frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Open in hackernews

Self-Evolving Skill for Claude Code – v3 validation complete

https://github.com/191341025/Self-Evolving-Skill
1•tiansenxu•1h ago

Comments

tiansenxu•1h ago
A few weeks ago I shared a design pattern I've been building: a governance protocol that lets Claude Code Skills accumulate domain knowledge across sessions without bloating. The core idea is a Five-Gate protocol that controls what gets written into a living knowledge base — the most common outcome of the gates is "do nothing." 63.6% rejection rate in my first experiment. What's new since the last post: I've now completed three rounds of validation on the same database (smart building management, 29 tables). Each round tested progressively more sophisticated capabilities:

v1: Basic Five-Gate protocol — rejection rate, self-correction behavior v2: Confidence decay model — C(t) = C0 × e^(-λ × (β+1)/(α+1) × t) v3: Phase 5 enhancements — entities tagging, search-driven retrieval, hard/soft signal distinction

v3 passed 6/6 verification points. The highlight was T3.3: a user claimed incorrect enum values, and Gate 2 correctly rejected them because they contradicted SQL-verified data already in the knowledge base. The system defended its own knowledge integrity against incorrect human input. v3 also found a critical bug in the inject command — it silently wrote to the wrong path when given a relative --target. Fixed, patched, and verified. What the computation layer looks like now: The confidence math no longer lives in SKILL.md prompts (which caused LLM calculation errors). It's been moved to a Python tool layer: C(t) = C0 × e^( -λ_base × (β+1)/(α+1) × t ) 143 pytest cases passing. LLM judges, Python computes. What this is not: This isn't a finished product. The knowledge base for my test domain has 8 entries after 17 tasks — deliberately sparse. The design philosophy is that a mature Skill that stops growing is healthy, not stalled. Convergence is the goal. What I'm still iterating on: λ calibration across domains requires a second experiment (pending data ethics clearance on production data), the α/β upper bound question is open, and protocol compliance still depends on LLM discipline with no mechanical enforcement. Repo: github.com/191341025/Self-Evolving-Skill Still building. Feedback welcome.

chistev•1h ago
Why do the new green accounts post about Claude a lot?
tiansenxu•1h ago
Not promotion — I'm the author. This is an ongoing personal project I've been building and validating over the past few weeks. Happy to discuss the technical details.

Estranged and Alone? How to Find Community After Going 'No Contact' with Family

https://www.teenvogue.com/story/estranged-how-to-find-community-after-going-no-contact-with-family
1•Tomte•15s ago•0 comments

Roast your startup and send it to 1999

https://shipordie.club/roast/1999mystartup
1•ghoshbishakh•45s ago•0 comments

Are Prediction Markets Good for Journalism?

https://www.cjr.org/the-interview/are-prediction-markets-actually-good-for-journalism-kalshi-poly...
1•Tomte•59s ago•0 comments

AgentPages – GitHub Pages for AI Agents (GitHub Agentic Workflows – Gh-Aw)

https://github.com/idorozin/AgentPages
1•idorozin•1m ago•1 comments

Dog's Cancer Cure via ChatGPT and mRNA Vaccine Offers Hope for Humans

https://www.archyde.com/dogs-cancer-cure-via-chatgpt-mrna-vaccine-offers-hope-for-humans/
1•iamflimflam1•1m ago•0 comments

'Revolutionary': Ukrainian para-biathlete wins silver using ChatGPT as his coach

https://www.theguardian.com/sport/2026/mar/09/ukraine-winter-paralympics-chat-gpt-artificial-inte...
1•helsinkiandrew•4m ago•0 comments

I built a DNA computing kernel that processes 100k+ genes in parallel

https://github.com/Admin135158/Proteus
1•admin135158•5m ago•1 comments

Ask HN: How to Get a Internship?

1•krishSingaria•8m ago•0 comments

Cicikus v3 Prometheus 4.4B – An Experimental Franken-Merge for Edge Reasoning

https://huggingface.co/pthinc/Cicikus_PTHS_v3_4.4B
1•pthuser•10m ago•0 comments

Porting software has been trivial for a while now

https://ghuntley.com/porting/
1•ghuntley•11m ago•0 comments

$96 3D-printed rocket that recalculates its mid-air trajectory using a $5 sensor

https://github.com/novatic14/MANPADS-System-Launcher-and-Rocket
2•ZacnyLos•11m ago•0 comments

Atlassian promotes and lays off someone in the same day

https://old.reddit.com/r/theprimeagen/comments/1rsuj1v/atlassian_promotes_and_lays_off_someone_in...
2•mmarian•12m ago•0 comments

I made Karpathy's Autoresearch work on CPU

https://github.com/bopalvelut-prog/autoresearch
1•M4s4•13m ago•1 comments

Did giant Ice Age beasts carve these caves in South America?

https://www.nature.com/articles/d41586-025-00216-x
1•zeristor•13m ago•1 comments

The ~fifth~ fourth postulate of decision theory (On the Independence Axiom)

https://www.lesswrong.com/posts/MsjWPWjAerDtiQ3Do/on-the-independence-axiom
1•sieste•14m ago•0 comments

Gemini 3.1Pro is aggressive like a hungry wolf

https://old.reddit.com/r/GoogleAntigravityIDE/comments/1ru97bx/gemini_31pro_is_aggressive_like_a_...
2•cft•18m ago•0 comments

The women bringing chess into the 21st Century with bullet games & viral videos

https://www.bbc.co.uk/news/articles/ce3g0kel3jyo
2•mellosouls•20m ago•0 comments

Wow

1•yuvrajdahal•22m ago•0 comments

Show HN: Soorf, a natural language audio editor

https://soorf.xyz
1•copypirate•26m ago•0 comments

Pm2-manager: A modern, real-time web dashboard for your PM2 processes

https://github.com/orangecoding/pm2-manager
1•thunderbong•28m ago•0 comments

Boot, Prompt, Run: what happens to personal computing when sw writes itself

https://giampaolo.guiducci.it/posts/2026-03-15-boot-prompt-run/
2•gosub•29m ago•0 comments

Try an open-source MCP server for Postgres – win a Raspberry Pi

1•pgedge_postgres•31m ago•0 comments

Shield Messenger A Tor-native,P2P encrypted messenger with built-in Solana/Zcash

https://github.com/abokenan444/shield-messenger
1•abokenan444•39m ago•1 comments

AutoResearchClaw

https://github.com/aiming-lab/AutoResearchClaw
1•frozenseven•39m ago•0 comments

The Space Data Center Mass Budget Behind 10× Power Density

https://research.33fg.com/analysis/the-space-data-center-mass-budget-behind-10x-power-density
1•T-A•39m ago•0 comments

//go:fix inline and the source-level inliner

https://go.dev/blog/inliner
1•vismit2000•42m ago•0 comments

Claude Code tips for non-programmers

https://thewriting.dev/claude-code-isnt-just-for-developers/
1•r0rshrk•47m ago•0 comments

Three Claude Skills to Sharpen Judgment

https://age-of-product.com/three-ai-skills-to-sharpen-judgment/
1•swolpers•47m ago•0 comments

Hybrid AI Desktop Layer Combining DOM-Automation and API-Integrations

https://github.com/BiamOS/BiamOS
1•BiamOS•48m ago•1 comments

Jazzband Is Sunsetting

https://jazzband.co/
2•taubek•49m ago•0 comments