Launched this week

Tabstop

Launched this week

From client material to deployed site with GPT-6 Astra

3 followers

From client material to deployed site with GPT-6 Astra

3 followers

Tabstop is a bespoke web delivery system designed around shared human/agent boundaries. For this challenge, GPT-6 Astra works through machine-valid Define, Provision, Implement and Release gates to turn real client material into a deployed CMS-backed site.

Payment Required

Launch tags:Web App•Software Engineering•Developer Tools

Launch Team / Built With

ReplayAutonomous app testing for teams who ship at AI-speed

Promoted

OpenAI

GPT-6 Astra Challenge

How did Astra change the scope or ambition of what you built?

Maker

Astra changed the ambition from making Tabstop merely usable by agents to proving that an agent could carry a real delivery problem end to end. Instead of building a special “AI website builder,” I exposed the existing delivery lifecycle as explicit, machine-valid phases — Define, Provision, Implement and Release — over the same project artifacts humans use. That let me aim for a much stronger demonstration: give Astra realistic client material and one high-level instruction, then have it model the content, provision infrastructure, build the bespoke site, respond to independent review and release a working CMS-backed system. The important shift was from agent assistance inside individual tasks to a delivery system where humans and agents can work at the same durable boundaries and hand work between each other without relying on hidden context.

Report

2d ago

Maker

📌

Tabstop is the delivery system behind my bespoke web work. For this challenge I wanted to test a different idea from the usual “AI website builder”: can an agent work through the same explicit project boundaries that a human can, using the real project artifacts rather than a giant implementation prompt? The demonstration starts with a fictional but realistic architecture-practice client package containing ordinary client material and requirements — not a prewritten CMS schema, route list or design solution. GPT-6 Astra receives one high-level instruction and works through four phases: Define → Provision → Implement → Release Astra performs the project work. Tabstop validates the phase boundaries. Those boundaries are persisted in the project, so a human or another agent can inspect the same state and continue without needing the previous worker’s context. The technical article, run evidence and final result are linked from the launch page.

Report

2d ago

OpenAI

GPT-6 Astra Challenge

How did Astra change the scope or ambition of what you built?

Maker

Report

2d ago

Maker

📌

Report

2d ago