• [Blog] Utkarsh’s insights on the mathematical impossibility of some expectations of AI agents in 2025. TLDR: error compounding, token economics, tool engineering.

  • [Blog] OpenAI projected to release GPT-5 in less than a month, while their open-weights model is still delayed (Blog, Tweet)

  • [Blog] OpenAI publishes a report on the economic impacts of ChatGPT. The finance & insurance sector contributes the most enterprise adoption.

  • [Blog] Clément Delangue, HuggingFace CEO, discusses the inclusion of “open-source and open-weight AI” in the White House’s AI Action Plan (PDF).

  • [Paper] NVIDIA paper argues Small Language Models are the future of Agentic AI.

  • [Blog] OpenAI signs $30B / yr deal with Oracle to develop more Stargate data centers in the US. Official blog post from OpenAI.

  • [Blog] OpenAI raises another $8.3B at $300B valuation.

  • [Blog] Figma more than triples in NYSE debut after selling shares at $33.

  • [Blog] Jared recollects his first fundraising experience after co-founding GroupMe.

  • [Blog] ChatGPT Study Mode.

  • [Post] Photo sharing app that runs on SMS.

  • [Blog] Recency of citations in AI results.

  • [Paper] System 1 vs System 2 thinking: Hierarchical Reasoning Model.

  • [Paper] Microsoft publishes a speculative study on the most (ir-)replaceable human jobs in the future, based on signals collected from Bing Copilot conversations. TLDR: information-gathering jobs more replaceable, physical labor less.

  • [Paper] Self-evolving agents.

  • [Quote] AI Snake Oil.

Have a great week!

Updated:

Comments