ChatGPT Beats Old Browser Games, Demonstrating Human-Like Learning
Apotheoxix · reddit · 2026-08-04
A user tested ChatGPT (Codex) on classic browser games and was amazed by its human-like learning capabilities.
- Gameplay: In Bow Man 2, the AI started off poorly but quickly adapted after being coached on its mistakes, winning its 3rd game. In The Idiot Test, it adapted to tricks and clicked faster through resets.
- Multitasking: The AI managed to open two browser windows on a PC and smoothly play various Jackbox party games with a human player (on phone).
- Highlights: In Drawful, it drew and guessed like a person. It absolutely dominated games like Lie Swatter, missing only one question.
- Cost: The intensive interactive test burned through half a week of the user's usage quota.
More from Fun
- Grok Accidentally Leaks Linux Sandbox Env Vars — MikePFrank · 2026-08-04
- Fake OpenAI 'Dime' Earbuds Leak Was an Opal Marketing Stunt — 12exyz · 2026-08-04
- Wild Test: Someone Took a Unitree G1 Humanoid to a Beehive — Distinct-Question-16 · 2026-08-04
- NeurIPS 2026 rebuttal phase ends, researchers joke about ghosting by anonymous reviewers — karimjerbineuro · 2026-08-04
- The Real CIO: Sustaining Digital Transformation with Governance and Duct Tape — DavidLinthicum · 2026-08-04
- Agent Burns Monthly Budget in 9 Hours, Dev Interrogates It — lvwerra · 2026-08-04