1. X
  2. Dhruv Batra
Log inSign up
Dhruv Batra
3,303 posts
Image
user avatar
Dhruv Batra
@DhruvBatra_
Co-founder & Chief Scientist @yutori_ai. Prev: Senior Director leading FAIR Embodied AI @MetaAI and Professor @GeorgiaTech.
dhruvbatra.com
Joined January 2016
754
Following
21.3K
Followers
RepliesRepliesArticlesArticlesMediaMedia
  • Pinned
    user avatar
    Dhruv Batra
    @DhruvBatra_
    Nov 19, 2025
    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
    Image
  • user avatar
    Dhruv Batra
    @DhruvBatra_
    Jul 24
    The most valuable bit of information in an eval is often — which method is #2? (Because the #1 spot faced a strong selection pressure) For instance, what this plot conveys is a confirmation from Anthropic that GPT 5.6 Sol pareto-dominates Fable 5 on coding benchmarks.
    Image
    Image
    Image
    Image
    Image
    user avatar
    Claude
    Anthropic
    @claudeai
    Jul 24
    Replying to @claudeai
    Opus 5 is also highly efficient. It outperforms other models for a similar or lower cost per task:
  • user avatar
    Dhruv Batra
    @DhruvBatra_
    Jul 23
    We eval'd Muse Spark 1.1 on Online-Mind2Web — a computer-use / browser-use benchmark • Better than Opus 4.8 • Slightly worse than GPT 5.4 (but possibly not a statistically significant difference) Qualitatively, Muse tends to manipulate URLs more than other models: • To
    Image
  • user avatar
    Dhruv Batra
    @DhruvBatra_
    Jul 22
    Another conjecture falls — 4 prompts with a public model. The pattern (for now) seems to be search for counter-examples to disprove conjectures, rather than constructive proofs. Maybe because the former lend themselves to grinding and persistence more?
    user avatar
    Dmitry Rybin
    @DmitryRybin1
    Jul 22
    Dinitz-Garg-Goemans conjecture is false. This graph theory problem was open for ~30 years. The graph below has fractional flow cost 58. Any unsplittable flow (with capacity violation <=15) has cost at least 60. Chat with GPT 5.6 Pro where this was found: chatgpt.com/share/6a60b2eb…
    Image
  • user avatar
    Dhruv Batra
    @DhruvBatra_
    Jul 20
    fun chat about the similarities between physical agents (robots) and digital agents (computer use / browser use). thanks for having me!
    user avatar
    Ravid Shwartz Ziv
    @ziv_ravid
    Jul 20
    New episode of the Information Bottleneck with Dhruv Batra (@DhruvBatra_ ), co-founder of Yutori and former head of Embodied AI at Meta FAIR. He spent years training robots to navigate 3D scans, and now he builds web agents, and he says it's the same job, and the browser is just

Log in or sign up for X

See what’s happening and join the conversation

Continue with phone
or
Log in with username or email
Terms·Privacy·Cookies·Accessibility·Ads Info·© 2026 X Corp.
Advertisement
Advertisement