Surendra Patil

Building highly resilient Decentralized AI Infrastructure for Web3.0.

[TFTC @TFTC21]

Someone leaked Claude Code's source.

44 hidden feature flags. Background agents, multi-agent orchestration, cron jobs, voice mode, browser control...

All already built, just not turned on yet.

Anthropic isn't shipping fast. They're uncorking what's already done.

[TFTC @TFTC21]

Jensen Huang: "If that $500,000 engineer did not consume at least $250,000 worth of tokens, I am going to be deeply alarmed. This is no different than a chip designer who says 'I'm just going to use paper and pencil. I don't think I'm going to need any CAD tools.'"

[Surendra Patil @surendra_patil]

Don't Just Watch the Future. Train It !!!
#TrainYourOwnAI

[HBCSE @HBCSE_TIFR]

Explore AI & IoT in our computer lab: Come Curious, Leave Understanding! Touch, Train & explore how machines think #NationalScienceDay @HBCSE_TIFR.

[Avi Chawla @_avichawla]

Fine-tune 100+ LLMs directly from a UI!

LLaMA-Factory lets you train and fine-tune open-source LLMs and VLMs without writing any code.

Supports 100+ models, multimodal fine-tuning, PPO, DPO, experiment tracking, and much more!

100% open-source with 50k stars!

[BURKOV @burkov]

100k tokens. This is where the excitement from LLM-produced code turns into frustration, no matter the LLM.

Grok just doesn't see anything beyond that threshold. Claude just provides guesses. Gemini rewrites everything every time, and it just gets uglier and doesn't work.

No serious application's codebase is under 100k tokens long.

I doubt that any substantial LLM improvement will allow it to be effective with a 1M token codebase. Until then, vibe coding is for prototypes and proofs of concept only.

Devin, you are our last hope! Don't disappoint us.

[Surendra Patil @surendra_patil]

UAE is set to use AI to draft and update laws, speeding up legislation by 70%! This AI-driven system will enhance accuracy and align laws with global standards, marking a new era in governance. #UAE #AI #LegalTech #Innovation

[Financial Times @FT]

The government expects the radical move to speed up lawmaking by 70%, but experts have warned the technology continues to 'hallucinate'. https://t.co/LEJzPXM51C

[Harvard Business Review @HarvardBiz]

Creating a personal experience map is one of the smartest ways you can invest your time. https://t.co/5otVFKm0mA

[Sam Altman @sama]

GPT-4.1 (and -mini and -nano) are now available in the API!

these models are great at coding, instruction following, and long context (1 million tokens).

benchmarks are strong, but we focused on real-world utility, and developers seem very happy.

GPT-4.1 family is API-only.

[Min Choi @minchoi]

🚨 It’s FINALLY here.

Google just dropped Gemini Live with real time screen & camera sharing.

You can now ask Gemini about anything you see.

Rolling out on Pixel 9, Galaxy S25 & Android first.

[Watcher.Guru @WatcherGuru]

JUST IN: Jim Cramer says not to panic.

[Surendra Patil @surendra_patil]

@monad_xyz 0x3a813Cf9ceEd3dc0b7Ea2b0E1caf55c27ba1Fa1B

[Surendra Patil @surendra_patil]

@nillion 🚀 Hello to a new era of data sovereignty!
🌐 Breaking free from centralized data silos!

[Nillion @nillion]

Nillion is now Open Source!

Following our our alpha mainnet launch, we’re making our core technologies available to everyone.

It’s time to build a more private internet together.

[Surendra Patil @surendra_patil]

NVIDIA's open-sourced cuOpt is a game-changer for blockchain + AI: 3000x faster optimization enables real-time transaction routing, DeFi rebalancing, and multi-agent coordination. Evaluates millions of possibilities on the fly. The future of intelligent Web3 just arrived. #cuOpt

[Surendra Patil @surendra_patil]

The line between sci-fi and reality continues to blur, with NVIDIA making GR00T N1 training data available is a game-changer. I'm fascinated by the dual-system architecture mimicking human cognition for reflexes and reasoning. Training robots got as easy as human demonstrations.

[TechCrunch @TechCrunch]

Nvidia CEO Jensen Huang debuts Groot N1, a general-purpose foundation model for humanoid robots, in Disney’s BDX Droids at GTC 2025

[Surendra Patil @surendra_patil]

MIT researchers created Exo 2, a language that lets developers write just ~100 lines of code to match performance of state-of-the-art HPC libraries built with 100,000+ lines. A few students with the right tools can now compete with NVIDIA's engineering army.

[MIT CSAIL @MIT_CSAIL]

A new programming language from MIT could enable high-performance coding that can compete w/state-of-the-art libraries w/a few hundred lines of code, instead of tens or hundreds of thousands.

Called "Exo 2," the approach enables reusable scheduling libraries external to compilers: https://t.co/PlUOXU6GGz

[Surendra Patil @surendra_patil]

The 4 questions that transformed our implementation strategy:

1. What if we wait?
2. What are we missing?
3. How will we measure success?
4. Can we reverse course?

It's like we obsess over consensus but often miss asking: "What options did we miss?" #BlockchainInnovation

[Surendra Patil @surendra_patil]

Imagine an AI that adapts and evolves with you! Manus AI is redefining collaboration in AI! Its multi-agent system enables seamless integration of specialized models, fostering a new era of AI-driven solutions. Journey to a more autonomous future is here! #AI #ManusAI #OpenSource

[Surendra Patil @surendra_patil]

🚀✨ Now is the supreme moment! 💡 Concentrate on excellence! Construct with the finest, pursue innovation relentlessly with passion, and become the brightest star in the space. Let's forge the future, one block at a time! 🛠️✨ #NextEraCreators #InnovateWithHeart #EyesOnTheFuture

[Surendra Patil @surendra_patil]

@SmokeyTheBera Unable to DM you. Count us in @ZenscapeOne