New video is live.
Most AI labs use synthetic data or distillation because it is cheap, fast, and effective. Microsoft just made a very different bet with MAI-Thinking-1: cleaner pre-training data, no third-party teacher model, and a much more transparent report than usual. It
LLMS
-
Microsoft’s MAI-Thinking-1 uses cleaner data, no teacher model
By
–
-
LangSmith Engine identifies issues, clusters patterns, and fixes agents
By
–
That's why we built LangSmith Engine, an agent built to improve your agent. Engine identifies issues from traces, clusters patterns, drafts fixes, and proposes evals to prevent regressions. Connect your tracing project + start improving your agent today:
-
Aisle preceded Mythos in LLM vulnerability detection using small open model
By
–
Bien que Mythos ait popularisé l'idée de détecter des vulnérabilités avec les LLM, Aisle le faisait plus tôt. D'un point de vue ingénierie, il est intéressant de constater qu'un petit modèle à poids ouvert avec un système de recherche structuré est compétitif pour cette tâche.
-
Cursor adds team leaderboard for plugins, skills, MCPs with one-click install
By
–
Cursor now shows you a leaderboard of the most popular plugins, skills, and MCPs across your team.
— Cursor (@cursor_ai) 23 juin 2026
Add any to your setup with one click from the new Customize page. pic.twitter.com/1DXQXkGTVnCursor now shows you a leaderboard of the most popular plugins, skills, and MCPs across your team. Add any to your setup with one click from the new Customize page.
-
Cursor and MagicPath for Interactive Component Variations
By
–
The best way to design and build with your repository is Cursor/Codex + MagicPath.
— Pietro Schirano (@skirano) 23 juin 2026
Watch me import components from my repo into the MagicPath canvas inside Cursor, then explore completely different variations.
All while they stay interactive and faithful to the source code 👇 pic.twitter.com/ZNMZoaA8AvThe best way to design and build with your repository is Cursor/Codex + MagicPath. Watch me import components from my repo into the MagicPath canvas inside Cursor, then explore completely different variations. All while they stay interactive and faithful to the source code
-
Jack Clark predicts recursive AI improvement by 2028
By
–
Recursive self-improvement: Anthropic co-founder Jack Clark says it could arrive by 2028.
— Chubby♨️ (@kimmonismus) 23 juin 2026
AI systems could help invent their own successors – with Claude 10 building Claude 11, and so on – potentially “without any researchers involved.” pic.twitter.com/MYzkxIbge5Recursive self-improvement: Anthropic co-founder Jack Clark believes it could arrive by 2028. AI systems could help invent their own successors – with Claude 10 building Claude 11, and so on – potentially "without
-
Noam Shazeer joins OpenAI; Character.AI valued at 1 billion despite difficulties
By
–
Noam Shazeer, co-author of "Attention Is All You Need", left Google to join OpenAI, while his startup, Character.AI, reached a valuation of one billion dollars in March 2023. Despite the company's difficulties in establishing revenue
-
Ling and Ring Technical Report 2.6 on Large-Scale Agentic Intelligence
By
–
Ling and Ring Technical Report 2.6 Efficient and Instantaneous Agentic Intelligence at the Scale of Trillions of Parameters
-
The mechanical accumulation of useless code in agentic coding
By
–
With agentic coding, complexity accumulates mechanically: useless code ends up in the codebase, migrates to the context window, degrades the model's reasoning capabilities, and leads to even more useless code (often to
-
Nemotron Office Hours: The Nemotron 3 Model Family
By
–
Nemotron Office Hours: The Nemotron 3 Model Family | Nemotron Labs