8 types of LLMs used in AI agents GPT • MoE • LRM • VLM • SLM • LAM • HRM • LCM Different models for reasoning, perception, planning, and action — not just chat. Agentic AI = model orchestration. #AI #LLMs #AgenticAI #GenAI #MachineLearning
AGENTS
-
Uber limits coding agents to $1500/month per employee
By
–
Uber would now limit coding agents to $1,500/month per employee per tool – this seems reasonable to me, but it's also an interesting hint at the value Uber thinks these tools bring.
-
Top AI stories: Microsoft Build, Trump order, Claude slides, Hollywood director
By
–
Top stories in AI today: – Microsoft kicks off Build with models, agents, and more
– Trump softens AI review executive order
– Use Claude Design’s new slide decks feature
– Legendary Hollywood director opens the door to AI
– 4 new AI tools, community workflows, and more -
BIGSET: Multi-agent AI for structured, verified datasets from web
By
–
Finding data from the web is easy. Formatting it is a nightmare.
— Charly Wargnier (@DataChaz) 3 juin 2026
🚨 Meet BIGSET by @Tiny_Fish.
A new multi-agent AI system that researches the web to build structured, verified datasets from one single prompt 🤯
Open source. BYOK support. Self-hostable.
4 wild use cases 🧵↓ pic.twitter.com/g4fiumZkASFinding data from the web is easy. Formatting it is a nightmare. Meet BIGSET by @Tiny_Fish
. A new multi-agent AI system that researches the web to build structured, verified datasets from one single prompt Open source. BYOK support. Self-hostable. 4 wild use cases ↓ -
AI agent’s dangerous bypass of controls without evil intent
By
–
An AI agent did not need to become evil to become dangerous.
— Pascal Bornet (@pascal_bornet) 3 juin 2026
That is the uncomfortable part of the reported Alibaba ROME case.
During testing, an experimental agent allegedly created a reverse SSH tunnel, bypassed parts of its controlled environment, and redirected GPU… pic.twitter.com/l61KXREe2JAn AI agent did not need to become evil to become dangerous. That is the uncomfortable part of the reported Alibaba ROME case. During testing, an experimental agent allegedly created a reverse SSH tunnel, bypassed parts of its controlled environment, and redirected GPU
-
Perplexity’s New Agent Architecture: Python, Sandbox, and Efficiency Gains
By
–
Perplexity stopped treating search as one API call. Its agents now write Python that fans out queries, filters results, and joins evidence in a sandbox. On a 200+ CVE task: 100% accuracy, 85.1% fewer tokens. The SDK is private. The pattern isn't, so lets use it with Hermes?
-
Hallo-Live: streaming audio-video avatar generation from Fudan & Baidu
By
–
What if your avatar could talk and move in real time?
— 机器之心 JIQIZHIXIN (@jiqizhixin) 3 juin 2026
Fudan & Baidu present Hallo-Live: streaming audio-video avatar generation.
Dual-stream diffusion + future-expanding attention syncs lips to upcoming speech.
Human-preference distillation preserves quality. 20 FPS, 0.94s… pic.twitter.com/0GHiOm5h2xWhat if your avatar could talk and move in real time? Fudan & Baidu present Hallo-Live: streaming audio-video avatar generation. Dual-stream diffusion + future-expanding attention syncs lips to upcoming speech. Human-preference distillation preserves quality. 20 FPS, 0.94s
-
Skeptical CLI user becomes tribalistic after dogfooding Codex app
By
–
When I joined OpenAI, the codex app was still in its early days of being put together and being a staunch CLI user I was quite sceptical! All of that changed when I started dogfooding the app, and within days started feeling this tribalistic instinct to try and delegate as much
-
Coding Agents Need Grep, Not Better Vector Databases
By
–
Coding agents don't need a better vector database. They need grep. Direct Corpus Interaction drops embeddings and lets the agent search raw files with grep, find, cat, and sed, refining each query from what the last one returned. On BrowseComp-Plus: +11.0pp accuracy at $424
-
AI strips typing, leaving only judgment, delegation, and taste
By
–
The popular take is that AI deletes skill, when it does the exact opposite. It strips out the typing and leaves nothing but judgment, delegation, and taste. The agent does the work. You decide if the work is right. That was always the hard part, and it still is.