Agents In Use
What’s new todayLast update
160 use cases added today

How to use Codex to build a playable personal website

A creator used Codex with Astra to build a playable personal website on original Halo 3 Guardian assets. They manually placed pickups and memories and repeatedly tested the game, which still had some rough edges.

Source published news.ycombinator.comAgent: Codex
OUR ADAPTATION

Suggested prompt

Suggested prompt written by Agents In Use, not the source author. Replace bracketed inputs and review tool permissions first.

Build a playable personal website using [environment assets] in Three.js. Add [movement controls], [personal memories], and [quests or mini-games]. Provide tools for placing content against the geometry, then iterate on [playtest feedback] about movement and navigation.

Evidence at a glance

Sourced account
Source poster / publisher
runshouse
Original account
Read the source ↗
Published
Source link checked
Reported result
The author shared a playable personal website with memories, quests, and mini-games, while acknowledging pickup lag, imperfect lifts, and browser audio issues.
Human involvement
The author manually placed pickups and repeatedly played the website to fix bugs and improve movement and user flow.

A linked account supports attribution. It does not mean we independently reproduced its result.

Report an issue with this caseOpen correction email

Opens your email app with the case link included. Nothing is sent automatically.

How the agent did it

  1. Load original Guardian geometry and textures through Three.js.
  2. Use Codex with Astra to implement the playable systems and quest routing.
  3. Place pickups manually using callout maps and a coordinate locator.

The author built a personal website around Halo 3's Guardian map, filling it with photos, stories, quests, mini-games, and Easter eggs. Visitors pass through a main menu and lobby before exploring with keyboard and mouse. The project used original Guardian geometry and textures from the fan-run Halo Archive, loaded through Three.js.

Codex with Astra wrote essentially all the code layered onto those assets, including shaders, lighting, physics, collision, movement, lifts, weapon handling, and quest routing. The author reported 20,552 lines of TypeScript across 118 files. They placed every pickup manually using callout maps and a development coordinate locator.

Human work centered on selecting memories, deciding where they belonged, choosing game mechanics, and repeatedly playing to fix bugs and improve movement and flow. The author shared the site as an unofficial, noncommercial fan project. Remaining issues included occasional power-weapon pickup lag, imperfect lifts, and browser audio behavior. Who it's for: Creators who want a personal site built around an explorable environment and can invest time in content placement and playtesting.

Prompt
Suggested prompt (written by Agents In Use, not the author's)

Source: Hacker News · runshouse

Frequently asked questions

Did Codex generate the Guardian map assets?

No. The project used an original Guardian asset export from the fan-run Halo Archive. Codex generated the code needed to make the environment playable.

What work did the creator do manually?

The creator selected memories and mechanics, placed every pickup against the geometry, and repeatedly played the site to fix bugs and improve movement and flow.

More details

Remaining steps

  1. Play repeatedly to fix bugs and refine movement and user flow.

About this account

Codex with Astra wrote essentially all implementation code, while the creator directed content, placement, mechanics, and iteration.

Platform: Hacker News. Attribution: https://news.ycombinator.com/item?id=49732259. Prompt: Suggested (written by Agents In Use, not the author's). Sourced by the Agents In Use pipeline.

Credit: runshouse on Hacker News

Source published

Added to Agents In Use

Source checked

Tools and topics

Codex · Astra · Three.js · TypeScript

Personal website · Three.js · Halo · TypeScript

Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.