Built a business, knowledge systems, and AI agents that run my life before I wake up

The category asks for range across tools, domains and complexity. Mine came from one workflow I did not design, applied to whatever was in front of me.

I had no experience with AI before March 2026. I had been avoiding even the Google search AI. I signed up for the Summit standing beside my bed, after midnight, on a whim after seeing a Facebook reel. I listened and took notes the whole Summit. My first conversation with Claude was March 23. The program started April 8. I enrolled to build a business, and what took over instead was my own life, starting with the fact that I could not hold what I had already built in my head. What helps a person with that, turns out to be what helps an agent with it too. That is where the workflow came from.

Domains, none of which were built to demonstrate range. A registered business with a documented brand voice and a 20-table knowledge base. A study hub for a Master of Athletic Therapy student, with a 750-hour practicum tracker and an AI assistant calling Claude through a server-side proxy. A Mediterranean trip planner covering four ports and two cities, with real users voting on real plans. Dark-themed astronomy guides, built dark because they are read at 2am in a field. An interactive backyard designer, built after commercial tools would not do what I needed. A 17-section photography field guide that works offline, because the lake has no signal. A research method, built against a deliberately difficult subject so it would have to handle the awkward cases.

Tools, added only when a problem demanded one. Claude across six surfaces. Airtable, seven bases. Perplexity, Drive, Gmail and Airtable joined by MCP. GitHub, Netlify, Vercel. NotebookLM, Obsidian, Suno, ChatGPT and Fable where each was better at something specific. No n8n.

Complexity. Four scheduled agents run before I wake up. They are governed, not merely scheduled. Each may only touch records it owns. One is forbidden from writing to what it assesses. One is told that a confident recommendation drawn from comparing only filenames and sizes is worse than no recommendation, because it would be actionable and wrong. Alerting has a four-layer fallback, because absence should never render as silence.

The part I would point at. Twice, my AI proposed research rules that sounded rigorous and would have quietly excluded the material the project existed to find. The second time I named it. It agreed: "That's twice now you've caught me setting up the same bias pattern, methodological rigor structured as exclusion." That correction is now a permanent rule in the research skill I use for every domain.

Honest ceiling. The pipeline in is built. The pipeline out is not but the plan is there and the work is constant. One dashboard surface is still hardcoded.

How to read this. Open 00-START-HERE-The-Ultimate-AI-Master.html first. It is a single self-contained page that opens offline, and it is the guide to everything else. It sets the work out as a map, fifteen points around one core, and for each point it gives the claim, the file that proves it, what to notice, and exactly where to look. Twenty-six 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, The_Ultimate_AI_Master_Project_Narrative_2026-08-08.md carries the full argument. The remaining files can be opened during or after, for detail on anything you want to check yourself.

Links and cost. sept2026-trip-planner.netlify.app, live and public. Personal dashboard passphrase gated, a standalone copy is attached. 9,327.20 CAD to March 31, plus roughly $1,000 in top-ups. The impact on my life has been priceless.

I did not set out to cover ground. I built one thing which snowballed to the next. The snowball has continued to roll.

Jen'sEdit: as I do with my AI. In the attachments is an HTML called Museum Portfolio (Standalone) which is what this submission should have been with most of those attachments built in. But life and children happen and must take priority. So unfortunately I am not able to submit the way I hoped but am submitting nonetheless.