Choose a topic
We consider cost, safety, compliance, and how the idea can be meaningfully tested.
Independent experiment log
Turning curiosity
into experiments
AI, software, web services, game development, automation, and online business. We test ideas found online and document what happened, along with the conditions and process.
Ideas found online, tested in the real world.
TRY IT ✦ OBSERVE IT ✦ RECORD IT ✦ VERIFY IT ✦ SUCCESS OR FAILURE
FIELD NOTES / EXPERIMENT LOG
The latest experiment record
We gave an AI God a world that began with 300 people. Across 745 years it used 46 Actions. Humanity still became extinct.
EXP
#004
#004
COMPLETE
I had an AI copy a drawing 100 times. What was left at the end?
A picture telephone game, 100 rounds each for three subjects. The subject came through to the end, while a tail became an arm and bacon became red lines.
EXP
#003
#003
COMPLETE
How far can one sentence take an AI build?
Fifteen builds from one-sentence requests with no testing. All 15 were usable within what was checked, and the limit was not found.
EXP
#002ADDITIONAL
EXPERIMENT
HUMAN WORLD
#002ADDITIONAL
EXPERIMENT
HUMAN WORLD
TERMINATED
I Gave an AI God a World of 300 People. Humanity Was Extinct 745 Years Later
The God checked the world 152 times and used 46 Actions. After a recovery from 22 people to 484, the final human died naturally.
EXP
#001
#001
PARTIAL
I let AI handle the whole game build
The first version worked but scored 1 / 5 for fun. The human-directed revision reached 3 / 5.
SIDE EXPERIMENT / DAILY DRAW
Your lucky number today
One number from 1 to 99 and one theme for the day. A small break for the curious.
YOUR NUMBER--
What number will appear?
HOW IT WORKS / EXPERIMENT METHOD
Start by testing
Try it in practice
AI work, human work, cost, time, errors, and unexpected turns are recorded separately.
Make sense of the result
Success, partial success, failure, and cancellation are all documented with the facts behind them.
ABOUT THIS LAB
AI moves the work forward.
A human checks it.
AI supports research and production, while a human checks the actual screens, behavior, and results before publication.
We keep failures and cancellations as well as successes, so others have something useful to judge similar ideas by.