The Brains — everything the assistant knows about you, and how it answers.
Ask about your body, your business or your family and the answer should be built from your own dated record rather than a guess.
As of September 10, 08:02 PM
What the status words mean
proven · done, not independently checked · partly done · not started · blocked · parked0%not started
24 steps: 0 proven · 0 partly done · 24 untouched · 0 blocked · 0 parked
agent runs · not recordedtokens used · not recordedreview notes · not recordeddollars charged · not recorded
The plan as first written
Nick asks about his body, his business or his family and gets an answer built from his own dated record, held in one of exactly two places he owns — the business Hub and the family app — better than the answer he gets today, safe on every hard flag, and provable. Every path that answers him, and every test, runs on the free door through the subscription accounts he already pays for. In his own words this morning: *"we need to be able to test without the paying so much."*
Steps
| Id | Plain words | Percent | State | Proof | What is left |
|---|---|---|---|---|---|
| 1 | Write down every feed and which of your two databases it lands in | 0% | not started | Python3 projects/business/single-brain/validate_business_spine.py, AND the register's own check built in this step run red on a register with one feed's destination deleted and green on the completed register — the validator alone never closes this step | The step |
| 2 | Prove every answer traces back to a record in one of the two | 0% | not started | Python3 projects/ops/life-os/audits/A1/retrieval_probe.py | The step |
| 3 | Get the outside task board out of the answers, and hand the app-side rewiring over | 0% | not started | Python3 projects/business/single-brain/test_hub_spine_source.py, AND the open-loops answer produced again from the Hub's own record with the old and new answers pasted side by side — a deleted capability never closes this step | The step |
| 4 | Fix the crash that wasted three runs, and build the test the other paths are graded by | 0% | not started | Python3 ../deck-shared/py/test_lane_cli.py | The step |
| 5 | Make the Mac answer you through the free door, not the card | 0% | not started | Node projects/personal/skippy-app/_test-cli-lane-node.mjs --target local (created by STEP 4) | The step |
| 6 | Make the cloud assistant answer you through the free door too | 0% | not started | Node projects/personal/skippy-app/_test-cli-lane-node.mjs --target https://skippy-cloud.fly.dev (created by STEP 4) | The step |
| 7 | Take away the cloud's standing permission to reach for the card | 0% | not started | Node projects/personal/skippy-app/skippy-code/_test-cloud-responsiveness.mjs | The step |
| 8 | Run all twenty-three health questions for nothing, through one door | 0% | not started | Node projects/personal/health/engine/qa-battery/chat-battery.mjs --a11-run --surface local --ids H01-H15,S01-S08 --out projects/ops/life-os/audits/A11/LIVE/FREE-RUN-ALL-23 | The step |
| 9 | Say plainly whether the new health brain is worth releasing, and what would make answers better | 0% | not started | Python3 projects/personal/health/engine/qa-battery/a11_compare.py | Waiting on Nick: release, redesign, or park |
| 10 | The health engine's deeper work, run under its own plan and reported back here | 0% | not started | Node projects/ops/life-os/audits/A11/LIVE/route-contract.mjs | The step |
| 11 | Run the sealed fifteen questions blind, then hand the result to someone independent | 0% | not started | Node projects/personal/health/engine/qa-battery/chat-battery.mjs --a11-run --surface local --ids U01-U15 --case-file projects/ops/life-os/audits/A11/LIVE/unseen-cases.json --out projects/ops/life-os/audits/A11/LIVE/UNSEEN-RUN-15 | Waiting on Nick: final acceptance of the health engine is his — the spoken and on-device proof questions are run by the agent itself under his standing grant of 2026-09-04 |
| 12 | Prove a saved record can be pulled back out, before anything real goes in | 0% | not started | Python3 projects/ops/life-os/audits/BUSINESS-INTAKE/real_batch_rehearsal.py, with the store snapshotted and read back before the first withdrawal and the snapshot's path named in the step's record | The step |
| 13 | Land Rizza's finance material and DinDin's recruiting material on your approve lines, and read every record back | 0% | not started | Python3 projects/ops/life-os/audits/BUSINESS-INTAKE/contract_checks.py, with the saved identifier one of rizza-2026-09-08-knowledge or dindin-2026-09-08-knowledge and its report digest re-measured at save time | Waiting on Nick: the approve line for each batch, or a decline |
| 14 | Take in how the team actually works — the calls they make, the histories, the open loops | 0% | not started | Node projects/ops/life-os/audits/BUSINESS-INTAKE/hub_rehearsal.mjs | The step |
