New stories

Live mirror
30 storiesupdated just nowView source snapshot
  1. Contrails.org – Contrail Avoidance for the Climate(contrails.org ↗)
    discuss
  2. Landing the Space Shuttle – A Flying Machine and the Thrill of a Lifetime(eaa.org ↗)
    discuss
  3. Show HN: AgentReady – can AI assistants read your site?(lumnika.com ↗)
    discuss
  4. Show HN: Leo – a Markdown engineering process for AI coding agents(github.com/alex-zaporozhan ↗)
    discuss
  5. Mistaken alignment is not misalignment(c.mov ↗)
    1comments
  6. Vieta's Formulas(wikipedia.org ↗)
    discuss
  7. More Tree-sitter, more neocaml, more elisp(lambdafoo.com ↗)
    discuss
  8. How to keep a knowledge base up to date when the product changes every day(frigade.com ↗)
    discuss
  9. Emily St. John Mandel on Writing a Novel of Dystopian Counterlives(lithub.com ↗)
    discuss
  10. The proof passed. Can you see why it works?(github.com/8braid ↗)
    discuss
  11. Show HN: A chess analyzer that runs in the browser(github.com/huytd ↗)
    discuss
  12. A better way of blocking macOS updates(zoey-on-github.github.io ↗)
    discuss
  13. Ask HN: eBPF verifier packet tracking boundaries under volumetric socket drops?
    discuss
  14. Nvidia and CrowdStrike Develop New Cybersecurity AI Models(wsj.com ↗)
    discuss
  15. Steve Bannon and Bernie Sanders Condemn Tech 'Oligarchs' and Demand A.I. Reforms(nytimes.com ↗)
    discuss
  16. Founding documents from companies that shaped industries(relentless-investments.com ↗)
    discuss
  17. Waymo to start Tokyo robotaxi service next year in Asia push(japantimes.co.jp ↗)
    1comments
  18. Interview with Paul Bogen [video](youtube.com ↗)
    discuss
  19. Divide, Consult, Conquer: Capability Laundering Through Aligned LLMs(arxiv.org ↗)
    discuss
  20. Singapore launched a novel five-year national campaign to nurture reading habits(cnn.com ↗)
    discuss
  21. Show HN: Biom – A visual workspace where your AI agents' work lands(biom.dev ↗)
    discuss
  22. Hackaday Europe 2026: Bare Metal Made Easy(hackaday.com ↗)
    discuss
  23. Show HN: Plurnk (Yet *Another* AI Harness)(github.com/plurnk ↗)
    discuss
  24. Independent Lens – Ghost in the Machine: AI's troubled history, current impacts(pbs.org ↗)
    1comments
  25. Ask HN: Who has been contacted by 5+ hiring platforms?
    3comments
  26. Is There Anything Google's Fruit Fly Brain Can't Do?(nytimes.com ↗)
    discuss
  27. Callum Williams on Cybersecurity Prices(marginalrevolution.com ↗)
    discuss
  28. Probably the best laptop with Linux support and Local LLM AI [video](youtube.com ↗)
    discuss
  29. Atlas-Finance: Evaluating AI Agents Inside a Bank(joinhandshake.com ↗)
    1comments
  30. Disney's Films Used to Look Different(animationobsessive.substack.com ↗)
    discuss

Show HN: Plurnk (Yet *Another* AI Harness)

1 pointsby 56m agogithub.com
0 comments
I am desperate for feedback on my homebrewed harness, "Plurnk."

I've been dogfooding it on my RTX5070Ti with only 16GB VRAM as my daily driver for weeks now and it's genuinely better than everything else available (I'm biased).

1. Curation, not Compaction

The model's fully responsible (bitter lesson) for deterministically curating its own context.

2. ANTLR Grammar

The dozen verbs for the harness are optimized for model pretraining, looking like markdown while being fully integrated with an EBNF grammar that plugs into an AST for superior workflow and recovery.

3. Universal Resources

Everything's addressable by the model through a pseudo-URI interface, including the log entries.

4. Omnipatterns

Everything in the repo goes through treesitters which build a complete graph that the model can search with glob, regex, jsonpath, xpath, graph, and sqlite fulltext.

5. Standards

MCP2 tools, AG-UI client interface, A2A "agent to agent", plugs into everything through OAI spec + models.dev

6. Interfaces

CLI, TUI, and a Neovim plugin, easy to extend, very pluggable

7. Lean Agent

Service/Client architecture with lean TUI allows dozens of agents to run on weak hardware. 200MB footprint.

8. Lean Context

Achieves all of the "batteries includes" features of a fat harness while the sysprompt is only slightly larger than pi agent's.

It started as a proof of concept for my bespoke opinions on what "bitter lesson" actually means, and it's gone so well that I believe others trying to build agentic workflows, especially with local and humble constraints, could benefit from the proven concepts.

A quiet thread, for now.Start the conversation on HN ↗