← Engineering index

Genysys / Independent founder work

Can a training decision earn its cost?

My long-term aim is to lower the cost of expert care. I started with a narrower job. An athlete already has a plan and a wearable. Can one adjustment help them complete their next important session, at a cost a customer will pay?

Arciv prototype screens showing suggested and logged training choices in an illustrated episode.
Original Arciv product walkthrough. The two screens illustrate an episode. A completed on-device pilot episode is still to be demonstrated.

The product is Arciv. The first gate is one replayable decision.

The prototype keeps what the software knew on the day, what it recommended, what the athlete actually did, and what happened at the next planned session. That matters when a wearable later corrects a reading, or when an athlete does the finisher the software suggested skipping. Otherwise an attractive outcome can be credited to advice that nobody followed.

The founder-pilot software was merged in July 2026. The later work to reconstruct a whole episode from frozen evidence is an open build item. My first step is to run one real episode on my own device and replay it without later information. After that, I would ask one independent training team to test use, delivery cost and willingness to renew. There is no demonstrated paid renewal or clinical outcome in this record.

Implementation / checked November 2026

What has actually been built?

The engineering record and the product illustration answer different questions.

The merged founder-pilot code connects wearable observations to a plan, a recorded recommendation, the athlete’s action and a later outcome review. The Android client and backend retain the evidence available at the original decision.

174backend tests reported in the July merge
67/67release evaluations reported in that merge
Openhistorical episode-replay build item

Those checks concern code behavior. The public record still needs a real on-device episode, a subsequent comparison with usual training, and a buyer willing to continue.

Build record · July 25 founder-pilot merge, October 31 build contract and open issue. Readable extract from the private repository.  Read the source exhibit

Proposed pilot / design for discussion

One coach. Eight adults. Four weeks.

A proposed feasibility study after the founder replay passes. This scope has no announced enrollment, customer or funding commitment.

QuestionProposed comparisonDecision rule
Can we trust the record?Count every eligible decision from each athlete’s training calendar, including sessions with no entry. Report attempts, completions and missing episodes by athlete and period.Proposed gates: capture at least 90% of calendar-eligible decisions in each period, replay more than 95% of completed Arciv episodes, and allow zero use of data that arrived after the decision.
Can people use it?Each athlete uses their usual log for two weeks and Arciv for two weeks. Stagger order between participants.Aim for median deliberate entry under 30 seconds a day. Record coach review minutes and the count completing the next key session in each period.
Will a buyer pay?Agree a trial price and a renewal decision with the team before starting. Ask for a paid continuation at the end.Fund a larger comparison only after the team renews at a price that covers delivery and the use evidence warrants another test.

If only two of twenty eligible decisions are logged, coverage is 10%. Even two perfect replays would stop the pilot. Eight people over four weeks can expose missing data, high review burden or weak buyer interest. Training order and small numbers limit what a difference in completion rates could establish.

Business case / sensitivity

Five coach minutes use the entire fee.

Illustrative pricing for one four-week pilot cycle. The numbers are assumptions for judging a model, rather than quoted terms.

8 athletes × $20 = $160 of revenue per four weeksAt $60 per coach-hour, five minutes per athlete each week costs 8 × 4 × 5 ÷ 60 × $60 = $160. At ten minutes, coach time alone costs $320.

At that price the group needs fewer than five review minutes per athlete-week to leave room for software, support and margin. I would measure coach time before hiring for a larger rollout. The first funding would buy a verified founder replay and one bounded trial. Its actual budget depends on setup, coach cost and the agreed scope.

A clinic would need its own comparison against usual care, a named clinician who owns the decision, and proof that review burden and outcomes justify payment. That is a later hypothesis.