vani

i evaluate deep research apis so you don't have to

[vani @vaniagrwall]

@hellohsuh @jiholimm yay!

[vani @vaniagrwall]

i love mobbin

[vani @vaniagrwall]

love mobbin!

one thing that trips me up is when i click “next” in the screen viewer, i expect to see the next step in that website’s flow, but it takes me to a screen from a different website.

to follow the original flow, i have to click the website’s logo and scroll through its screens to find the ones i’m looking for. would love for a way to potentially continue through the same flow directly from the screen i’m on 👀

[Andrey Styskin @styskin]

Keenable 1.0 is out.

Agents want the original document and pay for every token. 1.0 returns only the part they need, cutting tokens 50% on research tasks, and adds 100M knowledge cards.

It’s the new champion in web search for agents, and I’m proud of what we’ve built.

https://t.co/FWo9O59evf

[vani @vaniagrwall]

wip onboarding :) does this feel simple & welcoming 🌻

[vani @vaniagrwall]

@hellohsuh @whereischarly @browserbase this is so cool

[vani @vaniagrwall]

RT @hellohsuh: I wish I recorded my reaction when I saw this live. Truly blessed to have had the opportunity to work with @whereischarly t…

[Andrey Styskin @styskin]

@ArtificialAnlys @KeenableAI Thank you @ArtificialAnlys for running this. The @KeenableAI team has been cooking.

[vani @vaniagrwall]

@wwwjim we have lots of interesting and personalized camps & activities very immediately available close to you 🏀

[vani @vaniagrwall]

working on an app for finding + booking summer camps & children's activities 🌻

does this layout feel intuitive?

[vani @vaniagrwall]

@hellohsuh hannabanana ui library

[vani @vaniagrwall]

What does a research harness add when the model stays the same?

@AnthropicAI's Fable 5.2 answered 12 ResearchRubrics questions across four setups: closed-book, a search-and-read loop using @ExaAILabs, GPT Researcher, and LangChain Deep Agents.

Closed-book was fastest and cheapest. Deep Agents scored slightly higher, but at nearly 18× the cost and 6× the runtime. The Exa baseline also scored higher than GPT Researcher at less than half the cost.

Across these 12 questions, adding research tools or a harness produced no clear rubric-score improvement.

https://t.co/fcZwSU9jlS

[vani @vaniagrwall]

@SolRuck THANKS

[Jimmy Lee @wwwjim]

I am so thrilled about our game engines ability to create scenes like this, every website will be a world, soon.

[vani @vaniagrwall]

@maayanalbert @soleio @proto_ide @hellohsuh

[Sol Ruck @SolRuck]

Spot the difference 👀

[Jimmy Lee @wwwjim]

@flipgoing we also offer our support at https://t.co/ZEG7qwCCMG, @vaniagrwall and @SolRuck just joined

[Jimmy Lee @wwwjim]

We brought @SolRuck and @vaniagrwall onto our team at the @internetxstudio, so you should follow them 👊

[vani @vaniagrwall]

@cynthwangg @hellohsuh