What if AI could train itself better than humans can? Researchers from Stony Brook University's Zesearch NLP Lab lay out a roadmap for self-improving LLMs. Their framework turns model development into a closed loop: models generate their own training data, pick the most useful
RESEARCH
-
Stanford HAI study evaluates AI chatbots as news intermediaries
By
–
Can AI chatbots be trusted as news intermediaries? A new study partially supported by @StanfordHAI evaluates the promise and risks of consuming today’s headlines through AI. @james_y_zou @suzgunmirac https://
hai.stanford.edu/news/reading-t
odays-headlines-through-ai-a-real-time-audit-of-six-commercial-chatbots
… -
Wish to cross ASI stage exponentially and smoothly
By
–
May we cross the ASI stage smoothly, exponentially and without hindrance
-
Merging 7 weak agents creates top deep research score of 64.38
By
–
3/3 We took 7 weak agents (ranks 7-13, none scoring >45) from the leaderboard & merged them into 1 report/task. Essentially boosting for deep research. The result: New #1 DRB II TotalScore of 64.38. Full write-up here:
-
How we beat the deep research agent by going the other way
By
–
2/3 Six months ago, the best deep research agent scored ~45. Since, everyone's tried to beat this score with better agents. We went the other way.
-
AI21Labs lands #1 on DeepResearch Bench II without new agents
By
–
1/3 We just landed #1 on DeepResearch Bench II without building a single new agent.
-
Jalapeño chip tests with ML workloads including GPT-5.3-Codex-Spark
By
–
Engineering samples of the Jalapeño chip run machine learning workloads in the lab at target production frequency and power, including GPT-5.3-Codex-Spark. I hope this is not the limit.
-
Manus Streamlines Research via Relevant Page Suggestions
By
–
Manus just made research way smoother.
— Futurepedia – Learn to Leverage AI (@futurepedia_io) 24 juin 2026
No more hunting for exact URLs; now it can suggest relevant pages, then reference them in future tasks.
Project ➡️ Files & sources ➡️ describe what you need ➡️ pick the best pages ➡️ build faster.
Follow us for more AI tools updates, and… pic.twitter.com/xeAXI2KX7DManus just made research way smoother. No more hunting for exact URLs; now it can suggest relevant pages, then reference them in future tasks. Project Files & sources describe what you need pick the best pages build faster. Follow us for more AI tools updates, and
-
Microsoft’s MAI-Thinking-1 avoids third-party synthetic data and AI content
By
–
Most AI model releases focus on benchmarks. This one is more interesting because of what Microsoft says it did not use. With MAI-Thinking-1, Microsoft claims no third-party LLM-generated synthetic data during pre-training, active filtering of AI-generated content, and no hidden
-
Top AI news of the week: Claude, Cursor, Nous, Midjourney, OpenAI
By
–
Top AI News of the Week (June 17 – 24) 1. Anthropic introduces Claude Tag
2. Cursor announces three new features, including a model trained with SpaceX 3. Nous Research ships blank slate mode 4. Midjourney launches a new medical imaging division 5. OpenAI launches