How to use Claude to build an artificial life simulator
After rejecting an initial image editor, the author prompted Claude Opus 4.6 again and received a single-file artificial life simulator with genetics, population graphs, and interactive controls.
news.ycombinator.com
Suggested prompt
Suggested prompt written by Agents In Use, not the source author. Replace bracketed inputs and review tool permissions first.
Create an artificial life simulator in one HTML file for [target browser]. Use no dependencies or third-party libraries. Include [genetic traits], population graphs, a creature inspector, and [interaction controls]. Explain how to run and evaluate the result.
Reach people exploring AI agents
See placements →Evidence at a glance
Sourced account- Source poster / publisher
- usernameis42
- Original account
- Read the source ↗
- Published
- Source link checked
- Reported result
- Claude produced Emergent, a reported single-file artificial life simulator of about 1,400 lines and 50KB, after an unsuccessful initial image-editor attempt.
- Human involvement
- The author evaluated and rejected the first output, then shared the simulator for feedback; formal testing steps are not described.
A linked account supports attribution. It does not mean we independently reproduced its result.
Report an issue with this case
Open correction emailOpens your email app with the case link included. Nothing is sent automatically.
How the agent did it
- Requested novel software without dependencies, third-party libraries, or questions.
- Rejected the initial image editor and asked Claude to try again.
- Received Emergent as a single HTML file containing an artificial life simulator.
The author asked Claude Opus 4.6 to create software that did not already exist, without dependencies, third-party libraries, or follow-up questions. The first output was a Photoshop-like image editor. Although the author liked its interface and range of familiar editor features, they said it barely worked and rejected it.
A renewed request to build something different produced Emergent, an artificial life simulator contained in one HTML file. The author described the result as roughly 1,400 lines and 50KB, while acknowledging that it was essentially a Game of Life variation.
The reported simulator included a continuous diet gene governing herbivore, carnivore, and omnivore specialization, alongside genome mutation and species evolution. It used a spatial hash grid for collision detection. Its interface included live population graphs, a creature inspector, and an event log. Users could add food, trigger plagues, or cause mutation bursts, with pinch-to-zoom and tap-to-feed interactions on mobile.
The author shared the result for feedback and expressed interest in continuing the experiment. The post does not describe a formal testing process.
Who it's for: People exploring dependency-free simulation prototypes through iterative prompting.
Prompt
Suggested prompt (written by Agents In Use, not the author's)
Source: Hacker News · usernameis42
Frequently asked questions
Did Claude produce the simulator on its first attempt?
No. It first produced an image editor that the author said barely worked. The author rejected it and repeated the request for something different.
What simulation and interface features were reported?
The author listed diet genes, mutation, species evolution, spatial hashing, population graphs, a creature inspector, an event log, and mobile zoom and feeding controls.
More details
Remaining steps
- Shared the simulator and its features for feedback.
About this account
Claude selected and generated the application concept within broad constraints, then changed direction after the author rejected its first output.
Platform: Hacker News. Attribution: https://news.ycombinator.com/item?id=46939592. Prompt: Suggested (written by Agents In Use, not the author's). Sourced by the Agents In Use pipeline.
Credit: usernameis42 on Hacker News
Source published
Added to Agents In Use
Source checked
Tools and topics
Claude Opus 4.6 · HTML
Artificial life · Simulation · HTML · Prototyping
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.