iiinova — where your digital self keeps evolving

Ultimately, we’re headed to a world where computer-use models “agentify” the long-tail of the web.

100

Most computer-use/browser-use benchmarks are GUI-only. Models (including Navigator n1.5) now support hybrid actions — UI interactions (click, type, scroll) and programmatic actions (e.g., execute JS).

100

In the last year, performance on Online Mind2Web has gone from ~40% success to basically saturated. So what’s next?

100

All benchmarks are transient attempts at measuring progress. Ultimately, what matters is how a model performs when users use it. But there’s a sentiment online that computer-use models aren’t progressing quickly. Not true.

100

𝗡𝗮𝘃𝗶𝗴𝗮𝘁𝗼𝗿 𝗻𝟭.𝟱 “𝘀𝗼𝗹𝘃𝗲𝗱” 𝗢𝗻𝗹𝗶𝗻𝗲 𝗠𝗶𝗻𝗱𝟮𝗪𝗲𝗯: 𝟵𝟳.𝟯% 𝘀𝘂𝗰𝗰𝗲𝘀𝘀 𝗿𝗮𝘁𝗲. While some teams self-report, this result is independently evaluated and verified by OSU NLP Group and Careerflow Human Data Labs.

Dhruv Batra @dhruvbatra.bsky.social · 21/05/2026

Including your CPA and the IRS?

SoTA performance at a fraction of the input price. Available now at yutori.com/api

Navigator n1 and n1.5 together set a new Pareto frontier in accuracy and cost.

n1.5 saw and read the page, then wrote a custom JS function to extract all results at once.

Asked to extract all available product sizes and colors: • n1 clicked 22 buttons • n1.5 finished in 5 steps

𝐄𝐱𝐩𝐚𝐧𝐝𝐞𝐝 𝐚𝐜𝐭𝐢𝐨𝐧𝐬 n1 could only output UI actions. n1.5 can also write JS to read and manipulate the webpage DOM. When filling out forms, n1.5 can one-shot fill multiple fields. Cuts down time in half. The model decides: code for efficiency, UI actions for generality, vision for verification.

𝐈𝐧𝐭𝐫𝐨𝐝𝐮𝐜𝐢𝐧𝐠 𝐍𝐚𝐯𝐢𝐠𝐚𝐭𝐨𝐫 𝐧𝟏.𝟓 The most capable computer-use model for the web. Pareto-domination: accuracy, latency, cost • SoTA across all benchmarks • +5-10% over GPT 5.5, Opus 4.7, n1 • +25% over Gemini • 2x faster, significantly cheaper

Dhruv Batra @dhruvbatra.bsky.social · 20/04/2026

yutori.com/api - 3 rounds of prompting (description, feedback+screenshot, feedback) - GPT 5.4 xhigh still lacking contrast. - Opus 4.7 max even with explicit description of content doesn't seem to get that the navigator agent has eyes. Output #1 was just a ball with no details.

000

Dhruv Batra @dhruvbatra.bsky.social · 20/04/2026

I gave Claude Code & Codex a video of @yutori_ai Navigator logo spinning and asked for code to regenerate it. Opus 4.7 max (left) vs GPT 5.4 xhigh (right) GPT 5.4 clearly better. Ground-truth / OG video in 🧵

210

Dhruv Batra @dhruvbatra.bsky.social · 10/03/2026

Two updates from Yutori: 1. We benchmarked GPT 5.4 on browser-use tasks • Matches/slightly-outperforms Opus 4.6 (+0.3%) • Big jump over previous OpenAI CUAs 2. Latest version of n1 • Outperforms GPT 5.4 and Opus 4.6 (+3%) • 2.5x faster, 4-5x cheaper.

100

Dhruv Batra @dhruvbatra.bsky.social · 11/02/2026

Most recent checkpoint of n1 vs Opus 4.6! On Navi-Bench and Westworld browser automation benchmarks: - Same accuracy - n1 is 2.5x faster - n1 is 5.6x cheaper Try it out via the Yutori API.

130

Dhruv Batra @dhruvbatra.bsky.social · 06/02/2026

Fun chat with Evan O'Donnell about the similarities between training robots and web agents, managing context for agents that run for months and years, the future of the AI-first web, and ideal form factor for embodied AI. www.thetimes.blog/p/agents-ne...

000

Dhruv Batra @dhruvbatra.bsky.social · 21/01/2026

Maybe coding is just amortized inference for LLMs. Maybe the reason we write programs down to files is just to save inference costs.

Dhruv Batra @dhruvbatra.bsky.social · 14/11/2025

Why? Because the web is a mess. And websites are fundamentally built for human consumption. If you want to automate everything that a human can do with a browser, then you have to perceive like a human. Stay tuned for more in the coming weeks! yutori.com/blog/the-bit...

yutori.com

The bitter lesson for web agents

Agents that generalize need to perceive the web like a human would.

Dhruv Batra @dhruvbatra.bsky.social · 14/11/2025

The bitter lesson for web agents The last 1 year has taught us a new bitter lesson that we think others are not yet grokking. Agents that *look at the web like humans* (screenshots of sites) navigate and generalize better than agents that read code (HTML, DOM).

140

Dhruv Batra @dhruvbatra.bsky.social · 23/10/2025

Solved: robustness to paraphrasing and false premises, OCR, world-knowledge based reasoning. Open: spatial reasoning, data-efficiency, learning compatible representations.

