1. Sharing AI Progress in Mathematics
    (openai.com)

  2. Mistral Large 4
    (mistral.ai)

  3. Decisions API is in public beta
    (developers.openai.com)

  4. EmbeddingGemma 2: An open, lightweight multimodal embedding model
    (blog.google)

  5. Nobel Prize in Physics 2026: Francis Halzen
    (www.nobelprize.org)

  6. OpenTPU – An open-source AI accelerator, developed by AI
    (github.com)

  7. Paramount Skydance has completed its $111B merger with Warner Bros. Discovery
    (arstechnica.com)

  8. Claude Code’s suggested message feature: I think the real customer is the model
    (www.zohaib.cc)

  9. OpenSSH 10.6
    (www.openssh.org)

  10. The Query Transformation Pipeline
    (readyset.io)

  11. Benchmark in Milliseconds
    (matklad.github.io)

  12. Vibecoding Photoshop: Time and pressure
    (andreklein.net)

  13. The Early History of Smalltalk (1993)
    (worrydream.com)

  14. Gleam doesn't compile to Erlang source anymore
    (gleam.run)

  15. LLMs may have helped my RSI
    (vaughanhilts.me)

  16. Erdosproblems.com Succumbs to the AI Onslaught
    (www.erdosproblems.com)

  17. Berthd
    (berthd.app)

  18. When random is not actually random enough
    (ersc.io)

  19. Toronto-Based VPN Provider Plans to Quit Canada over Lawful-Access Bill
    (citizenlab.ca)

  20. Show HN: I turned my iPhone and a $20 smart plug into an f-stop timer
    (peterszentkiralyi.eu)

  21. Subquadratic 3SUM and Subcubic APSP
    (arxiv.org)

  22. Mathematics of Geothermal Energy
    (www.ebsco.com)

  23. System-level ad-blocking in Android
    (kevinboone.me)

  24. Beam: Reflection's 501B open-weight model
    (reflection.ai)

  25. Polars 2.0
    (pola.rs)

  26. What's Earth's dominant species by mass?
    (signoregalilei.com)

  27. Example.com just launched the biggest redesign in decades
    (www.debugbear.com)

  28. Nature's capacity to 'bounce back' when species are lost is overestimated: study
    (phys.org)

  29. Show HN: Parseable, an open observability datalake, handles 100M time-series/min
    (www.parseable.com)

  30. Dust: Pretraining Transformers Without Backpropagation
    (qlabs.sh)