AI visibility is now part of software discovery. @wynter_com’s 2026 CMO Software Buying Survey shows a clear pattern. 84% of CMOs use AI tools for discovery
→ View original post on X — @coremention, 2026-03-12 12:38 UTC

By
–
AI visibility is now part of software discovery. @wynter_com’s 2026 CMO Software Buying Survey shows a clear pattern. 84% of CMOs use AI tools for discovery
→ View original post on X — @coremention, 2026-03-12 12:38 UTC

By
–
LLM2Vec-Gen represents a major paradigm shift for embeddings/retrieval. Why encode the query when the LLM already knows what to look for and can directly produce an embedding for it? Best part: it’s self-supervised, and it does all of this while the LLM remains completely frozen. Think about it: "solve x² + 3x − 4 = 0" has zero reasoning in it. But the LLM's response does. By encoding the response, the embedding captures the reasoning — and the better the LLM reasons, the better the embedding. This is why our results scale with model size. As LLMs get smarter, our embeddings automatically get better. LLM2Vec-Gen is also the first demonstration of the promise of @ylecun's JEPA for text embeddings. The alignment loss is JEPA — predict in representation space, not token space. The reconstruction loss goes beyond — it keeps embeddings decodable. This paradigm shift opens new frontiers: 🔬 Can we build a full JEPA for language where the teacher and student are the same LLM? ⚡ Can LLMs reason in compressed space without ever generating text? 🤖 Can agents reason in compression tokens and carry that directly into retrieval? 💬 Can agents talk to each other in compression tokens instead of text — dense, fast, and still human-readable? LLM2Vec-Gen is a first step toward all four. Vaibhav Adlakha (@vaibhav_adlakha) Your LLM already knows the answer. Why is your embedding model still encoding the question? 🚨Introducing LLM2Vec-Gen: your frozen LLM generates the answer's embedding in a single forward pass — without ever generating the answer. Not only that, the frozen LLM can decode the embedding back into text. 🏆 SOTA self-supervised embeddings 🛡️ Free transfer of instruction-following, safety, and reasoning — https://nitter.net/vaibhav_adlakha/status/2032065008603951187#m
→ View original post on X — @hugo_larochelle, 2026-03-12 12:37 UTC

By
–
2/ Want to test it out yourself without installing anything? You can run it right now in your browser via @Huggingface
's Spaces → https://
huggingface.co/spaces/Yathart
hS/LuxTTS
…

By
–
Oh man, that's wild. Perplexity's CTO explains that they don't use MCP. But I've been convinced of one thing for several weeks now: aside from Supabase's MCP, I'm not using MCP either. Why bother with MCP when agents can just call APIs directly in bash? Before, I got the point:
By
–
Soul Cast = Cinema Studio 2.0
— Jouhatsu | AI Influence Operator (@Jouhatsu_ai) 12 mars 2026
Vous créez des acteurs IA COMPLÈTEMENT personnalisables.
Pas juste des deepfakes.
Des personnages construits pour le storytelling. pic.twitter.com/MwwOkf8nfS
Soul Cast = Cinema Studio 2.0 Vous créez des acteurs IA COMPLÈTEMENT personnalisables. Pas juste des deepfakes. Des personnages construits pour le storytelling.

By
–
This guy literally just gave Claude (freaking) superpowers. Meet "Superpowers", a open-source framework that forces Claude Code to code with discipline How it works:
✦ Refuses to blindly start typing
✦ Asks Socratic questions first
✦ Teases out a rock-solid design spec
By
–
CLI: Good for single-user mode, hackers.
MCP: Good for multi-user mode, teams/enterprises.
By
–
CLI: Good for single-user mode, hackers.
MCP: Good for multi-user mode, teams/enterprises.
By
–
If you want power, you want to connect to more data sources. If you want to connect it to more data sources, please use something that gives you some semblance of controls. If you want controls, you want MCPs and preferably a gateway. CLIs are much harder to govern!
By
–
If you're trying to figure out how marketing, sales, eng, finance can connect more data sources to AI, please use MCPs, or something with controls. If you're hacking away on your claw, CLIs are awesome.