frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Open in hackernews

Show HN: 32KB deductive engine that catches LLM hallucinations

1•zhangxiaowen•1h ago
I built a hallucination detector that takes any AI-generated text, extracts verifiable factual claims, cross-checks each against search results, and outputs a credibility report with per-claim verdicts.

How it works: (1) Paste any AI response. (2) Extractor identifies factual claims — names, dates, numbers, citations. (3) Each claim gets searched independently via HTTP. (4) Comparator checks search evidence against claims. (5) Reporter scores overall credibility.

7 Python modules, ~27KB total. Uses Claude API for extraction/comparison and direct search for verification. Streamlit web UI with color-coded cards per claim.

The thesis: Hallucination is an architecture problem, not a scale problem. LLMs compute argmax P(most_likely), not P(true). More parameters make the guess more refined, but "most likely" ≠ "most true." So instead of making the guesser better, add an independent verification layer that runs on logic, not statistics.

The meta-irony: During code review, I had Claude write the code and Gemini review it. Gemini flagged claude-sonnet-4-20250514 as a "fictional model" and issued a critical blocking warning. The model is real — Gemini's training cutoff made it hallucinate about a model name while reviewing a hallucination detector. Then Claude summarized "all three AIs approved" when only two existed. Human caught both with one sentence each.

Built on a 32KB deductive reasoning engine (9 axioms, fractal-verified across 6 relationship scales). Also open source.

Detector: https://github.com/ZhangXiaowenOpen/hallucination-detector

All projects: https://github.com/ZhangXiaowenOpen

MIT + Heart Clause license. Solo dev + AI collaboration. Happy to answer questions about the architecture or why deductive verification will outlast RAG-based approaches.

Show HN: A luma dependent chroma compression algorithm (image compression)

https://www.bitsnbites.eu/a-spatial-domain-variable-block-size-luma-dependent-chroma-compression-...
1•mbitsnbites•1m ago•0 comments

No Other Choice

https://en.wikipedia.org/wiki/No_Other_Choice
2•tosh•1m ago•0 comments

AMD hints the next-gen Xbox console could launch next year

https://www.videogameschronicle.com/news/amd-hints-the-next-gen-xbox-console-could-launch-next-year/
1•smurda•1m ago•0 comments

Ax for Browser Automation Platforms: Browserless vs. Browserbase vs. Anchor

https://techstackups.com/comparisons/browserless-vs-browserbase-vs-anchor-agent-experience/AgentE...
1•sixhobbits•4m ago•0 comments

BKND Joins Supabase

https://supabase.com/blog/bknd-joins-supabase
1•ferhatelmas•4m ago•0 comments

Personal Information Firehose

https://adamwiggins.com/posts/personal-information-firehose/
1•tosh•5m ago•0 comments

Show HN: Agentic website to automate presales and more, by hotlines.ai

https://hotlines.ai/
1•avaid1996•6m ago•0 comments

Show HN: Once – An app that only works once per day

https://apps.apple.com/ro/app/once-today/id6758272971
1•andysteaua•7m ago•0 comments

I Am Building an AI-Powered Reverse Incubator

https://benjaminsen.substack.com/p/i-am-building-an-ai-powered-reverse
1•johlo•7m ago•0 comments

The dueling 'free grocery' stunts from Polymarket and Kalshi in NYC

https://www.businessinsider.com/polymarket-kalshi-free-grocery-store-marketing-stunt-nyc-2026-2
1•cft•12m ago•0 comments

Show HN: Tokenaru – commodity market for LLM tokens

https://tokenaru.com
1•bgleb•12m ago•0 comments

Are We at the End of the Industrial Age?

https://www.nytimes.com/2026/02/04/opinion/ai-jobs-employment-industry.html
2•thm•14m ago•0 comments

Confide: Encrypted, ephemeral and screenshot-proof messenger

https://getconfide.com/
1•rzk•17m ago•1 comments

Claude Code Plugins

https://github.com/ComposioHQ/awesome-claude-plugins
1•sunilkumardash9•18m ago•0 comments

An Open Letter to Jony Ives AI Companion

1•daly•21m ago•0 comments

JavaScript Bin Down in 2026

https://remysharp.com/2026/02/02/js-bin-down-in-2026
1•robin_reala•21m ago•0 comments

Glass Battery

https://en.wikipedia.org/wiki/Glass_battery
2•RGamma•24m ago•0 comments

Grounded Agency: The Type System Your Agent Framework Forgot to Build

https://github.com/synaptiai/agent-capability-standard
1•fornbogi•25m ago•1 comments

Long-term memory for OpenClaw agents with the mem0/OpenClaw-mem0 plugin

https://docs.mem0.ai/integrations/openclaw
1•ninadwrites•25m ago•0 comments

Show HN: Swiss army knife for SpiderWeb Router

https://github.com/knitprong/Devilfileprong-/commit/549364cb64afc348cfd60b18b95af71096a5cd12
1•devilfileprong•28m ago•0 comments

Stardew Valley Turns 10: The Big ConcernedApe Interview

https://www.ign.com/articles/stardew-valley-turns-10-the-big-concernedape-interview
1•thm•31m ago•0 comments

Trump's Profiteering Hits $4B

https://www.newyorker.com/news/a-reporter-at-large/trumps-profiteering-hits-four-billion-dollars
7•tromp•31m ago•0 comments

Skill Issues: An OpenClaw Malware Campaign

https://cantpwn.com/posts/skill-issues
1•djood•32m ago•0 comments

What Do You Get When You Put a Mummy Through a CT Scan?

https://www.nytimes.com/2026/02/03/health/mummy-virtual-autopsy.html
1•mitchbob•32m ago•1 comments

Ask HN: Why not just running OpenClaw in Docker?

1•fdeage•32m ago•2 comments

Show HN: AI Blocker by Kiddokraft

https://kiddokraft.org/wiki?name=ai-blocker
1•Rezhe•32m ago•0 comments

We built what Canva AI should have been

https://markup.one
2•cyrus_kelly•34m ago•1 comments

Proposal to Illion-Ise the Byte System

https://billibyte.site/
1•permo-w•34m ago•0 comments

TfL Status Page

https://tfl.luischav.es/
3•lucharo•34m ago•2 comments

Did we just see a black hole explode? Physicists think so

https://phys.org/news/2026-02-black-hole-physicists.html
1•pseudolus•36m ago•1 comments