Pranav Shyam
@recurseparadox
low co2 ppm enjoyer. research rcientist at google deepmind. past: generative models at openai; rl w/ juergen schmidhuber. ಕನ್ನಡಿಗ
Our MAX algorithm for pure, active exploration now scales to high-dimensional continuous environments to build task-agnostic models. Check it out: arxiv.org/abs/1810.12162

Feed the Children of Gaza, no child should be starving.
it takes great epistemic bravery to be skillful in pre-paradigmatic sciences. that shit is messy, and one must avoid disguising their cowardice as "love of plots that make straight lines"
Lately I’m convinced that impact of video models will be greater than LLMs. The ordinary person is not using Cursor or doing homework. Even models from the last 3 months have caused a huge uptick in crazy things I’ve seen and almost believed
This is not AI this is real TelAviv Airport you can confirm with Grok
Waymo in a new blog post: "We conducted a comprehensive study using Waymo’s internal dataset. Spanning 500,000 hours of driving, it is significantly larger than any dataset used in previous scaling studies in the AV domain. Our study uncovered the following: • Similar to LLMs,…
I suck too much to be truly good. But I'm pretty good for someone who sucks.
Found his handle @CharlieXYChen
xprof artwork, created by Charlie Chen
Updated 2.5 Pro! Notable improvements in Humanity's Last Exam and Aider Polyglot (code editing)
Our latest Gemini 2.5 Pro update is now in preview. It’s better at coding, reasoning, science + math, shows improved performance across key benchmarks (AIDER Polyglot, GPQA, HLE to name a few), and leads @lmarena_ai with a 24pt Elo score jump since the previous version. We also…
woo hoo!
🚨 BREAKING: GPT Image Killer is here. FLUX Kontext [pro] & [max] models are available on @fal day 0! This isn't just your usual image editing model – it preserves identity AND maintains consistency with the rest of the image. Try it today here -> fal.ai/flux-kontext