Meet Notes

· How to use this site

Where things stand

Compute is settled for now: Reid rents cloud machines to test, and Joe decides what to buy once those numbers are in. Reid is running the Ector County OCR and will send Joe a write-up today. Claude moves the meeting app onto Reid's cloud server and fixes the problems Reid's walk-through found.

Suggested Claude's overnight update: Reid ran his first real call on the new meeting page on Oct 26, and it showed the page is not yet easy for someone who did not build it: six problems, like a typed question vanishing and stray talk being counted as a yes. Joe and Reid also agreed that the meeting notes, Reid's intake app and the inventory view become one app, each part kept separate behind one front door. What matters next is getting the meeting app running on Reid's cloud server so anyone can use it, then fixing the walk-through problems.

Not confirmed yet. Claude checks it with you at the next call.

Status board

Projects · 3

Suggested All 3 are Claude's suggestions, not discussed yet. Each waits on the next call.

Topics · 19

AI server hardware
Converge , stage 3 of 4Waiting on ReidSuggestedOct 26
Reid sets up the rented cloud test; Joe picks the two setups and the rent-or-own point
Reading old county records: OCR plus a vision model
Understand , stage 1 of 4Waiting on ReidOct 26
Reid runs the OCR on the Ector County records and compares with Title Lab
Testing the new meeting page
Understand , stage 1 of 4Oct 22
Controls tried; reload test passed
What one AI agent run looks like, technically
Understand , stage 1 of 4Waiting on Joe and ReidOct 20
Answer the eight run questions
The interface any AI plugs into
Understand , stage 1 of 4Waiting on Joe and ReidOct 20
Decide one shared interface or per-agent tools
The experiment system and its arms
Understand , stage 1 of 4Waiting on ClaudeOct 20
Write the setup down once the arms are named
How runs are scored: the eval criteria
Understand , stage 1 of 4Waiting on JoeOct 20
Joe names the memo; Claude puts it on the page
One app for meetings, intake and our data, running in the cloud
Understand , stage 1 of 4Waiting on Joe and ReidOct 19
Confirm the site work list order at tomorrow's sync
Getting every deal's record to 100%
Understand , stage 1 of 4Oct 19
Seed with what we have; close the gaps later
Deal intake comes first: a complete inventory record for every closed deal
Converge , stage 3 of 4Waiting on Reid and JoeOct 19
Seed the inventory with the 48 deals already run through intake
County data plan
Explore , stage 2 of 4Waiting on Next callSuggestedOct 19
Name an owner to pick how to search the thumb drive's county records
A shareable map of what we own
Understand , stage 1 of 4Oct 19
Add-on to the framework, after intake
What happens after intake: the steps that close every loop
Understand , stage 1 of 4Waiting on IntakeOct 19
Wait until intake works, then test the steps after it
Where all of this is going
Understand , stage 1 of 4Oct 19
Possible vision, not confirmed
Reid's check: how often plain code matches documents to deals
Understand , stage 1 of 4Oct 19
Baseline set for check stubs: 81%
The answers we test against
Converge , stage 3 of 4SuggestedOct 16
27 documents proposed across 8 types, waiting for Reid. A few leases and the 2 tax documents need hand checks.
Measuring success
Converge , stage 3 of 4SuggestedOct 14
Two measures: how good each step is, and whether it runs end to end with little supervision. The key number: how often it was wrong and didn't ask. Still to write: what counts as a pass.
Phase 1 and Phase 2 questions
Converge , stage 3 of 4SuggestedOct 13
Phase 1 is settled enough. Phase 2 is answered for taking documents in and filling the inventory; the later steps wait for experiment 1.
Revenue and tax data
Explore , stage 2 of 4SuggestedOct 13
Reid loaded the historical revenue and tax data into our database. It still needs linking to deals and leases.
Decided · 4

The others wait on the next call.

Not filed yet 26

From Oct 26 · Joe, Reid

To-do
Send Joe a write-up after the call: what we discussed, what is done, what is left, and the plan for the repositories
To-do
Put the records into the sandbox database and define the actions from the transfer automation plan, using plain code first and an AI setup only where code falls short
Decision
The intake in the Command Center is the old first version and is retired; the one we build on is in the deal data project
To-do
Meet again tomorrow: wearables, and getting the environment set up so we can start processing records (tidy-up a week later)
To-do
Get Joe's answers to Reid's questions to Reid; they were given to Claude but never reached him
To-do
Finish the answer sheet for the intake app tests, so the setups on the Cletus rig are scored for accuracy, not only for agreeing with each other

From Oct 22 · Joe, Reid

To-do
Move the scanner work to next week so the title index gets attention first
Decision
Use the Excel inventory as the source for the title index

From Oct 22 · Joe, Reid

To-do
Move the scanner work to next week so the title index gets attention first
Decision
Use the Excel inventory as the source for the title index

From Oct 21 · Joe

Decision
This meeting (2026-10-21-qkv-fney-xag) is 30 minutes long.
File: Update: Reid, Oct 19 (night)
File: Update: Reid, Oct 19 (evening)
File: Update: Reid, Oct 19 (morning)

From Oct 16 · Joe, Reid, Keegan

File: Memo B: how we pick a winner for Cletus
File: Memo A: which AI-with-its-own-desk versions to test
File: Google Meet notes for the whole call (Gemini)

From Oct 15

File: AI server purchase: updated recommendation (Dell)
File: Research: four ways to build the co-worker
File: Compute cost and breakeven calculator
File: Inference Server Design Lab
File: Experiment 1: the proposed documents

From Oct 14

File: What a document record holds (draft for Reid)
File: Meet Notes UI/UX proposal

From Oct 13

File: Experiment 1 handoff (draft)
File: Transfer Automation, in plain English

Claude asks where these go at the next call.

Meetings

No vision yet. Claude will ask for one at the start of the call.