Yash Patil

Building in AI and open source. I track the fastest-rising repos daily at 📡 Sharing what's new in AI, what's worth trying, and what I'm building.

[Yash Patil @Yashpatil]

@ns123abc The scary part is that no AI hacking was needed. A convincing impersonation got past people who said they'd met him before. Anyone giving agents access to their email should think about this too, because agents are even easier to fool.

[Yash Patil @Yashpatil]

@jerryjliu0 @llama_index Makes sense. Every agent I build ends up stuck on the same step: getting clean data out of PDFs, tables and scans. Most of the time the model isn't the bottleneck, the documents are.

[Yash Patil @Yashpatil]

@chetaslua Banked resets are a smart move. What heavy users hate most is hitting the limit in the middle of a task, and saving a reset for a big day fixes exactly that. Curious if OpenAI answers with something similar.

[Yash Patil @Yashpatil]

@davidvgilmore The catch: those are Rayline's own numbers, and self-hosting isn't smooth yet. Stock vLLM can't load it and their public plugin is unfinished, so most people will use the hosted API for now. The weights are Apache 2.0 on Hugging Face.

https://t.co/YnUCrH4NW3

[Yash Patil @Yashpatil]

Most AI agents use their priciest model for every step. Even the easy ones.

Rayline open-sourced ARC, a tiny 0.8B model that picks the model for each step. Their test: 68.7% on DeepSWE vs 69.9% for Opus 5.5, at about 7% of the cost.

Video: @davidvgilmore

[Yash Patil @Yashpatil]

@ns123abc Closing the rental route matters more than it sounds. For a lot of labs, renting GPUs abroad was the plan B after the chip bans. I'd expect even more push on efficient training and home-grown chips next.

[Yash Patil @Yashpatil]

@RoundtableSpace 22 hours is the interesting number. That's not a one prompt demo, it's a full workday of the agent testing and fixing its own work. The real skill now is giving it a goal clear enough to stay on track that long.

[Yash Patil @Yashpatil]

@ai_for_success Building a tool you use every day is the real test. Demos can break, daily tools have to keep working. Curious how it handled the boring parts, like fixing the pipeline when one of your platforms changes its API.

[Yash Patil @Yashpatil]

@n8n_io What I like: the safety rules live in the server, not the prompt, so an email or web page can't give it orders. Passwords go in through a vault the model never sees. Apache 2.0.

It jumped from 436 to 536 stars in a day on https://t.co/C3iBwBhrL1

https://t.co/PTPRJVwPso

[Yash Patil @Yashpatil]

An AI agent that cancels your free trials for you. And it's open source.

Nod from n8n Labs gets its own computer and browser, so it can chase invoices, prep your calls and cancel subscriptions on real websites. It asks before it sends, pays or deletes anything.

Video: @n8n_io

[Yash Patil @Yashpatil]

@ArtificialAnlys @joinHandshakeAI Tools nearly doubling scores lines up with what people are building. 4 of the 50 fastest-rising repos on https://t.co/C3iBwBhrL1 today are MCP servers and 12 mention skills. The bet is on better tools, not just bigger models.

[Yash Patil @Yashpatil]

@mattshumer_ Players filing bug reports and agents shipping fixes within hours is basically a live game studio. Do you look over each version before it goes out, or do the agents push straight to players?

[Yash Patil @Yashpatil]

@AravSrinivas @dylan522p The name works because it's true. Once an agent can run code, browse and keep files around between tasks, it stops feeling like a chat app and starts feeling like a computer you rent.

[Yash Patil @Yashpatil]

@GoogleDeepMind The numbers: on a big map of 28 immune cell types, it found nearly twice as many causal variants as the original AlphaGenome (30% vs 16%). Scientists can add their own cell data without retraining. Free for non-commercial research via the API.

https://t.co/gEow0pLJro

[Yash Patil @Yashpatil]

Every cell in your body has the same DNA. So why does one tiny change cause disease in some cells and not others?

Google DeepMind's new AlphaGenome 2 looks at one cell type at a time. It was trained on 100M+ single cell profiles across 68 tissues.

Video: @GoogleDeepMind

[Yash Patil @Yashpatil]

@kimmonismus This gap gets longer with every launch. For people in Europe the best personal agents now show up months late, if at all, which is rough for builders who want to try things on day one.

[Yash Patil @Yashpatil]

@emollick You can see it on GitHub too. Of the 50 fastest-rising repos on https://t.co/C3iBwBhrL1 today, 10 are coding agent tools and 16 have Claude in the name or description. Developers didn't wait for a company rollout, they just installed things.

[Yash Patil @Yashpatil]

@testingcatalog Handy, but this is where I'd want real controls: which sites each login can be used on, and a log of every time Claude signed in. Letting an agent hold your passwords only works if you can see what it did with them.

[Yash Patil @Yashpatil]

@sundarpichai Why builders should care: it costs $0.10 per 1M input tokens and output is free. Google says it's nearly 5x cheaper than Gemini 3.5 Flash-Lite on its test suite, and Google already uses it for model routing and safety checks. Free tier in AI Studio.

https://t.co/fCLE1Tm30y

[Yash Patil @Yashpatil]

Google just made an AI model that doesn't write answers. It picks them.

Gemini Decision 1 scores up to 1,000 yes/no, multiple choice or rubric questions about one input (text, images, audio, video or PDFs) in a single pass, in about 150 to 250 ms.

Video: @sundarpichai