HN Explorer

GodelNumbering

↗ HN profile [ 81.5 ms ] full profile
Karma
0
Created
()
Submissions
0
  1. If your agent commits a crime, who is responsible? (signalbloom.ai)
    5 points 71
  2. Ask HN: Anyone interested in building a harness-only benchmark?

    There are a lot of LLM benchmarks but few, if any, harness benchmarks. I am thinking this would be a really good community project to build one. End goal: a leaderboard of harness performance (multiple axis) on a set of…

    5 points 5
  3. Token Arbitrage: Use Sol for coding, Luna for compaction, save 84% (dirac.run)
    3 points 2
  4. Claude gen-5 models show significant regression in BullshitBench (github.com)
    3 points 0
  5. GPT-5.6-terra used 48.5% more context than Mimo-2.5-pro (dirac.run)
    3 points 1
  6. Cache hit rate dropping by 20% doubles your agent's bills (dirac.run)
    2 points 0
  7. Don't Conflate 'Minimal' with Minimal Effort (dirac.run)
    2 points 0
  8. OSS models decisively overtook Proprietary models in openrouter market share (dirac.run)
    4 points 1
  9. Outsourcing plus local AI will soon become more economical vs. frontier labs (signalbloom.ai)
    323 points 91
  10. Cache hit rates of Inference are more meaningful than the headline costs (dirac.run)
    2 points 0
  11. Ask HN: Who needs contributors? (May 2026)

    Looking for contributors to your project? Feel free to post any project that may interest HN readers, with a strong preference towards open source. Please follow this general format: Project Name Project Description…

    2 points 0
  12. Peak lazy engineering: Let the code answer its own questions (dirac.run)
    2 points 0
  13. The Isolator (Helmet) (en.wikipedia.org)
    3 points 0
  14. Show HN: OSS Agent I built topped the TerminalBench on Gemini-3-flash-preview (github.com)

    Scored 65.2% vs google's official 47.8%, and the existing top closed source model Junie CLI's 64.3%. Since there are a lot of reports of deliberate cheating on TerminalBench 2.0 lately…

    393 points 56
  15. Hash anchors and Myers diff and single-token anchors: 60% cheaper AI code edits (dirac.run)
    5 points 1
  16. Show HN: I built Dirac, Hash Anchored AST native coding agent, costs -64.8 pct (github.com)

    Fully open source, a hard fork of cline. Full evals on the github page that compares 7 agents (Cline, Kilo, Ohmypi, Opencode, Pimono, Roo, Dirac) on 8 medium complexity tasks. Each task, each diff and correctness + cost…

    2 points 1
  17. Nvidia's Nemotron 3 Super is a bigger deal than you think (signalbloom.ai)
    2 points 0
  18. Why Task Proficiency Doesn't Equal AI Autonomy (signalbloom.ai)
    2 points 0
  19. Claude Code wipes out a production database (xcancel.com)
    5 points 6
  20. Renaissance Slashes Mega-Cap Tech Exposure in Major Defensive Pivot (signalbloom.ai)
    1 points 0
  21. Show HN: Realtime 13Fs and track live institutional ownership for any ticker (signalbloom.ai)

    What it does: - Polls SEC continuously at small intervals - Fetches all variants of the 13F since the last poll - Resolves every filing to its effective manager (one filing can contain data for multiple filers or proxy…

    1 points 0
  22. Show HN: the entire US ETF market mapped into 280 distinct categories (signalbloom.ai)
    2 points 0
  23. 5 points 0
  24. The math, mechanics, and risks of leveraged ETFs (signalbloom.ai)
    2 points 0
  25. ETFs now hold more than $3.1T worth of just top US companies (signalbloom.ai)
    91 points 9