Sources:
Anthropic, When AI builds itself:
https://anthropic.com/institute/recursive-self-improvement
… SkillOpt (arXiv): https://arxiv.org/abs/2605.23904 SkillSmith (arXiv): https://arxiv.org/abs/2606.01314 MOSS (arXiv): https://arxiv.org/abs/2605.22794 Co-Scientist, Google DeepMind: https://deepmind.google/blog/co-scientist-a-multi-agent-ai-partner-to-accelerate-research/
… Gary Marcus, No need to panic
LLMS
-
AI Self-Improvement: New Advances and Outlooks
By
–
-
Karpathy’s 2015 RNN coding experiments and 2017 gradient descent advances
By
–
2015: Karpathy publishes one of the earliest experiments (if not the first public article/blog with practical artifacts) on language models for coding (The Unreasonable Effectiveness of Recurrent Neural Networks). 2017: @karpathy again: gradient descent can write code better
-
NVIDIA Nemotron Coalition grows with new AI partners
By
–
The NVIDIA Nemotron Coalition continues to grow. We're excited to welcome new members: @hcompany_ai
, @NousResearch
, and @PrimeIntellect
. And a big thank you to our existing members: @bfl_ai
, @cursor_ai
, @LangChain
, @MistralAI
, NAVER Cloud, @perplexity_ai
, @ReflectionAI_
, -
tt-vscode-toolkit: interactive learning environment
By
–
Introducing tt-vscode-toolkit, an interactive learning environment designed to accelerate developing on Tenstorrent hardware. The toolkit brings project templates directly into VS Code, with lessons covering model deployment, agent frameworks, video generation, and more. Get
-
LangSmith LLM Gateway enforces cost limits and detects PII
By
–
Enforce cost limits Detect PII Act on violations
…All without leaving LangSmith. ICYMI: LangSmith LLM Gateway is the runtime governance layer that lives where you already build, observe, and evaluate your agents in LangSmith. -
Google DeepMind releases Gemma 4 QAT models for on-device use
By
–
Google DeepMind released new Gemma 4 QAT models that make the model family much more efficient for local, on-device use. Using Quantization-Aware Training, the models are trained with compression in mind, which reduces memory needs while preserving more quality than standard
-
AI monetization: human judgment, not the tool, is the differentiator.
By
–
That test has nothing to do with the tool. It measures whether a human shaped the result. The channels that kept their monetization used the exact same AI as the ones that got wiped. The only variable was the judgment sitting between the model and the upload.
-
Cursor AI: Visual Prompts to Bridge Vision and Understanding
By
–
Give Cursor visual prompts to shrink the gap between what you see and what the agent understands. Learn more:
-
Sarvam AI builds full-stack ‘Made in India’ AI platform with NVIDIA H100 GPUs
By
–
Sovereign AI at population scale isn’t theory anymore, it’s shipping. Sarvam AI is building a full-stack, “Made in India” AI platform that: Trains 100B+ parameter MoE models efficiently across 4,096+ NVIDIA H100 GPUs Delivers millisecond-level, multilingual voice
-
Codex imports UI from repo, recreates as editable interactive canvas
By
–
Codex can import UI from your repository into MagicPath, understand its components and design system, and recreate it as an editable, interactive experience on the canvas.
— Pietro Schirano (@skirano) 5 juin 2026
Here is our entire left-side panel with ALL the working functionality. pic.twitter.com/CIKPsq5n9nCodex can import UI from your repository into MagicPath, understand its components and design system, and recreate it as an editable, interactive experience on the canvas. Here is our entire left-side panel with ALL the working functionality.