How the builds fit together
How the builds connect, and how I use them day to day.
None of these builds started as a plan. Each one started with a task I was tired of doing by hand, and over time they started working together.
Keeping track of many sessions
On a normal day I have six to twelve Claude Code sessions open, one per customer environment or project. Each one keeps its own context, so I never re-explain a project from scratch. How I keep that many sessions straight is its own build, and it drops Nov 3.
What the builds produce
The visible output is mostly things a sales engineer hands to someone else: demo videos from AutoLoom, animated workflow slides from GIF Studio, call clips for the account team, and a weekly digest of what is new on our demo environment.
Correcting the AI
This is the part I would keep if I had to drop everything else. When the AI gets something wrong, I say so, and it saves the correction as a short rule with the reason and the incident behind it. There are 169 of those now. On our team's shared toolkit, the same idea runs as a ledger: lessons that keep recurring get promoted from written rules to automated checks, because written rules get ignored. See the memory loop.
The same applies to me. A coaching loop reviews my calls and counts my habits by name, so I can see whether I am actually changing.
The bots
The rest is bots: a morning briefing, a daily note on my writing, time entries filed from my calendar. Two of them I built, ran, and then switched off when the data behind them was not good enough to broadcast.
If you're a seller starting out
Pick the task you dislike most and start there. Save every correction you give the AI; most of what works for me came from those.