The team acknowledges that some agents are still very limited each in their own way:
– Deep research cannot interact with webpages (no Textbrowser can)
– Operator has trouble reading through long pages personal take: Operator should theoretically be able to also do deep research
RESEARCH
-
Deep research and Operator limitations; Operator could do deep research
By
–
-

ChatGPT agent live announcement combining Deep Research, Operator, Terminal code editing
By
–
ChatGPT agent live announcement underway, with @sama onboard!
This Agent patches together
– Deep Research (TextBrowser )
– Operator (GUI Agent) – Terminal code editing
to unlock all these capacities together in an agent. -

Gemini’s Tokenization of YouTube Videos for Large Recommender Models
By
–
criminally underrated talk from @devanshtandon_ Gemini powers YouTube's Large Recommender Model by **tokenizing every video on youtube** (SemanticID) – a vocabulary several OoMs larger than English, CONTINUOUSLY PRETRAINED every day. can reason across titles/descriptions, throws
-

OpenAI prepares a unified agentic model
By
–

OpenAI is about to announce a “Unified Agentic Model” Will it be GPT-5 in fact? The one to be released at September? Wondering if “Agentic” may mark another tier after “Reasoning” models.
-

SLMs and Mini-Agents Power Reliable Agentic AI Workflows
By
–
At #DataAISummit, @LaminiAI CEO @realSharonZhou explored how SLMs and mini-agents can power reliable agentic workflows. Learn about memory RAG, MoME-based memory tuning, and real-world use cases like text-to-SQL and code analysis: https://
databricks.com/dataaisummit/s
ession/composing-high-accuracy-ai-systems-slms-and-mini-agents?utm_source=twitter&utm_medium=organic-social&utm_scid=701vp00000d7qrxia3
… -

WebSailor paper uses agentic RL post-training to boost Deep Research scores
By
–
Recent WebSailor paper by Alibaba-NLP, shows how to post-train models for Deep Research – good insights in there, about creating a dataset then training recipe. I particularly like how the agentic RL at the end of post-training improves scores by ~4 p.p. across the board: RL +
-
OpenAI’s Secret Internal Model Nearly Wins World Coding Competition
By
–
OpenAI's Secret INTERNAL Model Almost Wins World Coding Competition… https://
youtu.be/HctuXVQci4E?si
=Zs-KAWUAGyzMpOYw
… via @YouTube -

Opinion Dynamics: Statistical Physics Approaches in AI
By
–
Opinion dynamics: Statistical physics and beyond Starnini et al.: https://
arxiv.org/abs/2507.11521 #ArtificialIntelligence #DeepLearning #MachineLearning -
Memory-Native Language Models: The Missing Architecture Innovation
By
–
again, the AI labs are obsessed with building reasoning-native language models when they need to be building *memory-native* language models – this is possible (the techniques exist)
– no one has done it yet (no popular LLM has a built in memory module)
– door = wide open -

FakePsyho Wins AtCoder World Tour Finals 2025 Heuristic
By
–
Congrats to @FakePsyho for winning AtCoder World Tour Finals 2025 Heuristic Humanity has prevailed (for now!) Thanks OpenAI for sponsoring #AWTF2025, and getting #2 on this grand challenge. Proud of @SakanaAILabs & @AtCoder
’s ALE-Agent for reaching #5, on a shoestring budget!