White Wolf Studio · Johannesburg

AI agents build it.A human signs it off.

Claude Code and Codex plan, build, test and review every change. Then Dhashin reads every line and decides what ships. It's how we run four products of our own in production, and how we'll build yours.

In production: Polisync · Wonderkind · EziPzi · Houndbook

How every change ships

Five steps. Four are agents. The last one is a person.

The agents move as fast as they can. The final step waits for a human, every time. Here's an illustrative run. You get to be the human.

#142

Add CSV export to the reports page

Illustrative runAgents working
  • plan › 5 steps · 3 files · 2 risks noted

Agent does

Reads the ticket and the codebase, then writes a step-by-step plan: files to touch, tests to add, risks to watch.

Human checks

Is this the right problem, and the right size? The plan is approved before any code is written.

Agents are working through the plan. Tap a step to see who does what.

Who does what

Agents do the groundwork.
Judgement stays human.

The agents

Claude Code · Codex

  • 01Turn a ticket into a written plan
  • 02Implement it on a branch
  • 03Write and run the tests
  • 04Run types, lint and CI until green
  • 05Give every diff a first-pass review

The human

Dhashin Reddy

  • 01Decide what is worth building
  • 02Approve the plan before code
  • 03Make the architecture calls
  • 04Read every diff, line by line
  • 05Approve what ships. Every time.

What the loop runs on

  • Claude Code
  • Codex
  • Laravel
  • React
  • TypeScript
  • Inertia.js
  • Tailwind CSS
  • PostgreSQL
  • PHPUnit
  • GitHub Actions
  • Railway

What we do

Brand, product and launch. One loop.

End-to-end work for founders and teams who want agent speed without giving up a human's eye for quality.

Brand & Identity

Naming, visual language and typography. Taste is the part we never hand to a model: every brand call is made by a person.

Web Apps & Sites

Marketing sites and product UIs in React, TypeScript, Laravel and Inertia. Built by agents, tested by agents, read line by line by a human.

Launch & Systems

CI, deployment, analytics and handover docs, so the same plan-build-test-review loop keeps running after we hand it over.

The proof is on our own shelf

Four products of our own, in production. Every change goes through the loop.

Polisync home page beside an illustrative WhatsApp message about a missed premium, with Retry payment, Change debit date and Call me buttons

Polisync

Pilot

WhatsApp policyholder communications for South African insurers, UMAs and brokerages. A failed debit order becomes an approved WhatsApp message with Retry payment, Change debit date and Call me — and the reply goes straight back to the insurer's system.

SaaSWhatsApp Cloud APIInsurtechPOPIA
See all our work

"Discipline over motivation. Quality and speed. Build things that last."

— Studio philosophy

Have a project in mind?

Tell us what you're building. The agents start on day one. A human is accountable for every line.