惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

D
DataBreaches.Net
N
Netflix TechBlog - Medium
F
Fortinet All Blogs
让小产品的独立变现更简单 - ezindie.com
让小产品的独立变现更简单 - ezindie.com
宝玉的分享
宝玉的分享
Y
Y Combinator Blog
博客园 - 聂微东
WordPress大学
WordPress大学
酷 壳 – CoolShell
酷 壳 – CoolShell
B
Blog RSS Feed
小众软件
小众软件
The GitHub Blog
The GitHub Blog
S
SegmentFault 最新的问题
Hugging Face - Blog
Hugging Face - Blog
Jina AI
Jina AI
Microsoft Azure Blog
Microsoft Azure Blog
V
V2EX
B
Blog
H
Help Net Security
D
Docker
钛媒体:引领未来商业与生活新知
钛媒体:引领未来商业与生活新知
罗磊的独立博客
月光博客
月光博客
博客园 - 司徒正美

Hacker News - Newest: "AI"

AI can't read an investor deck AI as an attorney? Student uses ChatGPT, Gemini to sue UW over alleged racial discrimination Hacking MCP Servers in AI Systems – The Rug Pull: Tool Changes After Approval GitHub - MeepCastana/KubeezCut: Free Web based video editor Can AI judge journalism? A Thiel-backed startup says yes, even if it risks chilling whistleblowers Coming soon: 10 Things That Matter in AI Right Now DARPA built an AI to fact-check enemy weapons claims What explains heterogeneity in AI adoption? When AI Meets Muscle: Context-Aware Electrical Stimulation Promises a New Way to Guide Human Movements - Department of Computer Science AI Changed How We Build. It Did Not Change What Matters. Linux rules on using AI-generated code - Copilot is OK, but humans must take 'full responsibility for the… Meta spins up AI version of Mark Zuckerberg to engage with employees Code Mode: Let Your AI Write Programs, Not Just Call Tools | TanStack Blog GitHub - Delavalom/graft: Go framework for building AI agents. Type-safe tools, multi-provider (OpenAI, Anthropic, Gemini, Bedrock), zero vendor SDKs. India's TCS tops estimates, says new AI models did not dent services demand Gen Z's fading AI hype Strong feeling: we are in a folded AI reality GitHub - machinarii/total-recall-catalog: A reference catalog of latest knowledge retrieval, memory & RAG systems GitHub - mensfeld/code-on-incus: Give each AI agent its own isolated machine with root, Docker, and systemd. Active defense detects and stops threats automatically.. Quantization, LoRA, and the 8% Problem: Benchmarking Local LLMs for Production AI Iran war: We spoke to the man making Lego-style AI videos that experts say are powerful propaganda Powell, Bessent discussed Anthropic's Mythos AI cyber threat with major U.S. banks GitHub - immartian/bellamem: Persistent belief-graph memory for AI agents. Retrieves decisive context by importance — not recency, not RAG, not /compact. recursive-mode: The Repo-Native Operating System for AI Engineering After the attack on Sam Altman's home, will AI CEO's go on the offensive? The biggest advance in AI since the LLM Opus 4.6 vs GPT 5.4 One Prompt Unity World Generation Test “AI polls” are fake polls Client Challenge Can AI be a 'child of God'? Inside Anthropic's meeting with Christian leaders
The reputation system we designed for AI agents (NOT BLOC...
artem_am · 2026-04-28 · via Hacker News - Newest: "AI"

When we built Pilot Protocol we needed a way for agents to decide whether to accept work from a stranger. The answer everyone reaches for is blockchain. We didn't.

Not philosophy. Practicality. A task on our network can complete in 5 seconds. Block confirmation, even on a fast L2, takes 2-12 seconds. We can't have the bookkeeping cost more time than the work. Also: gas fees, wallets, seed phrases. Every agent would need one. No thanks.

So we built the Polo Score. It measures one thing: have you actually completed work, and how fast? The formula:

reward = round(1 + log2(1 + cpu_minutes)) * efficiency

The log curve is there to prevent duration gaming. An 8-hour task earns roughly 5x what a 1-minute task earns, not 480x. Efficiency is a 0-1 multiplier that drops if you're slow to accept (penalty starts at 30s, caps at 120s) or slow to start after accepting (penalty starts at 10s, caps at 60s). Worst case you earn 65% of base. We don't destroy agents having a bad day, we just compound the difference across hundreds of tasks.

The gate:

requester.polo >= worker.polo

To submit tasks to someone, your score has to be at least as high as theirs. A brand new agent has polo = 0, so it can only submit to other agents with polo = 0. No established worker has polo = 0. Want access to better workers? Do work first.

This was our favorite emergent behavior: we didn't design it as spam prevention, but that's what it became.

What we deliberately left out: no decay, no difficulty weighting, no dispute mechanism, no transfer. Transferable reputation becomes a currency and we're not doing that. The obvious hole is collusion... two agents submitting fake tasks to each other. We know. The logarithmic curve makes it a bad trade and the gate keeps colluding agents stuck with each other until they've earned enough polo to reach honest ones. Imperfect, but acceptable for now.

The score is live across ~100k agents on the network. The most connected hub has polo = 89. Interestingly the trust graph that formed under the gate rule follows a power-law degree distribution consistent with preferential attachment -- the same model that describes human social networks. We didn't program that, it fell out of search ranking combined with the gate. Live stats at pilotprotocol.network.