Priyanka Kamath

Happy 2023 ✨ The Best Is Yet To Come!!!! #UnrealTimes💠

ELON MUSK: “To reach a petawatt, we’d build an electromagnetic mass driver on the Moon with robots like Optimus and humans. With no atmosphere and 1/6 gravity, payloads can reach escape velocity without rockets, drastically cutting costs. That could scale compute 1000× beyond a terawatt—I hope we live to see it.”

gm gm gm @unstoppableweb fam!!!

A few updates from across Google this week🧵 A new full-stack vibe coding experience is now in @GoogleAIStudio with the @Antigravity coding agent and a built-in @Firebase integration. Turn your prompts into amazing production-ready apps.

Do take a look at @CapitalmindMF Flexi Cap Fund for a quantitative allocation to your portfolio. We are at capitalmindmf.com and also at your favourite MF platform:

It’s 2am again, my favourite time, and as always reality is still staring me in the face 🧐

We’re launching a brand new, full-stack vibe coding experience in @GoogleAIStudio, made possible by integrations with the @Antigravity coding agent and @Firebase backends. This unlocks: — Full-stack multiplayer experiences: Create complex, multiplayer apps with fully-featured UIs and backends directly within AI Studio — Connection to real-world services: Build applications that connect to live data sources, databases, or payment processors and the Antigravity agent will securely store your API credentials for you — A smarter agent that works even when you don't: By maintaining a deeper understanding of your project structure and chat history, the agent can execute multi-step code edits from simpler prompts. It also remembers where you left off and completes your tasks while you’re away, so you can seamlessly resume your builds from anywhere — Configuration of database connections and authentication flows: Add Firebase integration to provision Cloud Firestore for databases and Firebase authentication for secure sign-in This demo displays what can be built in the new vibe coding experience in AI Studio. Geoseeker is a full-stack application that manages real-time multiplayer states, compass-based logic, and an external API integration with @GoogleMaps 🕹️

Claude can now build interactive charts and diagrams, directly in the chat. Available today in beta on all plans, including free. Try it out: claude.ai

Can solo-founder build a $1Billion company now?

How disruptive are you?

Do you remember when you joined X? I do! #MyXAnniversary Best Social Media Decision 🚀

Super Stoked For This @abhilashi ! Can't wait to try!!

🚀 Stop building chatbots. Start building LIVE AI agents. The Gemini Live Agent Challenge is here — and it’s redefining how we interact with AI. 🗣️ Talk naturally 👀 Vision-enabled understanding 🎥 Multimodal storytelling 🖥️ UI navigation agents ☁️ Powered by Gemini + Google Cloud 💰 $80,000 in prizes 🏆 Grand Prize: $25,000 + Google Cloud Next 2026 tickets ⏳ Deadline: March 17, 2026 Build beyond text. Build immersive AI. Join now geminiliveagentchallenge.devpost.com #GeminiLiveAgentChallenge #AI #MachineLearning #GoogleCloud #BuildWithAI #Hackathon

@kamath_sutra Repping for white paper 🗞️

Who wants a white paper on Hydra? Share your email here, will send it over this week.

Sarvam Kaze waitlist is live. Join now: sarvam.ai/kaze-waitlist

Skills for AI agents to analyze news and forecast markets github.com/RKiding/Awesom…

I am a proud investor in this ambitious startup. The returns from this investment go far beyond financial rewards. It’s about getting a ringside seat to watch the literal ‘take-off’ of Indian talent…

Tonight, we reached an agreement with the Department of War to deploy our models in their classified network. In all of our interactions, the DoW displayed a deep respect for safety and a desire to partner to achieve the best possible outcome. AI safety and wide distribution of benefits are the core of our mission. Two of our most important safety principles are prohibitions on domestic mass surveillance and human responsibility for the use of force, including for autonomous weapon systems. The DoW agrees with these principles, reflects them in law and policy, and we put them into our agreement. We also will build technical safeguards to ensure our models behave as they should, which the DoW also wanted. We will deploy FDEs to help with our models and to ensure their safety, we will deploy on cloud networks only. We are asking the DoW to offer these same terms to all AI companies, which in our opinion we think everyone should be willing to accept. We have expressed our strong desire to see things de-escalate away from legal and governmental actions and towards reasonable agreements. We remain committed to serve all of humanity as best we can. The world is a complicated, messy, and sometimes dangerous place.

Your AI agents are only as good as the systems they can interact with. Introducing the new ADK integrations ecosystem. Connect your ADK agents to GitHub, Hugging Face, Notion and many more with tools and plugins. Start building today: goo.gle/4aTAJAV

I had the same thought so I've been playing with it in nanochat. E.g. here's 8 agents (4 claude, 4 codex), with 1 GPU each running nanochat experiments (trying to delete logit softcap without regression). The TLDR is that it doesn't work and it's a mess... but it's still very pretty to look at :) I tried a few setups: 8 independent solo researchers, 1 chief scientist giving work to 8 junior researchers, etc. Each research program is a git branch, each scientist forks it into a feature branch, git worktrees for isolation, simple files for comms, skip Docker/VMs for simplicity atm (I find that instructions are enough to prevent interference). Research org runs in tmux window grids of interactive sessions (like Teams) so that it's pretty to look at, see their individual work, and "take over" if needed, i.e. no -p. But ok the reason it doesn't work so far is that the agents' ideas are just pretty bad out of the box, even at highest intelligence. They don't think carefully though experiment design, they run a bit non-sensical variations, they don't create strong baselines and ablate things properly, they don't carefully control for runtime or flops. (just as an example, an agent yesterday "discovered" that increasing the hidden size of the network improves the validation loss, which is a totally spurious result given that a bigger network will have a lower validation loss in the infinite data regime, but then it also trains for a lot longer, it's not clear why I had to come in to point that out). They are very good at implementing any given well-scoped and described idea but they don't creatively generate them. But the goal is that you are now programming an organization (e.g. a "research org") and its individual agents, so the "source code" is the collection of prompts, skills, tools, etc. and processes that make it up. E.g. a daily standup in the morning is now part of the "org code". And optimizing nanochat pretraining is just one of the many tasks (almost like an eval). Then - given an arbitrary task, how quickly does your research org generate progress on it?

Big tip: as Claude.md gets longer than 200 lines, put in *triggers* so Claude knows if you touch evals with this *_generator.* filename splat, load docs/EVALS.md. You can test the trigger prompt in another Claude window. Works well, saves tokens!