Onkar Deokate’s ParkEase case study describes a Claude-assisted build of a car-wash partner app as a governed team workflow, not a one-shot code-generation exercise. The author reports 72 commits, 42 recorded decisions and six defects discovered in already-merged code. Those are project-specific figures, not independently audited results or a general measure of AI coding quality.
What the ParkEase task involved
Task 14 was a partner-facing car-wash app for ParkEase, a peer-to-peer parking marketplace in India. Its intended jobs were to receive wash offers, capture before-and-after photos, maintain a price menu and check earnings. Deokate describes the surrounding product as a NestJS API and worker, an Expo React Native app, an admin panel and a marketing site.
The author reports using Expo SDK 57, NestJS 11 on Fastify, Drizzle and PostgreSQL 18 with PostGIS. These are details of this 2026 case study, not recommendations or a claim that every component was independently checked.
How Claude was used as part of a gated workflow
Deokate describes /flow as a project-aware routing and governance process. It set phase gates rather than treating an agent’s output as approval. Coding was not allowed until design approval; the workflow then called for planning, test-driven implementation, review, a separate design audit, verification evidence and a pre-PR gate.
Free tools Windows power users keep installed
One-click scans. No signup required.
#1 Best Overall
Design first, then a plan
The design phase explored three tappable directions. The selected “Bay” direction made the before-and-after photo pair a persistent two-slot obligation, turning a product decision into a concrete interaction and data requirement. Planning was treated as a hypothesis to test, not a specification that could be trusted without review.
Separate review lenses
The reported review stack included security, silent failures, database behavior, TypeScript, React, test adequacy and specification conformance. Design auditing was separate from code review, and fixes received scoped re-reviews. The author says this process surfaced six defects already in merged code.
Make decisions and handoffs inspectable
The workflow used a decision ledger, artifact handoffs through files and saved context to resume interrupted work. A gate required evidence; an agent’s report alone did not satisfy it. This distinction matters because a confident completion message is not proof that a device behavior, integration or requirement has been verified.
Rank #2
Six defects the author says were already merged
Deokate reports that reviews and walkthroughs found these defects in merged code. This is the author’s account of one project, not an independent audit.
The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →- Proof-photo upload contract mismatch. The mobile app sent multipart form data while the API expected a proof-photo ID.
- Driver wash screen queried an unpopulated name field. The screen requested a user name that, according to the author, was not populated in the codebase.
- Business verification could not reach its pending state. The required state was unreachable through the implemented flow.
- Missing route index files. Washer and valet partners could land on an unmatched route.
- Idempotency operations ran fire-and-forget. Store and release operations could fail to complete as intended, leaving a key locked.
- An error code blurred two different outcomes. The client could not distinguish an unregistered partner from a server failure.
Why a plan and tests still needed scrutiny
The case study also describes mistakes that were not simply obvious syntax errors. An upload plan omitted an idempotency key. A later retry design reused a key and could replay a stale Cloudinary signature. An earnings query correlated a transaction identifier to itself, so it could sum postings too broadly. A two-job test exposed that earnings issue after two reviews had missed it.
Another test assumption about rupee-to-paise conversion conflicted with the stated minimum. The author also reports regressions caused by attempted fixes: memoizing an offer card carried an expired state onto a recycled FlashList cell, while stale-lock takeover risked running a committed-but-unstored request twice. Scoped re-review caught both before merge, according to the author.
Rank #3
These examples explain the value of testing both the initial implementation and the repair. A passing test suite is only useful to the extent that its cases exercise the relevant behavior; here, the two-job scenario mattered because it could reveal an over-broad earnings aggregation that prior reviews had not caught.
What the reported counts do—and do not—show
Deokate reports 72 commits, about 22,800 added lines across 185 files, 1,013 mobile tests, 397 integration tests against real PostgreSQL, 42 recorded decisions and six defects found in already-merged code. The author also reports a design review score of 23/40. These are counts and judgments from the author’s account; they are not independently audited, a controlled comparison, or a benchmark of Claude or AI-assisted development generally.
Recommended Free Tools
The numbers show the scale of the work as the author presents it, but they cannot by themselves establish productivity, correctness or product readiness. In particular, test totals say how many tests were reported, not whether every important real-world path was exercised.
Rank #4
What remained unverified
The author explicitly says the real camera, a real Cloudinary upload, TalkBack on Android and Maestro end-to-end flows were not verified. The reported automated tests and live walkthrough therefore should not be read as proof that those device, upload and accessibility checks passed, or that the app was fully verified.
For teams borrowing this approach, the practical distinction is between evidence from tests that did run and important checks that remain open. Deokate’s stated principle is: “An honest ‘I couldn’t check this’ beats a confident green checkmark.”
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Was this workflow faster?
Deokate does not claim that it was. Rate limits and quota changes stretched the work across two days. The case study’s argument is about inspectability: recording decisions, requiring evidence at gates and keeping gaps visible helped the team see what had been decided and what still needed checking. It does not establish that this workflow is faster than another process.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
The author’s broader lesson is that a plan is a hypothesis and the review stack tests it. For a small product team, that means treating AI output as work to validate, separating different kinds of review, re-checking fixes for regressions and tracking unverified real-world behavior rather than equating generated code or a large test count with completion.
Read Onkar Deokate’s ParkEase case study on DEV Community.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