000

Dhruv Batra @dhruvbatra.bsky.social · 23/10/2025

As part of the award ceremony, VQA team presented a recap of vision-and-language research over the last decade — solved problems, progress, and open-challenges for mutimodal LLMs.

100

Dhruv Batra @dhruvbatra.bsky.social · 21/10/2025

Lots to be done. Thank you to all our collaborators and the research community for this recognition!

000

Dhruv Batra @dhruvbatra.bsky.social · 21/10/2025

Fun-fact: the T-shirt I'm wearing is an inside joke about the quality of 2015 models. However, every few years we rediscover the lesson that on difficult tasks, VLMs silently regress to being nearly blind. x.com/DhruvBatra_/...

Dhruv Batra @dhruvbatra.bsky.social · 21/10/2025

VQA challenge series won the Mark Everingham prize at #ICCV2025 for stimulating a new strand of vision-and-language research. It's extra special because ICCV25 marks the 10-year anniversary of the VQA paper. When we started, the idea of answering any question about any image seemed outlandish.

1122

Dhruv Batra @dhruvbatra.bsky.social · 15/10/2025

The problem with “AI slop” isn’t the AI — it’s the slop. People act like AI is the issue, when it’s actually part of the fix. If we're honest: most of what we make, most of the time, is slop by our own standards. That’s the generator–discriminator gap in creative work that Ira Glass talks about.

Dhruv Batra @dhruvbatra.bsky.social · 10/06/2025

Somebody is a fan of Abundance

Dhruv Batra @dhruvbatra.bsky.social · 16/04/2025

It is so refreshing to see conferences innovate on the reviewing model and run actual experiments (!) as opposed to fighting change.

020

Dhruv Batra @dhruvbatra.bsky.social · 01/04/2025

Good. Autonomous interface locomotion is the fundamental robotics problem of our time. The more the merrier.

000

Dhruv Batra @dhruvbatra.bsky.social · 01/04/2025

My entire robotics career has led to this.

151

Dhruv Batra @dhruvbatra.bsky.social · 28/03/2025

The answer to many "why X?" questions: Because the laws of physics do not prohibit X and the forces of biology gave us curiosity.

Dhruv Batra @dhruvbatra.bsky.social · 27/03/2025

The web is the ultimate boss-level for agents — dynamic, non-deterministic, and noisy; some mistakes are inevitable and so far, every agent fails eventually. Yutori is building superhuman agents for this ultimate digital environment. Join our waitlist for early access to our product! yutori.com

yutori.com

Yutori

We’re building AI agents that can reliably do everyday digital tasks for you on the web, towards an AI chief-of-staff for everyone.

221

Dhruv Batra @dhruvbatra.bsky.social · 27/03/2025

𝐈𝐦𝐚𝐠𝐢𝐧𝐞 𝐚 𝐰𝐨𝐫𝐥𝐝 𝐰𝐡𝐞𝐫𝐞 𝐧𝐨 𝐡𝐮𝐦𝐚𝐧 𝐡𝐚𝐬 𝐭𝐨 𝐝𝐢𝐫𝐞𝐜𝐭𝐥𝐲 𝐢𝐧𝐭𝐞𝐫𝐚𝐜𝐭 𝐰𝐢𝐭𝐡 𝐭𝐡𝐞 𝐰𝐞𝐛 𝐚𝐠𝐚𝐢𝐧. Where teams of AI assistants coordinate to book flights, manage budgets, or file paperwork—proactively surfacing insights and correcting errors. Only problem — no one knows how to build AI agents that actually work.

100

Dhruv Batra @dhruvbatra.bsky.social · 27/03/2025

I started something new last year with a wonderful group of people. We showed a demo in Jan. Today, we’re telling our story — show before you talk! 𝘞𝘦 𝘢𝘳𝘦 𝘳𝘦-𝘪𝘮𝘢𝘨𝘪𝘯𝘪𝘯𝘨 𝘩𝘰𝘸 𝘱𝘦𝘰𝘱𝘭𝘦 𝘪𝘯𝘵𝘦𝘳𝘢𝘤𝘵 𝘸𝘪𝘵𝘩 𝘵𝘩𝘦 𝘸𝘦𝘣 — one of humanity’s greatest inventions and a a mess overdue for an overhaul. yutori.com

1101

Ah, understood. No idea about the tracing of that meme.

000

Seems like the ultimate thing to rally around, no? To the extent there is any purpose, what's the alternative?

I'm already there for low-stakes queries.

Where's the skepticism coming from? Now that web search and citations are in there, isn't it easy to verify and thus become more confident?

MAPS - CVPR 2026 Workshop @mapscvpr.bsky.social · 14/03/2025

Reposted by Dhruv Batra

📢Excited to announce our upcoming workshop - Vision Language Models For All: Building Geo-Diverse and Culturally Aware Vision-Language Models (VLMs-4-All) @CVPR 2025! 🌐 sites.google.com/view/vlms4all

11711

Dhruv Batra @dhruvbatra.bsky.social · 06/03/2025

Using a locally-running LLM to translate a review is explicitly prohibited by @iccv.bsky.social Why? Whom does this possibly harm?

000

Dhruv Batra @dhruvbatra.bsky.social · 26/02/2025

The way it's always been done isn't handling the current scale well (as evidenced by the feedback from the authors). Yes, outsource to a company, pay for creation of new tools, start new companies, all of the standard ways of addressing a growing market.

300

Dhruv Batra @dhruvbatra.bsky.social · 26/02/2025

Why is it volunteer work? Why doesn't an organization that takes in millions in sponsorship professionalize?