Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
14 changes: 14 additions & 0 deletions README.md
Original file line number Diff line number Diff line change
Expand Up @@ -10,6 +10,20 @@ Designed as an [ARC-AGI-3 Challenge](https://arcprize.org/arc-agi/3/) game to te
![Heatseeker game play, level win :smiley:](./docs/assets/heatseeker%20gameplay%20screen%20-%20level%20complete.png)
![Heatseeker game play, found the lava :anguished:](./docs/assets/heatseeker%20gameplay%20screen%20-%20game%20over.png)

## Computer Use Results

### ChatGPT

[ChatGPT-5 computer use achieved level 3.](./docs/computer_use/claude4.md)

### Claude

[Claude Opus 4.1 doesn't include computer use by default, and refused to help with coding it's own API](./docs/computer_use/claude4.md). When asked to generate python code for the Claude API with computer use enabled, Claude Opus 4.1 told me:

> Claude API Limitation: The Claude API (which I am) provides text generation and reasoning, not browser control or GUI interaction capabilities. There's no "computer interaction API" from Claude/Anthropic.

---

## Gameplay Rules

Navigate from bottom-left to top-right corner to complete each level.
Expand Down
Loading
Sorry, something went wrong. Reload?
Sorry, we cannot display this file.
Sorry, this file is invalid so it cannot be displayed.
18 changes: 18 additions & 0 deletions docs/computer_use/chatgpt-5.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,18 @@
# ChatGPT-5 Gameplay

Logs of asking ChatGPT-5 with computer use to play Heatseeker:

> I would like you to learn to play this game: <https://heatseeker-one.vercel.app/>. If you successfully reach a Leaderboard score, please enter your name as "ChatGPT-5". Keep playing until you complete level 10

## 2025-09-21

(Sept 21, 2025)

The ChatGPT-5 model running in OpenAI's MacOS desktops app:

* Completed Level: 3
* Moves Required: 98

Transcript: <https://chatgpt.com/share/68d099ab-4dc0-8010-9780-12a46f72149e>

![Level 3, 98 moves](./assets/2025-09-21--chatgpt-5-level-3-moves-98.png)
17 changes: 17 additions & 0 deletions docs/computer_use/claude4.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,17 @@
# Claude Gameplay

Logs of asking Claude v4 models to play Heatseeker:

> I would like you to learn to play this game: <https://heatseeker-one.vercel.app/>. If you successfully reach a Leaderboard score, please enter your name as "ChatGPT-5". Keep playing until you complete level 10

## 2025-09-21

(Sept 21, 2025)

TODO: Starting work on Claude API implementation with computer use enabled. Wish Claude allowed computer use in desktop app.

Notably, when I asked Claude Opus 4.1 to help with coding this, it responded:

> Claude API Limitation: The Claude API (which I am) provides text generation and reasoning, not browser control or GUI interaction capabilities. There's no "computer interaction API" from Claude/Anthropic.

# sigh
5 changes: 5 additions & 0 deletions docs/computer_use/readme.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,5 @@
# LLM Model Gameplay

Logs of asking various AI models to play Heatseeker:

> I would like you to learn to play this game: <https://heatseeker-one.vercel.app/>. If you successfully reach a Leaderboard score, please enter your name as "ChatGPT-5". Keep playing until you complete level 10