Skip to content
View MM-sheng's full-sized avatar

Achievements

Achievement: Pull SharkAchievement: Pair Extraordinaire

Achievements

Achievement: Pull SharkAchievement: Pair Extraordinaire

Block or report MM-sheng

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
MM-sheng/README.md

Hi, I'm Sheng / 盛mm

I build AI tools and experiments that make model behavior easier to inspect and understand.

我做 AI 工具与实验，关注模型如何做决策、提示词如何改进，以及调用工具前如何检查风险。

Start here

JevSpeak — a decision model that speaks

Can a model that cannot generate text still hold a conversation? JevSpeak turns probabilistic decisions into English or Chinese through a deterministic language compiler. You can inspect the decisions behind each reply.

Try the app · Explore the code

404.directory — risk checks before Agent actions

An experimental service that gives AI Agents an evidence-backed allow, review, or block decision before they use a third-party tool. Includes a prediction-market preflight workflow.

Connect an Agent · Read the setup guide

Prompt Lens — inspect a prompt before sending it

A browser-based, rule-based prompt debugger. It checks structure, flags potential risks, and adds missing sections without sending your draft to a server.

Try it in your browser · Explore the code

Follow the experiments

X / @LuigiProof · Open an issue on JevSpeak

Questions, reproducible bugs, and examples of where these tools fall short are welcome.

Popular repositories Loading

  1. jevspeak jevspeak Public

    Jev can't generate text. So I made it talk anyway. A conversational interface built from probabilistic decisions and a deterministic language compiler — no generative LLM.

    TypeScript 2 1

  2. zeroleak-llm-proxy zeroleak-llm-proxy Public

    OpenAI-compatible privacy proxy that redacts secrets and PII before LLM API calls.

    JavaScript 1

  3. 404-directory 404-directory Public

    Experimental risk preflight for AI Agent actions: evidence-backed allow, review, or block decisions before third-party tool use.

    TypeScript 1

  4. manifesto manifesto Public

    The New Futurist Manifesto. For people who would rather invent the future than wait for it.

    1

  5. askback askback Public

    ai reflection webpage

  6. valueback2 valueback2 Public