8
Jagath Vytheeswaran retweeted
Code’s here. Fork it. Hack it. 🛠️
github.com/hyperbrowserai/ex…
AI devs, what’s your move? 🤖🚀
Jagath Vytheeswaran retweeted
First came pre-training scaling; then came inference-time scaling.
Now comes judge-time scaling.
Despite progress in AI through scaled inference-time compute, AI remains unreliable in open-ended, non-verifiable domains. The key limitation is not generation—it is evaluation.
Therefore, the next big leap for AI comes from better judging.
In service of this future, today we release Verdict, a library for scaling judge-time compute.