Seanku's profile picture. Mgmt.Consultant/Business Exec. w/ interests in Science,Tech,IT & Fintech. #ML,#AI, #Cybersec #Robotics,#Cloudcomputing #IoT, #BigData,#DevOps, #Startups

Pinaki Dasgupta ,MBA ✨

@Seanku

Mgmt.Consultant/Business Exec. w/ interests in Science,Tech,IT & Fintech. #ML,#AI, #Cybersec #Robotics,#Cloudcomputing #IoT, #BigData,#DevOps, #Startups

Public Report: Google Private AI Compute Review nccgroup.com/research-blog/…


Pinaki Dasgupta ,MBA ✨ reposted

here is maya1, our open source voice model: We’re building the future of voice intelligence @mayaresearch_ai team is incredible; amazing work by the team. remarkable moment.

Dheemanthredy's tweet image. here is maya1, our open source voice model:   

We’re building the future of voice intelligence

@mayaresearch_ai  team is incredible; amazing work by the team.  

remarkable moment.

Pinaki Dasgupta ,MBA ✨ reposted

Holy shit... this might be the next big paradigm shift in AI. 🤯 Tencent + Tsinghua just dropped a paper called Continuous Autoregressive Language Models (CALM) and it basically kills the “next-token” paradigm every LLM is built on. Instead of predicting one token at a time,…

rryssf_'s tweet image. Holy shit... this might be the next big paradigm shift in AI. 🤯

Tencent + Tsinghua just dropped a paper called Continuous Autoregressive Language Models (CALM) and it basically kills the “next-token” paradigm every LLM is built on.

Instead of predicting one token at a time,…

Pinaki Dasgupta ,MBA ✨ reposted

Two 23 year old Indians just dropped the #2 open-weight AI voice model in the world, trained purely on free credits! Maya1 is #20 globally, better than even Google's best. 3B params, runs on one GPU and does 20+ emotions with < 100ms latency You can just do things.


Deep sequence models tend to memorize geometrically; it is unclear why

important research paper from google... "LLMs don't just memorize, they build a geometric map that helps them reason" according to the paper: – builds a global map from only local pairs – plans full unseen paths when knowledge is in weights; fails in context – turns a…

slow_developer's tweet image. important research paper from google...

&quot;LLMs don&apos;t just memorize, they build a geometric map that helps them reason&quot;

according to the paper:

– builds a global map from only local pairs
– plans full unseen paths when knowledge is in weights; fails in context
– turns a…


SPICE: Self-Play In Corpus Environments Improves Reasoning #self_Improving #AI

Holy shit… Meta might’ve just solved self-improving AI 🤯 Their new paper SPICE (Self-Play in Corpus Environments) basically turns a language model into its own teacher no humans, no labels, no datasets just the internet as its training ground. Here’s the twist: one copy of…

rryssf_'s tweet image. Holy shit… Meta might’ve just solved self-improving AI 🤯

Their new paper SPICE (Self-Play in Corpus Environments) basically turns a language model into its own teacher no humans, no labels, no datasets just the internet as its training ground.

Here’s the twist: one copy of…


From Pixels 2 words : native VLM

NEO just proved every major AI lab built their vision models wrong 💀 OpenAI, Google, Anthropic... they all use the same approach: train a vision encoder, bolt it onto an LLM, pray the alignment works. NEO said "what if we just... didn't do that?" and built a native…

rryssf_'s tweet image. NEO just proved every major AI lab built their vision models wrong 💀

OpenAI, Google, Anthropic... 

they all use the same approach: train a vision encoder, bolt it onto an LLM, pray the alignment works.

NEO said &quot;what if we just... didn&apos;t do that?&quot;

and built a native…


ChatGPT Atlas vs Perplexity Comet: How the two AI-first web browsers compare in terms of #AI capabilities and agentic mode - moneycontrol.com/technology/cha…


DeepSeek-OCR , a system that turns long docs into small pictures so that AI models can see & understand a lot more info., much faster & cheaper, than if they had to read every single word. @deepseek_ai

🚨 DeepSeek just did something wild. They built an OCR system that compresses long text into vision tokens literally turning paragraphs into pixels. Their model, DeepSeek-OCR, achieves 97% decoding precision at 10× compression and still manages 60% accuracy even at 20×. That…

godofprompt's tweet image. 🚨 DeepSeek just did something wild.

