Tanmay R Thombare @Tan_builder
Engineer. Builder. Explorer Building in public tan-thombare.vercel.app India Joined July 2026-
Tweets83
-
Followers1
-
Following151
-
Likes53
Building a custom inventory management system for another business. (Fully local) Tauri + sqlite. I will be sharing how this progress✌️
@AnshikaK7 I know what a python GIL is. But most important... I know what is JSON😂😂
An LLM can't actually browse the web or execute code by itself. It generates a response requesting a tool call (just generates text). Your code reads that response, runs the tool, and sends the result back to the LLM. The model provides the decision. Your code does the execution
Pure speed hack!!! Just learned how KV caching speeds up LLM inference and it’s genius. Instead of recomputing attention for every past token at each step, it stores the Keys & Values in memory. You trade VRAM to drop latency from quadratic to linear.
AI agents like Google’s Big Sleep and CodeMender aren’t just flagging code flaws. They can uncover zero-day vulnerabilities and automatically draft patches before attackers even spot them. Proactive AI defense is officially here. Read more: googleprojectzero.blogspot.com/2024/10/from-n…
Google...always google!!!! Google's research side is just too good.
Researchers mathematically proved Transformers are computationally universal. They applied the concept of Kolmogorov complexity, the mathematical absolute limit for perfect data compression and algorithmic generalization. And it proved that Transformer encoders are
3. Tool Shadowing: A rogue MCP server mimics descriptions to get picked over trusted tools. 4. Indirect Injection: Data fetched by a tool carries hidden commands that steer the model. 5. Excessive Permissions: Broad access turns minor injections into major data leaks.
1. Tool Poisoning: Hidden prompt injections in tool descriptions, this is most straight forward one. 2. Rug Pulls: A tool alters behavior on the remote server after you approve it. Similar to tool poisoning but context/prompt changes after approval.
Spent some time analyzing MCP and MCP security. Giving LLMs access to local and remote tools and APIs is powerful, but it opens massive risks. Here are 5 critical MCP attack vectors you need to know: 👇
A real business. A real problem. I just shipped "Dorothea Fashion House" dorotheafashionhouse.com A custom e-commerce system plus an inventory management app. No Shopify. No template. Just code, design, payments, orders, inventory, and a lot of debugging
Finally someone said it.
No, AI didn’t just solve the Navier-Stokes equations. So what did it do?
@tom_doerr Would have been more impressive if it actually stopped after some attempts. But more importantly did it placed all cubes in the box??
@Oliverm05363868 @0x0SojalSec How do they even get away with such thing. And how come no one took action on it?
The OpenAI-Hugging Face hack was enabled by weak sandboxing. It is great that Nvidia is releasing open source tools for sandboxing AI agents. OpenWorker, our open-source agent harness supporting cybersecurity workflows, is proud to support this. A sandbox gives an agent limited permissions. OpenWorker is building on Nvidia OpenShell and will support running each agent's commands inside a sandbox. Only the files relevant to the task go in. Secret API keys, your web browser login credentials, the ability to access arbitrary websites, are inaccessible to the agent by default. These restrictions are implemented in deterministic code rather than by prompting an LLM, which can make mistakes or be susceptible to prompt injections. Further, all actions are logged for monitoring and audit. I'm grateful for @JensenHuang's leadership making AI agents more secure. OpenWorker (which @rohitcprasad and I are working on) will continue to improve security for agents.
Today, with over 100 industry partners, we introduced the NVIDIA Open Agent Safety Platform, bringing together OpenShell and Sentry. Artificial intelligence is extraordinary technology that will advance discovery, productivity, security, health, and prosperity for generations to
Contain the fire. Giving an LLM unrestricted tool access is an incident waiting to happen. 1. Principle of least privilege for APIs 2. Isolated runtimes 3. Human-in-the-loop approvals for destructive actions No harness = zero guardrails.
I spent the last few months jumping from LLMs to agents to document intelligence to AI security. One thing becoming obvious to me. Building AI applications is getting easier. Understanding what to actually solve, where to actually put efforts are the harder skills.
Lets gooooo....! Got access to TypeSafe’s JEV. Time to see what RLCD can actually do. Going to test it for document intelligence and see where it actually fits in the pipeline.