| 15 | Check the new material against what the brain already believes, and surface every disagreement | 0% | not started | Python3 projects/business/single-brain/validate_finance_spine.py | The step |
| 16 | Say where future team dumps come from, whose mailbox, and what period | 0% | not started | Python3 projects/ops/life-os/audits/BUSINESS-INTAKE/rehearsal.py | Waiting on Nick: the source, the mailbox owner and the period per batch |
| 17 | Grade thirty real conversations, not replays | 0% | not started | Node projects/personal/skippy-app/skippy-code/_test-conversation-ledger.mjs, AND the count of corroborated real conversations read out of the live ledger with the replay count printed beside it | The step |
| 18 | Get the Mac relay running again so its calls can be counted | 0% | not started | Node projects/personal/skippy-app/skippy-code/_test-relay-port-scope.mjs, AND the relay's count rising by exactly one across a real call, read before and after | The step |
| 19 | Complete the record of which assistant answered what, on every channel | 0% | not started | Node projects/personal/skippy-app/skippy-code/_test-threads-conversation-roundtrip.mjs, AND every channel carrying a receipt re-read off the live server or a named gap | The step |
| 20 | Say what a question actually costs, piece by piece, across every connection type | 0% | not started | Node projects/personal/skippy-app/skippy-code/_smp-1-conversation-harness.mjs, AND one real context package's per-component counts re-summed to its stated total by the checker | The step |
| 21 | Apply the seven wording fixes on your taps and switch on the weekly check | 0% | not started | Python3 projects/ops/agents/check_plan.py projects/ops/life-os/audits/A3/PLAN.md | Waiting on Nick: the approval taps for the seven wording patches — checking each patch still applies to current source runs now, without a tap |
| 22 | Re-run the memory acceptance over all forty questions, not the eight it closed on | 0% | not started | Python3 projects/ops/life-os/audits/A1/bounded_acceptance.py | The step |
| 23 | Prove the email watch runs on its own or drop the claim, and switch on the weekly re-check | 0% | not started | Python3 projects/ops/life-os/audits/A1/weekly-guard-test.py | Waiting on Nick: one approval tap switches the weekly re-check on |
| 24 | Prove a health number can never come out of prose | 0% | not started | Python3 projects/ops/life-os/audits/A1/retrieval_probe.py, with the filter's shasum -a 256 digest after the step matching the one recorded before the deliberate mutation | The step |
Done
0 of 24 steps proven
Left
24 of 24 steps not yet proven
Blocked on
- Nothing currently blocked.
Formal remaining plan
| # | Step | Needs Nick |
|---|---|---|
| 1 | Write down every feed and which of your two databases it lands in | No |
| 2 | Prove every answer traces back to a record in one of the two | No |
| 3 | Get the outside task board out of the answers, and hand the app-side rewiring over | No |
| 4 | Fix the crash that wasted three runs, and build the test the other paths are graded by | No |
| 5 | Make the Mac answer you through the free door, not the card | No |
| 6 | Make the cloud assistant answer you through the free door too | No |
| 7 | Take away the cloud's standing permission to reach for the card | No |
| 8 | Run all twenty-three health questions for nothing, through one door | No |
| 9 | Say plainly whether the new health brain is worth releasing, and what would make answers better | Yes |
| 10 | The health engine's deeper work, run under its own plan and reported back here | No |
| 11 | Run the sealed fifteen questions blind, then hand the result to someone independent | Yes |
| 12 | Prove a saved record can be pulled back out, before anything real goes in | No |
| 13 | Land Rizza's finance material and DinDin's recruiting material on your approve lines, and read every record back | Yes |
| 14 | Take in how the team actually works — the calls they make, the histories, the open loops | No |
| 15 | Check the new material against what the brain already believes, and surface every disagreement | No |
| 16 | Say where future team dumps come from, whose mailbox, and what period | Yes |
| 17 | Grade thirty real conversations, not replays | No |
| 18 | Get the Mac relay running again so its calls can be counted | No |
| 19 | Complete the record of which assistant answered what, on every channel | No |
| 20 | Say what a question actually costs, piece by piece, across every connection type | No |
| 21 | Apply the seven wording fixes on your taps and switch on the weekly check | Yes |
| 22 | Re-run the memory acceptance over all forty questions, not the eight it closed on | No |
| 23 | Prove the email watch runs on its own or drop the claim, and switch on the weekly re-check | Yes |
| 24 | Prove a health number can never come out of prose | No |
Decisions this lane is waiting on
- None.