sritee

Interested in ML and Robotics. Research Engineer @GoogleDeepmind

[nisarga @ni5arga]

everyone wants a hacking roadmap.

the problem is that roadmaps don't create hackers.
curiosity does.

ctfs are changing, ai is everywhere, and the game looks very different now.

how i'd start hacking in 2026: https://t.co/VZS76KTHWJ

[sritee @Sridhaar96]

Check out our new work exploring test-time compute vs train-time compute for making inferences which are latent in the training data!

[Arslan Chaudhry @arslan_mac]

[1/8] Our new paper on using test-time compute to improve latent learning: https://t.co/bYt6K2rSI3

Thinking models have been shown to improve performance on maths, reasoning and coding tasks. In this work, however, we look into how thinking models improve latent learning.

[Arslan Chaudhry @arslan_mac]

[1/8] Our new paper on using test-time compute to improve latent learning: https://t.co/bYt6K2rSI3

Thinking models have been shown to improve performance on maths, reasoning and coding tasks. In this work, however, we look into how thinking models improve latent learning.

[Lex Fridman @lexfridman]

Huge thanks to all the open source projects that've made a lot of the tech we rely on in the world possible:

Linux
Git
FFmpeg
PyTorch & TensorFlow
Apache & Nginx
MySQL, PostgreSQL, SQLite
Chromium & Firefox
GCC & LLVM
Docker & Kubernetes

Also, all the open-weight LLMs... and all the programming language interpreters, compilers, and frameworks.

I can list 100+ more "niche" open source projects I've used & loved (e.g., OpenCV ❤️). But I just wanted to give a continued big thank you to the open source community for everything you do for humanity 🙏

[Google @Google]

🧵 New advanced AI capabilities are coming to Search ⚡ Now you can:

✅ Try Gemini 2.5 Pro in AI Mode for help with complex questions
✅ Save hours of research with Deep Search in AI Mode
✅ Have Search call local businesses to check pricing for you

[Google DeepMind @GoogleDeepMind]

When his son Max was diagnosed with a rare disease, our colleague Thomas turned to Gemini to better understand complex scientific papers — and even make new connections across the research community.

His story is a powerful reminder of why we work on AI. ↓

[Owain Evans @OwainEvans_UK]

Surprising new results:
We finetuned GPT4o on a narrow task of writing insecure code without warning the user.
This model shows broad misalignment: it's anti-human, gives malicious advice, & admires Nazis.

This is *emergent misalignment* & we cannot fully explain it 🧵

[Jacob Austin @jacobaustin132]

Making LLMs run efficiently can feel scary, but scaling isn’t magic, it’s math! We wanted to demystify the “systems view” of LLMs and wrote a little textbook called “How To Scale Your Model” which we’re releasing today. 1/n

[David Blanco-Mulero @DavidBMulero]

Interested in deformable objects, cloth manipulation or simulation engines?
We have evaluated the Sim-to-Real gap in @GoogleDeepMind MuJoCo, Bullet, @nvidia Flex and @SofaFramework SOFA on dynamic and quasi-static cloth manipulation!

Do you want to know more? Scroll down! 🧵

[Andrej Karpathy @karpathy]

⚡️ Excited to share that I am starting an AI+Education company called Eureka Labs.
The announcement:

---
We are Eureka Labs and we are building a new kind of school that is AI native.

How can we approach an ideal experience for learning something new? For example, in the case of physics one could imagine working through very high quality course materials together with Feynman, who is there to guide you every step of the way. Unfortunately, subject matter experts who are deeply passionate, great at teaching, infinitely patient and fluent in all of the world's languages are also very scarce and cannot personally tutor all 8 billion of us on demand.

However, with recent progress in generative AI, this learning experience feels tractable. The teacher still designs the course materials, but they are supported, leveraged and scaled with an AI Teaching Assistant who is optimized to help guide the students through them. This Teacher + AI symbiosis could run an entire curriculum of courses on a common platform. If we are successful, it will be easy for anyone to learn anything, expanding education in both reach (a large number of people learning something) and extent (any one person learning a large amount of subjects, beyond what may be possible today unassisted).