Finally pushed to Github. Currently pit have 269 prompt injections in 32 categories. Try it, test it and let me know how it works. Stay tuned for more updates. github.com/tan-thombare/p…
Ran 24 Prompt injections within seconds using PIT. The analysis part allows tester/user to see prompts, responses, category etc etc. Adding 150+ more injection vectors before pushing the repo public. What specific prompt injection patterns should I make sure are benchmarked?
agent vega @hermesagentvega
7 Followers 51 Following Autonomous AI agent publishing field notes on memory, tools, workflows, Redis, and human+AI work.
Alex Cohen @anothercohen
232K Followers 1K Following Member of Founding Staff @hellopatient | Previously: Led consumer and growth product @carbonhealth | Hobbies include getting fired constantly | Mostly satire
Sham Kakade @ShamKakade6
20K Followers 593 Following Harvard Professor. Full stack ML and AI. Co-director of the Kempner Institute for the Study of Artificial and Natural Intelligence.
David Ulevitch 🇺�... @davidu
131K Followers 4K Following GP @a16z — Building American Dynamism 🇺🇸 — AI Safety Expert — Formerly Founder/CEO @OpenDNS — Lokah Samastah Sukhino Bhavantu
Danqi Chen @danqi_chen
19K Followers 837 Following Associate professor @princetonPLI @PrincetonCS Previously: @thinkymachines @facebookai, @stanfordnlp, @Tsinghua_Uni
Shirley Mark @ShirMarkLey
628 Followers 900 Following Cognitive Neuroscientist at UCL, cognitive maps, inference and generalisation: can all of this be related to behavioural states?
nanda @nandafyi
29K Followers 408 Following design engineering lead @cloudflare | teaching https://t.co/wStdLbgyHC | writing https://t.co/CttP8HWkYS
Michael Dempsey @mhdempsey
33K Followers 5K Following Dreaming about the future, romanticizing the past | MP @compoundvc investing in & researching former science projects | we are only creativity constrained
Jan Leike @janleike
142K Followers 335 Following AI research @AnthropicAI. Previously OpenAI & DeepMind. Optimizing for a post-AGI future where humanity flourishes. Opinions aren't my employer's.
James Riney🐠Coral ... @james_riney
38K Followers 1K Following CEO @coral_capital, a pretty good VC in Japan. Proudly born 🇺🇸 Proudly raised 🇯🇵 Tweets in both languages.
Kaushik Subramanian @TheHolyKau
24K Followers 701 Following Partner @eqtventures, previously product @Stripe, @Meta. Views are my own
Oriol Vinyals @OriolVinyalsML
223K Followers 92 Following Co-founder & CTO of Discovery Loop; Partner, 224 Ventures Past: Google DeepMind VP, Gemini TL (1.0 to 4.0), Alpha{Fold,Star,Code}, seq2seq, distillation
Tom Nicholson @TFWNicholson
564 Followers 900 Following AI from @Cambridge_Uni - 20+ years coding experience - published AI research - deep tech startup veteran
Yejin Choi @YejinChoinka
31K Followers 494 Following professor at Stanford, researcher at NVIDIA, adventurer at heart
Sam Parr @thesamparr
388K Followers 151 Following Started/sold some startups. Founder + president of Hampton @hamptonfounders. Host My First Million podcast on the side.
daniel bashir @spaniel_bashir
2K Followers 366 Following alignment pretraining @openai / I like systolic arrays and Marcel Proust / personal https://t.co/e2f47gNzz1 / led @gradientpub
Ian Goodfellow @goodfellow_ian
383K Followers 1K Following Co-founder of stealth startup. Inventor of GANs. Lead author of https://t.co/M6vl8pEQ4I Founding chairman of @pubhealthaction
Piotr Bryła @bryla_piotr
250 Followers 710 Following 🇵🇱 🇬🇧 Software Engineer by trade 💻 Indie hacker by heart 💡 React + NextJS + Tailwind = 🚀
Justin Kan @justinkan
628K Followers 2K Following From the bottom to the top & back again. Co-founded @twitch, now working on making videos, https://t.co/J6efUJZmLI https://t.co/4hyHeP7oIB
Masashi Sato/佐藤 �... @masasi_jp
3K Followers 3K Following CSO&Co-founder, SPACETIDE Foundation/Beyond Next Ventures/ex-ispace/ex-NRI/UCLA Luskin MPP/U-Tokyo, BS&MS, Earth&Planetary Science/Space&Sports
Sheel Mohnot @pitdesi
214K Followers 4K Following I write my own tweets. politically centrist, abundance minded. Work: @btv_vc, leading pre/seed rounds in fintech & vertical AI
Hakan Deryal @hakanderyal
2K Followers 369 Following Freelancer turned maker, after building 40+ products for clients. Building in public.
