Hacker News

Top stories

Live mirror
30 storiesupdated just nowView source snapshot
  1. Training a 4B model to produce 81% faster query plans than Postgres(rohanbansal.com ↗)
    213 pointsby 2 hours ago
    33comments
  2. Xiaomi Mimo 2.6 live post-training dashboard(xiaomi.com ↗)
    97 pointsby 1 hour ago
    27comments
  3. macOS 27 Golden Gate – Review(arstechnica.com ↗)
    48 pointsby 1 hour ago
    31comments
  4. Small programming tricks(will-keleher.com ↗)
    295 pointsby 5 hours ago
    153comments
  5. Breaking the 1.58-bit Barrier for Ternary LLMs(arxiv.org ↗)
    17 pointsby 31 minutes ago
    —discuss
  6. Accurate Models of AMD Matrix Cores(arxiv.org ↗)
    41 pointsby 2 hours ago
    5comments
  7. Reversing Factorio's RNG(gegell.github.io ↗)
    56 pointsby 2 hours ago
    5comments
  8. Vectorized and performance-portable Quicksort (2022)(googleblog.com ↗)
    151 pointsby 2 hours ago
    24comments
  9. Performance Improvements in .NET 11(devblogs.microsoft.com/dotnet ↗)
    58 pointsby 2 hours ago
    3comments
  10. How good are frontier models at physics?(arxiv.org ↗)
    37 pointsby 2 hours ago
    12comments
  11. AWS says it can't restore some data from mideast facilities struck by Iran(wsj.com ↗)
    52 pointsby 4 hours ago
    13comments
  12. Dream-RSI: Recursive Self-Improvement through Evolving Worlds(arxiv.org ↗)
    163 pointsby 7 hours ago
    48comments
  13. Anatomy of a Texture(agentlien.github.io ↗)
    48 pointsby 7 hours ago
    9comments
  14. WalShadow: Sub-second Postgres replication to ClickHouse from physical WAL(clickhouse.com ↗)
    17 pointsby 1 hour ago
    2comments
  15. Mistral X Mozilla: Private, Multilingual AI Browsing(mistral.ai ↗)
    492 pointsby 13 hours ago
    178comments
  16. Show HN: An e-ink frame that hears birds and draws them as 1800s illustrations(github.com/arnegiacomo ↗)
    2003 pointsby 1 day ago
    234comments
  17. The Siberian Ice Maiden and the Scythian World(patrickwyman.substack.com ↗)
    39 pointsby 4 hours ago
    1comments
  18. Japan's book scene is moving from bookstores to libraries(untranslatedjp.substack.com ↗)
    36 pointsby 4 hours ago
    1comments
  19. Anecdotally, programmers dislike "reduce"(evanhahn.com ↗)
    35 pointsby 8 hours ago
    72comments
  20. Training Text-to-Image Models 3.6× Faster(linum.ai ↗)
    12 pointsby 2 hours ago
    1comments
  21. Show HN: Restarted – a 2026 remake of the classic 2015 startup generator(restarted.io ↗)
    11 pointsby 2 hours ago
    1comments
  22. Tell the speakers that you liked their talks(ohhelloana.blog ↗)
    248 pointsby 7 hours ago
    69comments
  23. Show HN: AttaLambda: a language where types and data are made of untyped lambdas(attalambda.com ↗)
    26 pointsby 5 hours ago
    —discuss
  24. A warning about 'model welfare'(mustafa-suleyman.ai ↗)
    159 pointsby 7 hours ago
    402comments
  25. Kyber (YC W23) Is Hiring a Forward Deployed Engineer(ycombinator.com ↗)
    9 hours ago
    —discuss
  26. The DeepMind Institute(deepmind.com ↗)
    101 pointsby 6 hours ago
    33comments
  27. Claude Cowork and chat are now one Claude(claude.com ↗)
    177 pointsby 5 hours ago
    184comments
  28. Douglas Adams and the exterminated Doctor Who adventure(bbc.co.uk ↗)
    67 pointsby 10 hours ago
    62comments
  29. How big are factorials?(thegreenplace.net ↗)
    70 pointsby 8 hours ago
    30comments
  30. Randomized query complexity can beat certificate complexity(arxiv.org ↗)
    3 pointsby 1 hour ago
    —discuss

Training Text-to-Image Models 3.6× Faster

12 pointsby 4h agolinum.ai
Open original View on HN
1 comments
34m agoHN ↗

Author here, feel free to drop questions below. Will try to answer to best of my ability!

Open thread on HN ↗