Hacker News

New stories

Live mirror
30 storiesupdated just nowView source snapshot
  1. CrankGPT – hand powered offline physical AI box (squeezlabs.github.io)
    —discuss
  2. Show HN: RIFC-light, offline, pseudocode based flow chart generator (github.com/ramathesungod)
    —discuss
  3. I Put 63 Buying Questions to ChatGPT and G2 Was Cited Zero Times (cuescout.com)
    —discuss
  4. Cloudflare: Birthday Week 2026 (cloudflare.com)
    —discuss
  5. How to Tell If Someone Is Cheating on Lichess (chesscheatdetector.com)
    —discuss
  6. Show HN: Jeva.cpp – a llama.cpp fork with JEV-compatible API for all LLMs (github.com/pragmatwice)
    —discuss
  7. Where Trust in Automated Review Comes From (jakelundberg.dev)
    —discuss
  8. Aspiring Firmware Engineer (chrisgammell.com)
    —discuss
  9. Log internet outages and show whether they're on your side or the ISP's (github.com/boazcohenj)
    —discuss
  10. Expert Asterisks (nedbatchelder.com)
    —discuss
  11. Show HN: Decide – Jev decisions in the shell, scripts, and agent skills (github.com/vsekhar)
    —discuss
  12. Analyzing the ClickFix social engineering technique (2025) (microsoft.com)
    —discuss
  13. Galileo's first civil authenticated position fix under spoofing conditions (esa.int)
    —discuss
  14. Dot Plot (Statistics) (wikipedia.org)
    —discuss
  15. Show HN: X402 Inspector (Find misconfigurations in your x402 API) (trango-compute.com)
    —discuss
  16. UK study finds high-mileage electric cars more durable than petrol (electrek.co)
    —discuss
  17. AMD Takes the Lid off of Zen 6 as EPYC 9006 (servethehome.com)
    —discuss
  18. 2026 Small World in Motion Competition (nikonsmallworld.com)
    —discuss
  19. Accountable Algorithms (pennlawreview.com)
    —discuss
  20. Show HN: Senzii – open-source staff scheduling with a native MCP interface (github.com/senzii-app)
    —discuss
  21. The Most Valuable Engineer Isn't Shipping Features – Rails World 2026 Keynote [video] (youtube.com)
    1comments
  22. My time away from tech (2026) (jezhou.com)
    2comments
  23. Tell HN: New Chat Control vote tomorrow
    —discuss
  24. Dstv Installation Glenferness Tech Installations (techinstallations.co.za)
    1comments
  25. Show HN: IngotDB – SQL-based memory for LLM agents (github.com/tjbroodryk)
    1comments
  26. Israel Keeps Expanding into Gaza Despite Cease-Fire, Satellite Images Show (nytimes.com)
    —discuss
  27. "For my birthday, my brother-in-law gave me a 4992-digit hex number" (mathstodon.xyz/robinhouston)
    —discuss
  28. The Homeless Community That Became a Self-Governed Village (nytimes.com)
    1comments
  29. Nvidia Vera SMT Performance Benchmarks Review (phoronix.com)
    —discuss
  30. Generative AI Gives Spacecraft the Autonomy Engineers Once Feared (ieee.org)
    —discuss

Show HN: Jevpipe – a fast System 1 for AI agents, as a Unix pipe

2 pointsby 1h agogithub.com
2 comments
1h agoHN ↗

I built jevpipe as a wrapper for TypeSafe's Jev System 1 model.

Jev answers yes/no, picks from a few choices or gives a score, with probabilities, and it's fast and cheap. See https://docs.typesafe.ai/introduction for more details on Jev. It performs on par with small LLMs while being faster and cheaper (independent study: https://www.ayautomate.com/blog/jev-vs-llm-benchmark), and in my benchmark it was better calibrated than DeepSeek V4.1 Flash (see below).

I wanted agents to be able to use such a System 1 in the middle of a task without writing a script each time. So jevpipe is a small CLI that can be used like grep or other Unix tools: pipe lines or files in and get classifications out.

Example:

  cat abstracts.jsonl | jevpipe filter "Does this paper report a randomized trial?"
  cat abstracts.jsonl | jevpipe map -q '{"design": {"type": "choicsign?", "criteria": {"rct": "randomized trial", "observational":"observational study", "review": "review or meta-analysis", "other": "anything else"}}}' | jq -r .answers.design.choice | sort | uniq -c

filter prints the lines that pass, map returns typed answers as JS basically program their own System 1 for anything and apply it to a large chunk of data for "gut feeling" decisions.

I also added an agent skill for jevpipe that teaches the agent when and how to use it.

If you are interested in technical details, I also did a benchmark because that seemed to be missing from the recent evaluations. But it is by no means coding only or just a search tool. https://github.com/fabianboth/jevpipe/blob/main/bench/README...

It needs an OpenRouter key. I'm not affiliated with TypeSafe.

1h agoHN ↗

very nice. I finally made typesafe.ai account and put down $5 whole dollars.