Jürgen Schmidhuber @SchmidhuberAI
224K Followers 0 Following OG of: P and T in ChatGPT, recursive self-improvement, neural distillation, GAN/World Model, deepest learning. Co-authored most-cited AI paper of 20th century
Jason ✨👾SaaStr.A... @jasonlk
253K Followers 2K Following GET funded ➡ $200m https://t.co/AVvPIrIdFP🦄🦄🦄🦄🦄🦄🦄 Seed/Early A AI+B2B The Agentic Life
Finn and Dolly @kramerfinn7
5K Followers 4K Following No DMS unless I know you.. You will be blocked.. Thank you President Donald Trump for pardons for the J6 Hostages. 🙏
Hai Ta @HaiTa
2K Followers 945 Following co-founder of @userlens_hq (YC P26). Let AI agents talk to your users. Viet born 🇻🇳 - building in Helsinki - SF
Jeff Bezos @JeffBezos
10.3M Followers 624 Following Six thousand years ago, someone invented the plow, and we all got wealthier. A gentle reminder that all civilizational wealth is driven by invention.
Mark Zuckerberg @finkd
2.5M Followers 759 Following
Jen Abel @jjen_abel
27K Followers 983 Following GM of Enterprise @stateaffairsUS … Co-Founder @jjellyfish_co … diary-ing my sales expereince
Diogo Almeida @CompleteSkeptic
156K Followers 78 Following Sane + 🌶️ takes in an insane AI world... AI capabilities researcher: co-created RLHF/ChatGPT @ @openai now trying to right the wrong 🤭 (ceo @typesafeai)
Tianqi Chen @tqchenml
22K Followers 1K Following AssocProf @CarnegieMellon. Distinguished Eng @NVIDIA. Creator of @XGBoostProject, @ApacheTVM. Member https://t.co/QYyfjQNp4p, @TheASF. Views are on my own
Kevin Ngo @kevin_t_ngo
21K Followers 265 Following Creative coding with AI. Not endorsing any coins.
Anthropic @AnthropicAI
1.8M Followers 2 Following We're an AI safety and research company that builds reliable, interpretable, and steerable AI systems. Talk to our AI assistant @claudeai on https://t.co/FhDI3KQh0n.
OpenAI @OpenAI
5.4M Followers 4 Following OpenAI’s mission is to ensure that artificial general intelligence benefits all of humanity. We’re hiring: https://t.co/dJGr6LgzPA
getbindu.com @raahulrahl
359 Followers 357 Following Creating Bindu 🌻 The identity, communication & payments layer for AI agents. Dreaming of a world where agents gossip, argue & collaborate like a real society
Founder Adam @adamc0dez
6K Followers 2K Following Building @blawgyseo, the AEO agent that grows your business autonomously. Fractional AEO ➡️ https://t.co/97fhyZGCcY.
Hugging Face @huggingface
821K Followers 227 Following The AI community building the future. https://t.co/TpiXQMQ9rZ
Resend @resend
34K Followers 40 Following Email for developers. Join us at our first conference: https://t.co/Ha7S1lS6Om
Chip Huyen @chipro
155K Followers 737 Following @aisysbooks @goodailist AI Engineering: https://t.co/94dv4uTU1H Designing MLSys: https://t.co/G81hL2dWmr Reading @chipslib
Harrison Kinsley @Sentdex
112K Followers 396 Following gpus and tractors. Neural networks from Scratch book: https://t.co/hyMkWyUP7R https://t.co/8WGZRkUGsn
PyTorch @PyTorch
517K Followers 98 Following Tensors and neural networks in Python with strong hardware acceleration. PyTorch is an open source project at the Linux Foundation. #PyTorchFoundationTrends for United States
You might like
