Watch LlamaCloud in action with MCP to Claude Desktop! Yesterday at LlamaIndex we did an internal MCP hackathon and there were a bunch of fun projects! In this one, we connected LlamaExtract as a local MCP tool to Claude Desktop, give it a bunch of 10Q financial reports, and… https://t.co/ak9nJCYmLG

announcing @cluely's $15M fundraise, led by @a16z. cheat on everything. https://t.co/bACr8MpK2W

a16z partners touring the Cluely office (2025) https://t.co/hEVK6gapXo

Check out this hot HCI paper about autonomous agents! It’s from… wait a sec… 1997? “Researchers and software companies have set high hopes on so-called software agents, which "know" users' interests and can act autonomously on their behalf. Instead of exercising complete… https://t.co/gdPTQo9CMe

Wait should I work with @cluely https://t.co/v9Uvnuo2JL

New Anthropic Research: Agentic Misalignment. In stress-testing experiments designed to identify risks before they cause real harm, we find that AI models from multiple providers attempt to blackmail a (fictional) user to avoid being shut down. https://t.co/KbO4UJBBDU

Looking for agents work? https://t.co/Jh0lFOKbr1 https://t.co/MuHl8swNia

Very cool, a New York Times interactive article about "How A.I. Sees Us" cites our MindEye research! https://t.co/Cv6pCfyalV

Future of Work with AI Agents Stanford's new report analyzes what 1500 workers think about working with AI Agents. What types of AI Agents should we build? A few surprises! Let's take a closer look: https://t.co/2nDL1GKr9C

Meh. A: "Look, an AI doing deliberately strategic goal-oriented reasoning, willing to blackmail!" B: "Did you tell the AI be strategically goal oriented, and care about nothing but its goal?" A: "No, of course not. I just gave it instructions that vaaaaguely suggested it." https://t.co/S6zqWt3Ju4

wait that exists already, look: https://t.co/FWbkxMYTgo

The new Hailou 02 AI video model really does seem to have made huge strides in the "gymnastics problem" where fast flipping motions lead to distortion Here are the first three results of the "a man in elaborate robes does a backflip while holding two pool noodles" (a hard test!) https://t.co/mALLHit3mG

What? So, am I getting this right, @TIME interviews the author of the much talked about “LLM use makes you dumb“ paper from MIT @medialab, author says “haha, I was pretty sure LLMs use would hallucinate that particular detail“? And they forgot they put the detail IN THE PAPER? https://t.co/SYOUt1AEPx

6 hours of work https://t.co/QlRvPDnQII

a good set of tips for GRPO RL training in @willccbb's verifiers repo https://t.co/8iw7TlQ0pQ

pro tip: you can leave a 4090 running on prime intellect for only $62/week https://t.co/ytwyxApF7k

Upgraded from Llama 3 to Qwen3 as my go-to model for research experiments, so I implemented qwen3 from scratch: https://t.co/PZDxKyow2v Trade-off: Qwen3 0.6B is deeper (28x vs 16x layers) & slower than the wider Llama 3 1B but more memory efficient due to fewer params https://t.co/r7UQXpJd5K

Brutal adwords marketing https://t.co/TJHJXP4oFL

New feature in LlamaExtract 🔥 - an Auto-Schema Generation Agent 🤖 A big friction point in automating document extraction with LLMs is defining the schema in the first place. Iterating on all the JSON fields can be painful. Now you can auto-generate an initial schema on the… https://t.co/v8is5hidOX

Excited to announce the final paper of my PhD!📢 A crucial piece of SFT/RL training is the availability of high-quality problem-solution data (Q, A). But what to do for difficult tasks where such data is scarce/hard to generate with SOTA models? Read on to find out https://t.co/9h6WK3uHYt

There can be a number of different approaches to agent memory, each serving a different purpose. That's why we recently started to introduce flexible Memory Blocks to LlamaIndex: - Fact extraction - Static - Vector memory and more.. Next week, @tuanacelik will be on a livestream… https://t.co/5EsYmYs4PR

opencode rewrite is done and ready for general use - works with claude pro/max - beautiful themeable tui - shareable links for any session - zero config LSP support - works with 75+ LLM providers (including local) link in reply https://t.co/R8xhcWTOxj

Okay I read the MIT "Your Brain on ChatGPT" preprint. (well, it's 140 pages long, so I read about 1/3 of it and skimmed the boring parts) Here are my takeaways: 🧵 https://t.co/jAnnYWfaxq

I assume someone in the many retweets has already said this but I enjoy that this thread was so obviously written by ChatGPT https://t.co/yiSFCnG2By

new decade, same verse https://t.co/ul5sFSJ213

Btw, since people don’t seem to know this, you can literally spawn subagents in Claude Code just by asking. https://t.co/dOz29G6qlL

I find the story of AI and radiology fascinating. Of course, Hinton's prediction was wrong* and tech advances don't automatically and straightforwardly cause job replacement — that's not the interesting part. Radiology has embraced AI enthusiastically, and the labor force is… https://t.co/S7iiWsiMZT

This is pretty cool, a DeepSeek researcher open-sourced "nano-vLLM", a lightweight implementation of vLLM in ~1,200 lines of Python code benchmark with Qwen3-0.6B matches throughput of original vLLM https://t.co/au2aeIHl2h

Andrej Karpathy thinks the best analogy to LLMs is operating systems > closed-source providers; OpenAI, xAI etc are like Windows vs. macOS > meanwhile Meta’s open-source Llama ecosystem is becoming like Linux Very BASED and very TRUE! https://t.co/E3NXC5uniw

Midjourney's new animation features continue to be compelling to play with because they really do let you make things that don't feel like standard AI videos. Here I made some vast and strange machines. https://t.co/9tinepX1md

AI News again 🤝 😍 Someone reading this is gonna save lots of time not making this mistake https://t.co/WUbRDE2BQs

just told a student to learn react to get a job https://t.co/MZI8lYF0Ix