Rayan Garg

@ritvikpandey21 @Pulse__AI Huge!

the team at @Pulse__AI put bytedance's dolphin OCR to the test against complex documents that matter for real business use cases. while it shows improvements in reading order detection, we found critical limitations across key areas: - 7.7% structured data extraction from financial charts - 60% accuracy drop without grid lines, 31% hierarchy preservation rate, and boundary confusion across multi-table documents. credit to the bytedance team for the bounding box improvements, which genuinely outperform many open-source alternatives. specialized document processing systems still remain essential where accuracy and structure preservation are non-negotiable. we're excited to see continued innovation in open source document AI - this is an important problem to solve.

Browser agents use computers the same way humans do, unlocking powerful use cases for personal assistants, browsers, and enterprise workflows. After talking to 20+ founders in the space, we're excited to put out the definitive market map for browser agents.

@AravSrinivas 🙋‍♀️Interested! Email: rg569@cornell.edu

expanding comet access to more folks as we near the finish line. reply here if interested. needs to be MacBook. Preferably Apple silicon but Intel is fine too. DM your email that you use for perplexity.

@ycombinator Huge!

Vesence is Cursor for Transactional Lawyers. They're building Agentic AI inside of Microsoft Office, turning Word into an IDE for contracts. Vesence is already live firm-wide with their first enterprise client, loved by both associates and partners. ycombinator.com/launches/NYo-v… Congrats on the launch, @HenrikTaro and @LudvigSwanstrom!

@GrantStenger Love the release!

We've cracked $100m in lifetime volume! Kinetic is officially live to the public!

Introducing CUB: Humanity's Last Exam for Computer and Browser Use Agents

@ycombinator @humanbehaviorai @OfficialAmogh @skyguy_skyguy Amazing!

Human Behavior (@humanbehaviorai) is building an AI that watches actual customers using your product, through session replays, to help you understand why they stay, convert, or leave. ycombinator.com/launches/NU8-h… Congrats on the launch, @OfficialAmogh, @skyguy_skyguy, and @chirag_plusplus!

Theta (@trytheta) allows AI agents to learn from their mistakes in real-time. Their memory layer has already improved the accuracy of OpenAI Operator by 43% with 7x fewer steps taken. ycombinator.com/launches/NTK-t… Congrats on the launch, @RayanGarg, @tsha444, and @_gurvir_!

@omarwasm I would love to be added!

I’m making a gc of the most cracked undiscovered builders + growth hackers I’ve met on here comment for an invite and I’ll DM you

@josephsemrai Incredible!

Meet Context Autopilot It learns like you, thinks like you, and uses tools like you. With SoTA context understanding, it's capable of most information work today. Watch it beat a team of industry experts:

@nehadesaraju @JackBlair87

peak sf is a guy asking me if it was worth it talking to his grandparents because he had a more productive conversation with a vc partner today than he's ever had with his grandparents