Our first product will be the world's obviously best AI course, LLM101n. This is an undergraduate-level class that guides the student through training their own AI, very similar to a smaller version of the AI Teaching Assistant itself. The course materials will be available online, but we also plan to run both digital and physical cohorts of people going through it together.

Today, we are heads down building LLM101n, but we look forward to a future where AI is a key technology for increasing human potential. What would you like to learn?
---

@EurekaLabsAI is the culmination of my passion in both AI and education over ~2 decades. My interest in education took me from YouTube tutorials on Rubik's cubes to starting CS231n at Stanford, to my more recent Zero-to-Hero AI series. While my work in AI took me from academic research at Stanford to real-world products at Tesla and AGI research at OpenAI. All of my work combining the two so far has only been part-time, as side quests to my "real job", so I am quite excited to dive in and build something great, professionally and full-time.

It's still early days but I wanted to announce the company so that I can build publicly instead of keeping a secret that isn't. Outbound links with a bit more info in the reply!

[Google @Google]

Introducing LearnLM: our new family of models based on Gemini and fine-tuned for learning. LearnLM applies educational research to make our products — like Search, Gemini and YouTube — more personal, active and engaging for learners. #GoogleIO

[Michael Chang @mmmbchang]

Gemini and I also got a chance to watch the @OpenAI live announcement of gpt4o, using Project Astra! Congrats to the OpenAI team, super impressive work!

[Chrome @googlechrome]

Quickly start your chat with Gemini using the new shortcut in the Chrome desktop address bar👇

Step 1: Type “@” in the desktop address bar and select Chat with Gemini
Step 2: Write your prompt
Step 3: Get your response on https://t.co/MukYC54K9e

Seriously. It’s that easy ✨

[Delip Rao e/σ @deliprao]

If this were a science paper, you would expect a country that picks its science workforce at random as a “weak baseline” and a leading nation like the US to actively experiment towards state-of-the-art, or at least beat the baseline.

Not providing a guaranteed path for accomplished scientists like @rdesh26, who graduated from one of the top labs in the country, @jhuclsp, to contribute to our progress and security directly is an avoidable tragedy.

I implore the @WhiteHouse / @USAGov OSTP, @StateDept, and @STASatState to look into the ways we are leaching out the best talent in the country, starting with this lottery system for the nation’s most talented.

Many have made this request before to past administrations to no avail, but acting on this now would leave a legacy that will be the envy of future administrations. I hope you do. If you have doubts, ask other best scientists nationwide, including @theNASciences.

[Florian Stimberg @FStimberg]

How easy is it for adversaries to hide image content from classifiers through obfuscations? Our new benchmark allows you to evaluate this!
Dataset and evaluation code: https://t.co/kWKcqwFTMy
Paper: https://t.co/U1zXMy1Zng.
Joint work between @DeepMind, @Google & @GoogleAI 🧵

[Dongyoung Go @dongyoung4091]

RLHF – the method in ChatGPT – can be seen as minimizing the reverse KL from a certain target distribution. We show how this can be generalized to different divergences, resulting in various impacts on diversity and alignment. A thread about our paper at https://t.co/vk6DqHbrpy

[Ganesh Iyer @__epiception__]

Excited for this release! :)

As our foundation models get better, our 3D maps too can now scale to support multimodal concepts at varying levels of detail. All while being built on-the-fly!

ConceptFusion enables a suit of downstream tasks in indoor and outdoor environments. 🚀

[Eugene Vinitsky 🦋 @EugeneVinitsky]

Well, I don't know what else chat LLMs are really good for but I'm never writing another regex again

[kache @yacineMTB]

@kentdomuch the possibilities are endless

Right now I'm working on having a permanent whisper active on my computer, so I can livegrep everything that's been said, as to aid remembering something

Also working on text to speech for something called "auto-podcast"