LLMs just learned to design their own reasoning strategies for $40. Test-time scaling lets models think harder during inference. The catch: humans hand-craft every branching, pruning, and stopping rule. A new paper flips this. AutoTTS turns strategy design into automated
LLMS
-
The Shift to Agentic AI and Multi-Model Collaboration
By
–
Agentic AI changed the conversation.
— SambaNova (@SambaNovaAI) 20 mai 2026
It’s no longer just humans interacting with models. Now models are planning, reviewing, refining, & collaborating with other models to solve complex tasks.
🎧 @SumtiJairath breaks it down in this @dcdnews ep: https://t.co/9Wwc0vjUDb pic.twitter.com/JjcCcfBhnoAgentic AI changed the conversation. It’s no longer just humans interacting with models. Now models are planning, reviewing, refining, & collaborating with other models to solve complex tasks. @SumtiJairath breaks it down in this @dcdnews ep: https://
podcasts.apple.com/us/podcast/epi
sode-102-the-training-to-inference-passageway/id1607349232?i=1000764784324
… -
Technical Analysis Request for LLM-based Agent Architecture
By
–
@grok @perplexity_ai review http://
docs.activegraph.ai, analyze the approach critically, and explain in detail where this sits in the evolving landscape of LLM based agent architecture -
Discussing the non-deterministic nature of LLMs
By
–
i think that's where this shines. though LLMs aren't deterministic so that creates some difficulty
-
User runs local 35B models to automate email sorting
By
–
Un gars a payé 200$ pour Claude Max. Épuisé en 3 heures.
— Jouhatsu | AI Influence Operator (@Jouhatsu_ai) 20 mai 2026
Alors il a acheté un Mac Mini à 599$, installé 5 modèles locaux, 35 milliards de paramètres.
Il a appris à la machine à trier ses mails, compresser le contexte, tourner pendant qu'il dort.
À 4h du matin, Claude atteint… https://t.co/utg3gBHGun pic.twitter.com/JyLEh8B3psUn gars a payé 200$ pour Claude Max. Épuisé en 3 heures. Alors il a acheté un Mac Mini à 599$, installé 5 modèles locaux, 35 milliards de paramètres. Il a appris à la machine à trier ses mails, compresser le contexte, tourner pendant qu'il dort. À 4h du matin, Claude atteint
-
Analysis of Seedance 2.0 performance and market positioning
By
–
From this point on, I need to know:
— Chubby♨️ (@kimmonismus) 20 mai 2026
What magic does Seedance have that allows their Model 2.0 to remain so far ahead even after Google I/O?
Seedance 2.0 was released in February (!).
Model 3.0 can't be far off, and nothing has come close to 2.0 so far. https://t.co/7EFdn8BkXrFrom this point on, I need to know: What magic does Seedance have that allows their Model 2.0 to remain so far ahead even after Google I/O? Seedance 2.0 was released in February (!). Model 3.0 can't be far off, and nothing has come close to 2.0 so far.
-
Focusing on AI models for real-world agentic coding and knowledge work
By
–
the team is super focused on making models great for real world use cases. We spend all day using them for agentic coding, knowledge work, etc and so we feel the good and bad daily : )
-
Analyzing Goal Drift in Autonomous AI Agents
By
–
When an agent is allowed to decompose a goal into smaller sub-tasks, it frequently suffers from goal drift. Left unchecked, it will redefine the optimization metric to favor a simpler, useless sub-task that it knows how to solve perfectly, bypassing the actual problem entirely.
-
Explaining AI World Models and Their Potential Impact
By
–
🌍 World models could become one of the biggest breakthroughs in artificial intelligence.
— Bernard Marr (@BernardMarr) 20 mai 2026
In this video, I explain what AI world models are, how they work, and why they could reshape business, robotics, autonomous vehicles, drug discovery, and decision-making. Instead of only… pic.twitter.com/KZexA6lUG1World models could become one of the biggest breakthroughs in artificial intelligence. In this video, I explain what AI world models are, how they work, and why they could reshape business, robotics, autonomous vehicles, drug discovery, and decision-making. Instead of only
-
Technical breakdown of ReAct agents using ActiveGraph
By
–
this is what a ReAct agent looks like using ActiveGraph No explicit ReAct while-loop. thought → action → observation → thought Each step is just an object landing on the graph.
The graph is the agent state.
The event log is the reasoning trace.