Before
They came to the program to build a business, but kept losing track of their own structures. A whole database slipped from view. A status board they described as finished in four separate handoffs had sat empty since June 18.
“Not facts. Structures.”
In their own words
The turning point
On May 4, while designing a flagging system, they admitted they would not remember to go back and update it. That honesty changed the design: the flags moved out of project instructions and into Airtable, recorded as a numbered decision that stale flags are worse than no flags. From there, they set out to build something that would remember instead.
“So I built something that would remember instead.”
In their own words
After
Every session now opens with a running log and closes with three documents, all landing in a 20-table Airtable knowledge base. Four scheduled Claude agents run in the early morning, read the previous day and report what was done and what was left open. They also built a study hub for their daughter and a trip planner for a group voting on a September trip.
What it lets people do
The practice lets one person keep a clear, dated record of decisions, open threads and near-distractions without relying on memory. The agents check each other and the human checks them, so mistakes surface instead of hiding. Its ultimate audience is women who are not afraid of technology but are tired of feeling behind it.
What they learned
They learned to design around their real limits rather than pretend those limits away. Rules earn their place only through documented failures, and they are open about what is not finished yet: the pipeline in is built, the pipeline out is not.
“Nothing here was made for this submission. The files carry their original dates, including the ones where I got it wrong.”
In their own words
Walking the Talk was built during AI Mastery.
Read the full submission
I enrolled to build a business. The problem I actually had was that I kept forgetting what I had already built. Not facts. Structures. An entire database I lost track of. A status board I described as finished in four separate handoffs, sitting empty since June 18.
On May 4, designing a flagging system, I said: "I am not going to remember to go in and change it." The design changed because of it. The flags came out of the project instructions and into Airtable, recorded as a numbered decision: stale flags are worse than no flags.
So I built something that would remember instead.
**What it is.** A working practice, not a product. Every session opens with a running log and closes with three documents: what happened and why, where things route next, and what almost got chased but did not. Everything lands in a 20-table Airtable knowledge base. Four Claude scheduled agents run between 2:45 and 5:30 every morning, read yesterday, and tell me what I did and what I left open. They have run 41 consecutive days without a gap.
**How I built it.** Claude across Chat, Code, Design, Cowork and scheduled tasks. Airtable for memory. Perplexity for research. Drive, Gmail and Airtable joined by MCP. GitHub, Netlify and Vercel for deployment. No n8n.
**Who it is for.** Me first. Then my daughter, who uses a study hub I built her for a Master of Athletic Therapy. Then a group voting on a real trip in September. Eventually, women in their forties and fifties who are not afraid of technology but are tired of feeling behind it.
**What changed.** The agents police each other. Each may only touch records it owns. When two flagged the same file 51 minutes apart, that collision produced a cross-routine check now installed in all four. They audit my edits and I audit them.
I also turned my working method into a skill. Nine revisions in 25 days, and it will not accept a new rule without a documented failure attached.
**Honest limits.** The pipeline in is built. The pipeline out is not. Two further stages are specified and not running.
**How to read this.** Open 00-START-HERE-Walking-the-Talk.html first. It is a single self-contained page that opens offline, and it is the guide to everything else. It sets the four months out as a dated timeline of twenty entries, hangs nine exhibits off it, and for each exhibit gives the claim, the evidence, what to notice, and the exact file and line to look at. Thirteen screenshots are built into the page itself, so those are not separate downloads. Its last section lists every attachment and what each one is for. After that, Walking_the_Talk_Project_Narrative_2026-08-08.md carries the full argument. The remaining attachments can be opened during or after, for detail on any point you want to check for yourself.
**Links and cost.** Trip planner, live and public: sept2026-trip-planner.netlify.app. Personal dashboard: passphrase gated, a standalone copy is attached.
Nothing here was made for this submission. The files carry their original dates, including the ones where I got it wrong.
One last thing that I am adding at submission time. My plan was to have one large submission with most of my attachments contained. Attached you will find the Museum Portfolio (Standalone) HTML I had developed in Design initially for my Awards submission. Life and children don't stop for AI, therefore I was not able to complete my full submission. I am attaching it anyway as it is like most of my work, in progress.
Written 11 August 2026· Submitted by the storyteller
Every story is written by a Mindvalley student or by someone our teachers have worked with. How we collect and check stories
