What GPT-6 Astra chose to build
GPT‑6 Astra delivered three distinct prototypes—a string‑based puzzle game, a supply‑chain simulation, and a booklet‑printing tool—each refined through a review process that highlighted input handling, data visualization, and export functionality. The projects illustrate how AI‑driven agents can it…
In the latest round of AI‑driven development, GPT‑6 Astra produced three independent prototypes that were subjected to a rigorous review process. The projects ranged from a simple string‑puzzle game to a complex supply‑chain simulation, and finally a tool that arranges a manuscript into a printable booklet. Each prototype was built by autonomous agents that were instructed to explore, play, or utilize the system, and then the output was refined based on feedback from a coordinating reviewer.
String Puzzle Game: Fixing Input and Route Logic
The first prototype, called Loose End, featured a finite string, three pegs, and four bells. Players moved the brass end of the string around the board, catching it on pegs to create a path that would ring each bell. The original version suffered from a key‑press issue: short arrow‑key taps were ignored, making the game difficult to control. A coordinating agent identified this problem and sent the code back for a fix. The revised version now accepts brief key presses and includes additional checks that test two winning strategies and trace a long move sequence back to the starting position. The game remains a single‑player, author‑created puzzle and does not simulate physical rope dynamics such as friction or elasticity.
During the review, the agent was asked to demonstrate how the string’s route changes the amount of remaining string and to confirm that the game ends only when the string returns to its starting point after ringing all four bells. The updated prototype meets these criteria and provides a clear visual representation of the string’s path, making it easier for players to understand the mechanics.
Supply‑Chain Simulation: Visualizing Order Histories
The second prototype, named The Order Echo, models the flow of goods through a shop, wholesaler, depot, and factory. The simulation allows users to select a week and a stage of the supply chain and then view the arithmetic of orders placed under different rules. For example, during a busy shopping week, the shop might order 12.5 crates while the factory’s peak order rises to 76 crates. The simulation also tracks outstanding orders, supplier backlog, and the impact of higher demand forecasts on ordering policies.
Reviewers requested a unified scale for order histories so that users could compare different stages side‑by‑side. The final interface displays both the original policy traces (in dotted lines) and the adjusted runs, enabling direct comparison of how changes in ordering rules affect the supply chain over a 48‑week horizon. Although the model uses continuous crate quantities and assumes unlimited upstream production, it is not intended as a business forecast but rather as a teaching tool for understanding supply‑chain dynamics.
Printable Booklet Tool: From Manuscript to Sheet
The third prototype, called Pocket Press, takes a cover, manuscript, and back cover and arranges them into an eight‑page booklet on a single A4 or US Letter sheet. The tool provides a reading view alongside the print arrangement and can export the layout as an SVG sheet or an editable source file. Reviewers asked for discoverable save actions, folding instructions, and the ability to reopen the source. The final version meets these requirements: users can download the printable sheet, read back the geometry and text, and restore the manuscript from the saved source.
One notable design choice is the unconventional page order used for folding: pages 5, 4, 3, and 2 appear upside down above pages 6, 7, 8, and 1. This layout is intended to become a book after cutting a central slit and folding the sheet. During testing, the print sheet extended beyond the preview area at a 1,280‑pixel browser width, causing cropping at the top and bottom. A 900‑pixel view resolves this issue, but the paper result has not yet been physically printed or folded. Pocket Press supplies a numbered test sheet to support the next phase of experimentation.
What Happens Next?
All three prototypes have passed the initial review stage, but further testing is required before they can be released to end users. The string puzzle will undergo usability studies to confirm that the new input handling improves gameplay. The supply‑chain model will be validated against real‑world data to ensure its educational value. Finally, Pocket Press will need a physical printer and paper to confirm that the folding instructions produce a usable booklet. Until these steps are completed, the prototypes remain in a polished but untested state.
These projects demonstrate how AI agents can generate functional prototypes and how human reviewers can guide refinement by focusing on user experience, data visualization, and export capabilities. The iterative process ensures that each prototype meets specific design goals before moving forward.
Why It Matters
By showcasing GPT‑6 Astra’s ability to produce diverse, reviewable prototypes, this work highlights the potential for AI to accelerate product design across multiple domains—from gaming to logistics to publishing.
Key Takeaways
- Loose End fixed input issues and added route‑testing logic.
- The Order Echo provides side‑by‑side order history comparisons.
- Pocket Press arranges manuscripts into printable booklet sheets.
- All prototypes underwent a structured review process.
- Physical testing remains pending for Pocket Press.
FAQ
- What is GPT‑6 Astra? A generative AI platform that builds prototypes based on agent instructions.
- How were the prototypes reviewed? A coordinating agent identified issues and sent the code back for refinement.
- Will these tools be released commercially? They are currently in a prototype stage; commercial release depends on further testing.
Why it matters
These prototypes illustrate how AI can streamline the design cycle, allowing developers to iterate quickly and focus on user‑centric improvements before committing to full production.
Key points
- Loose End fixes input and adds route checks
- The Order Echo visualizes supply‑chain orders on a shared scale
- Pocket Press arranges a manuscript into a printable booklet
- All projects were refined through a coordinated review process
- Physical testing is pending for Pocket Press
Frequently asked questions
What is GPT‑6 Astra?
A generative AI platform that builds prototypes based on agent instructions.
How were the prototypes reviewed?
A coordinating agent identified issues and sent the code back for refinement.
Will these tools be released commercially?
They are currently in a prototype stage; commercial release depends on further testing.
