970LABProject practice, collaboration methods, and lessons grown from real work.
First, I completed the 970LAB AI content workflow on a Mac mini. Later, I automated publishing for Guoma970 on a MacBook Pro. Two separate projects, and the next steps toward reusable instructions.
Read articleStarting with an existing model, then printing, checking the fit and feeding back what changed. Version two is in use; the six-part third version still needs a test print.
Read articleKeep work records outside the AI account: where you got to, why you made a decision, and what comes next. Another AI can continue the work, but you still check the results and make the decisions.
Read articleFrom a worksheet mismatch to independently drawn letterforms and guide-line calibration, with a free two-page alphabet tracing PDF.
Read articlePackaging voice, video, and content-operations workflows, testing them on another Mac, and clarifying the inputs. A process record, with no download in this update.
Read articleReal needs in our family, existing business and everyday work became products, websites, tools, workflows—and our own handwriting typeface.
Read articleI made Chinese and English videos with an authorised AI version of my voice, then asked the editor who knows my recordings to listen to the finished clips.
Read articleStarting with the renovation timeline, I am organising scattered experience into internal tools and sharing the process.
Read articleA familiar voice was only the start. The pauses, delivery and full edit still needed a human listen.
Read articleCheck the project, exact revision, delivery environment, recipient access, important workflow and handoff record before treating done as delivered.
Read articleAfter AI has produced a page, release still requires six evidence layers, stop gates, a human decision, and public readback.
Read articleBrands stay independent; development can run in parallel, while production release stays site-by-site and serial.
Read articleA page opening is only the start; content facts, Preview, production and public readback remain separate evidence layers.
Read articleThe goal is not to replace every role, but to leave a usable artifact at every handoff.
Read articleA real incident showed that safety boundaries must describe consequences, not only action names.
Read articleThe Xiguo Companion redesign starts with a child’s current task, language and interaction order.
Read articleOrganising models and tools around verifiable work instead of permanent job titles.
Read articleHelping people explain a problem should not require them to trade away privacy.
Read articleTrial, review and official release are different states.
Read article