Hacker News

Top stories

Live mirror
30 storiesupdated just nowView source snapshot
  1. Claude Code now reads AGENTS.md if there is no Claude.md(claude.com ↗)
    149comments
  2. Android 17 is the first since 3.x to add new APIs without releasing to the AOSP(grapheneos.social ↗)
    203comments
  3. How OpenAI Used Its Own LLMs to Design Its Jalapeño Chip(ieee.org ↗)
    11comments
  4. Saving another 100TB of RAM(cloudflare.com ↗)
    36comments
  5. Cloudflare Quick Tunnels(cloudflare.com ↗)
    236comments
  6. Xcode 27.1 Beta Release Notes(developer.apple.com ↗)
    63comments
  7. Cache-to-Cache: Direct Semantic Communication Between LLMs (2025)(arxiv.org ↗)
    11comments
  8. Photon-Emission-Guided Laser Fault Injection Enables RP2350 Secure Debug(ledger.com ↗)
    46comments
  9. How to Write with an LLM(sockpuppet.org ↗)
    254comments
  10. Show HN: Cactus Needle 3: 8-29MB automation models can match DeepSeek V4 Flash(cactuscompute.com ↗)
    72comments
  11. The first new cat species discovered in 100 years(nationalgeographic.com ↗)
    41comments
  12. A 1542 papal cipher cracked with simulated annealing(simonklee.dk ↗)
    discuss
  13. OpenJev(openjev.com ↗)
    239comments
  14. Cyclomatic Complexity in C#(ndepend.com ↗)
    12comments
  15. Two parallel neural ectoderm progenitors contribute to the developing brain(newscientist.com ↗)
    51comments
  16. The Implications of Linguistic Illegibility for LLM Security(arxiv.org ↗)
    17comments
  17. US troop deaths during Iran war exceed Pentagon count by at least four(reuters.com ↗)
    58comments
  18. C++26: Trivial infinite loops are no longer undefined behaviour(sandordargo.com ↗)
    169comments
  19. Size-Specialized Memory Allocation(go.dev ↗)
    3comments
  20. Warez: The Infrastructure and Aesthetics of Piracy (2021)(archive.org ↗)
    13comments
  21. Minimal Phone 2(minimalcompany.com ↗)
    158comments
  22. Inside ZCode: Silently uploading your Git history to the cloud(ferstar.org ↗)
    89comments
  23. How SpaceX streamlined the Raptor engine(construction-physics.com ↗)
    28comments
  24. A search-and-inference database from scratch in pure Zig(antfly.io ↗)
    16comments
  25. I vibed a proof of Conway's conjecture(overreacted.io ↗)
    179comments
  26. From Geometry to Algebra and Back Again: 4000 Years of Papers (2023) [video](youtube.com ↗)
    discuss
  27. Korea raises data breach fines to 10% of revenue(koreajoongangdaily.com ↗)
    77comments
  28. North Korean nuclear test sets off years of earthquakes(science.org ↗)
    149comments
  29. Mathematicians Build Long-Awaited Graph Sandwich(quantamagazine.org ↗)
    18comments
  30. Cekura (YC F24) Is Hiring(ycombinator.com ↗)
    discuss

More heat than light: The HN “Flame War” detector

5 pointsby 5y ago
1 comments
Not everyone knows about this, so I thought I'd write a short note[0].

HN has a "Flame War" detector. When it detects a flame war it applies a penalty to the item in question, causing it to drop in the rankings.

The definition of "Flame War" is an interesting one:

* when an item gets to more than 40 comments,

* ... if there are more comments than votes,

* ... ... then it's likely to be a Flame War.

The idea is that HN wants things to be "Intellectually Interesting". TPTB seem to think that if something is interesting then it will get upvoted, and that seems reasonable.

OTOH, if there is a flame war then it's likely to be the same small group of people arguing back and forth. So the number of comments rises, but they can't repeatedly upvote, so the number of votes peaks early. Further, if it genuinely is a flame war then it's unlikely to get upvotes "from outside", and so there we have the conditions.

Occasionally it gets a false positive, so I saw somewhere that whenever the detector gets triggered, the item in question is reviewed by the mods. (Let me add here that the mods make decisions, and while some people will sometimes disagree, I think on the whole they do an amazing job. To them: Thank you.)

I suspect that this is a case in point: http://hnrankings.info/25870571/

I think it's a reasonable idea, the proxy of "more comments than points" seems defensible, and having the backstop of the mods checking things is reassuring. It's always disappointing when what I think is a reasonable discussion suddenly sinks without trace, but I believe HN is the better for it. It would be interesting to know how often it's triggered, and how many are then reversed.

[0] I'll try to add references when I can.

5y agoHN ↗

For those interested in more similar things, there is an unofficial FAQ by minimaxir: https://github.com/minimaxir/hacker-news-undocumented#flame-...

Also, for future readers, the numbers and exact criteria may change without warning. Perhaps in five years the threshold will be 50 (or 30) instead of 40, perhaps 37+uvpotes/7 or whatever criteria the mods can imagine. IIRC it has no changed in a long time, so perhaps it will still be 40.