The first 90 days
are when regressions actually hit.
After a Rescue Sprint ships, the work is not finished, it is exposed. The first 90 days in production are when memory drifts, costs creep, and an AI-built app starts breaking in ways nobody on the team can replay. Maintenance is the continuity layer: independent ongoing verification with three tiers, monthly billing, cancel anytime.
5 active client cap. Currently 0 of 5.
You get a written read on production health every month, delivered by email. The engine runs, the report is generated, you read it. No human in the loop, no calls, no fixes included.
- +Monthly Vibe Check engine run against your live build
- +Written report: memory, audit trail, cost control, autonomy boundaries
- +Severity-ranked findings with fix estimates
- +Email delivery on the 2nd of every month
- +Cancel anytime with 30 days notice
- −No fix PRs
- −No scheduled calls or live review
- −No emergency response
Everything in Verify, plus one small fix PR shipped to your repo every month. The engine ranks the candidate fixes, the top one gets a PR with a written summary. You review and merge. Never auto-merged.
- +Everything in Verify
- +One small fix PR per month from the ranked findings
- +Scope cap: under 2 hours of review time, no auth, RLS, billing, secrets, or data-migration touches
- +Written PR summary plus before and after evidence
- +Cancel anytime with 30 days notice
- −No emergency hotfixes
- −No scheduled calls (handled async in PR review)
- −Larger fixes get quoted separately as a Sprint
Weekly engine runs, priority response, and roughly 4 hours per month of prioritized fix work. For teams whose AI build is in production and they want a senior engineer watching it without standing up a full retainer.
- +Weekly Vibe Check engine run
- +Roughly 4 hours per month of prioritized fix work
- +One scheduled 30-minute call per month
- +Priority response to regressions surfaced by the engine
- +Hardening hour ledger, shared monthly
- +Cancel anytime with 30 days notice
- −Not a 24/7 on-call rotation
- −Hours do not roll over month to month
- −Anything beyond ~4 hours quoted separately as a Sprint
Why Maintenance, and why now.
Independent verification
Maintenance runs on a separate engine from whoever shipped the build. The point is not to defend the original work, it is to find the gaps the team will not see because they are too close to it.
Built for AI-shipped apps
The engine maps the four places AI builds fail: memory, audit trail, cost control, autonomy boundaries. If your app was shipped with AI assistance, these are the exposures that matter in the first 90 days.
No retainer trap
Month to month. Cancel anytime with 30 days notice. No annual commit, no minimum, no penalty for downgrading from Hardening to Verify when the build settles down.
Not sure which tier?
Start with a Triage.
Triage is free and async. Tell me what you shipped, what is breaking, and I will tell you whether Maintenance is the right fit or whether you need something else first.
Matt