They built an OCR system that compresses long text into vision tokens  literally turning paragraphs into pixels.

Their model, DeepSeek-OCR, achieves 97% decoding precision at 10× compression and still manages 60% accuracy even at 20×. That…


Some #AI tools you need 2 know.

Key AI tools to use for things like video generation, data visualization, & coding assistance, v/René Remsik.

MIT_CSAIL's tweet image. Key AI tools to use for things like video generation, data visualization, &amp;amp; coding assistance, v/René Remsik.


Pinaki Dasgupta ,MBA ✨ reposted

fantastic simple visualization of the self attention formula. this was one of the hardest things for me to deeply understand about LLMs. the formula seems easy. you can even memorize it fast. but to really get an intuition of what the Q,K,V represent and interact, that’s hard.


AgentKit from @OpenAI allows u 2 build , deploy, and optimize agentic workflows.

Introducing AgentKit—build, deploy, and optimize agentic workflows. 💬 ChatKit: Embeddable, customizable chat UI 👷 Agent Builder: WYSIWYG workflow creator 🛤️ Guardrails: Safety screening for inputs/outputs ⚖️ Evals: Datasets, trace grading, auto-prompt optimization



Bitcoin hits all-time high above $125,000 reuters.com/world/asia-pac…

Seanku's tweet image. Bitcoin hits all-time high above $125,000 reuters.com/world/asia-pac…

DeepSearch: Overcome the Bottleneck of #reinforcementlearning w/ Verifiable Rewards via Monte Carlo Tree Search

How do you train small reasoning models more effectively? This is a problem many AI devs run into. RL fine-tuning, in general, tends to plateau, especially for 1–2B models. I think DeepSearch offers a really clean approach here. It takes the idea of Monte Carlo Tree Search…

omarsar0's tweet image. How do you train small reasoning models more effectively?

This is a problem many AI devs run into. RL fine-tuning, in general, tends to plateau, especially for 1–2B models.

I think DeepSearch offers a really clean approach here. It takes the idea of Monte Carlo Tree Search…


Tinker is a training API for Builders/Developers/Researchers etc. thinkingmachines.ai/tinker/


Caltech physicists have created the largest neutral-atom quantum computer to date, trapping 6,100 cesium atoms as qubits in a single array. #quantumcomputing

Caltech physicists have created the largest qubit array ever assembled: 6,100 neutral-atom qubits trapped in a grid by lasers. Previous arrays of this kind contained only hundreds of qubits. caltech.edu/about/news/cal…



Meet shortcut.ai -- an AI agent that can do anything a human can on Excel. It has all of Excel's features, but it can do your work for you

tryshortcut.ai

Excel AI Automation | Shortcut

Automate Excel tasks with AI. Build financial models, generate formulas, and save 10-15 hours per week on reporting.

🚨 BREAKING: AI can now do your spreadsheets work Spoiler: you won’t be using Excel & Google Sheets anymore. Shortcut is the new way to work with data. It even beats McKinsey analysts in Excel challenges 89.1% of the time 👇



Building an AI agent from scratch in #Python How to implement a single #AI agent with an #LLM API and no frameworks

Ever wondered what's actually inside an AI agent? Almost all the resources on agents hide behind abstractions and frameworks. This new blog post builds agents from scratch in Python. You will realize how simple Agents are under the hood.

Hesamation's tweet image. Ever wondered what&apos;s actually inside an AI agent?

Almost all the resources on agents hide behind abstractions and frameworks. 

This new blog post builds agents from scratch in Python. You will realize how simple Agents are under the hood.


Introducing Claude Sonnet 4.5 Model for building complex agents.

Introducing Claude Sonnet 4.5—the best coding model in the world. It's the strongest model for building complex agents. It's the best model at using computers. And it shows substantial gains on tests of reasoning and math.

claudeai's tweet image. Introducing Claude Sonnet 4.5—the best coding model in the world.

It&apos;s the strongest model for building complex agents. It&apos;s the best model at using computers. And it shows substantial gains on tests of reasoning and math.


Loading...

Something went wrong.


Something went wrong.