From consumer to producer.

Knowing how to use AI is not the achievement. Making something people pay for is. We translate the methods worth knowing, get the tools working in your hands, and build our own product the same way — reporting back what breaks.

Open the Wiki See the tool service →

What we do

01KNOW

Curated technique

We read the foreign material first, translate it into Chinese, and credit the original author with a link. You get a method you can use today, not another feed to filter.

Open the Wiki →

02ACCESS

Tool service

Knowing the method is not the same as being able to run it. We open the subscription through official channels and hand you a redemption link — no payment trace on your account, and someone accountable when it breaks.

See the service →

03PROVE

Built in the open

Transcribe is a product we built with these same methods, from nothing to paying customers. What went wrong and how we got around it goes straight back into the Wiki — first-hand, not repackaged.

Visit Transcribe →

Latest

View all →

Have the model flag contradictions before it summarisesA contradiction pass before synthesis catches most plausible-sounding wrong conclusions.

ClaudeResearch

2026/07/18Every

Turn 20 PDFs into a citable, traceable corpusEvery claim links back to a page number — built for cross-checked work.

CodexEnvironmentWorkflows

2026/07/02Anthropic

Make the model separate fact, inference and guessA three-way labelling pass that stops inference being written as fact.

Prompts

2026/06/11HiBridge

Leave your email, walk straight in

The first time an address shows up we create the account. No password, no approval. Signing in only lets you see your orders and redemption links — the Wiki never asks.