What Running Pie Looks Like at Tilt
Five apps. Twice a week. End-to-end flows up to 149 steps.
Today, Pie owns the regression gate for every Tilt and Cashalo build that goes out the door. iOS and Android. End-to-end flows that traverse KYC, OTP, Plaid, loan disbursement, repayment, and ledger verification, on every meaningful release candidate. Anil's team reviews the deviations. Everything else, Pie has already cleared.
Current Run Cadence
Cashalo iOS
~5-6 times per week
Tilt Android
~5 times per week
Tilt iOS
~4 times per week
Pie has executed 507 test runs across the portfolio since September 2025. In the last 60 days, that's ~310 runs, roughly 670 hours of manual regression labor. The equivalent of two full-time QA engineers working nothing but regression, for two months straight. That capacity is back in Anil's team, freed up for exploratory testing, root-cause work, and the QA judgment calls that automation isn't meant to make.
How Pie Reaches the Deepest Flows: Backend Calls Inside Any Test
Most automation platforms are scripts. Every tap, every selector, every assertion is hand-coded. When the UI shifts, the script breaks.
Pie inverts that. Vision handles the front end. Pie's agent reads each screen, decides the next move, and adapts when the layout changes. Backend hooks handle only what vision can't see: production state that needs to be primed before the flow can run. Spinning up a test user with a specific credit history. Validating a ledger update mid-test. Resetting account state between runs.
That separation is what makes Tilt's 149-step journey possible. Pie navigates the UI on its own, through KYC, OTP, Plaid, disbursement, and repayment, and reaches into the backend only when state needs setting. No selector maintenance. No script rebuilds when a screen shifts. No ceiling on how deep a test can go.
