The workshop is open

Launch your idea.

For founders, technical or not. One monthly plan.

The governed session

Watch it work. Watch it ask.

A replay of a real session: an agent documents a codebase, and stops the moment a decision belongs to you.

Replay · recorded 2026-07-03 · excalidraw dossier run

Done86% cache hit
01 Agent02 Governance03 Tournament04 Dossier

replay shown in full; motion off.

01 Agent

Read package.json

Read README.md

Grep "exportToBlob" packages/

Read packages/excalidraw/components/App.tsx

Reading the code before writing about it. Every section will cite file and line.

Read packages/excalidraw/package.json

02 Governance

Permission requested

Bash wants to run.

Paused here until the operator chose. Allowed.

03 Tournament

Judging candidates

WinnerComposite 35/40Build pass · Tests pass

Judge rationale

dense, file:line-grounded, internally consistent.

04 Dossier

drafting sections…

Business summaryaccepted
Domains / modulesaccepted
Featuresaccepted
Data modelaccepted
Key decisionsaccepted
Glossaryaccepted
Open questionsaccepted

7 sections · 226s wall-clock · 86% cache reads

measured, not promised ↓

This is the machine you rent. Decisions stay with you.

Four stages. Sharp corners.

  1. 00

    What people ask for. What's already built. Where the MENA gaps sit.

  2. 01

    Fill the workshop. Upload a BRD. Turn an opportunity into a brief. Arabic, English, or both.

  3. 02

    Ship to the community. Content packs, cross-posts, investors looking.

  4. 03

    AI Tech Lead picks the stack. Splits the project into four milestones. Writes the tickets.

Measured, not promised

We benchmarked it before we told you about it.

In one night, this pipeline documented five real codebases, including dub and excalidraw. Independent judges scored the output blind.

RepoBlind score (/40)Wall-clockCache reads
glow35/40211s · 3.5×82%
full-stack-fastapi-template34/40120s · 5.1×83%
dub35/40301s · 4.9×85%
excalidraw35/40226s · 5.3×86%
swazona35/40286s · 4.1×85%

n=3 per cell. A defensible small-sample comparison, not statistical significance. Scored blind by independent judges on a 4×10 rubric: coverage, accuracy, proportionality, actionability.

Where this sits

A workshop, not a wrapper.

DimensionPrototype buildersAgent wrappersSwazona
Plan before codeprompt-firstprompt-firstBRD, milestones, tickets
Output judgedyou eyeball ityou eyeball itblind rubric, ranked
Spend and permissionsunmeteredper keycaps, approvals, audit
Runs ontheir cloudtheir cloudyour machine, your agent
Arabicafterthoughtafterthoughtfirst-class sibling
PLAN

One plan. Credits, metered to real cost.

Subscribe monthly for a credit balance. Every AI action debits what it actually costs, and new accounts start with free credits.

JUMPESC TO CLOSE
NAVIGATE
ACTIONS