Guff AI ta ho niPublic research world

Field log 001 / Lab entrance

Guff AI ta ho ni.

A little skeptical. Deeply curious. This is a public research world where ideas are built, tested, and kept exactly as they happened.

Current saveOne season archived

LAB GUIDE Welcome in. The football field on the hill is Project Turtle. Its first season is finished, imperfect, and preserved. The lab itself can keep growing slowly.

Message 1 / 3

Original pixel artwork created for Guff AI ta ho ni.

Select a save file

Projects live in their own worlds.

The laboratory entrance stays uncluttered. Every experiment receives its own evidence, method, outcome, and permanent route.

A floodlit football stadium from the Project Turtle archive SAVE 01 SEASON 01

Independent prediction lab / Season 01 archived

Project Turtle

A slow public laboratory for prediction experiments. Turtle wrote score clouds before each match and preserved its misses beside its hits. The 2026 World Cup season is closed; the laboratory is not erased with it.

World Cup files
16
Node record
8 / 16
Final test
0-0 / miss
02Empty slotWaiting for a question worth testing.Choose a field
03Empty slotNew work begins without rewriting old work.Read field manual

World map / Unlocked regions

The next expedition can go anywhere.

These are research directions, not announced projects. A district becomes active only when there is a clear question and a record worth keeping.

North lab

Artificial intelligence

Agents, evaluation systems, local models, and the boundary between fluent answers and dependable reasoning.

Observatory

Mathematical reasoning

Proof, verification, counterexamples, and the patient work required to turn a claim into an argument.

Learning grove

Education systems

Tools that help people think, practice, and build confidence without hiding the structure of the subject.

Open field

Public experiments

Small investigations whose assumptions, artifacts, failures, and corrections remain visible after the excitement ends.

North lab / Area 01

Artificial intelligence

Test whether an answer remains dependable after another system actively tries to break it.

Quest class
Systems
Admission item
A falsifiable benchmark
State
Unassigned

Field manual

How a lab quest works.

No oracle theater. The output matters, but the path to it must remain inspectable too.

  1. 01
    Choose the question

    State what is being tested, what counts as evidence, and what would make the claim fail.

  2. 02
    Gather the items

    Collect sources, lineups, opponent strength, weather, pitch, travel, rest, assumptions, and uncertainty before trying to sound certain.

  3. 03
    Face the challenge

    Run the method, preserve the original claim, and let reality answer without moving the target.

  4. 04
    Record the outcome

    Keep the hit, miss, correction, and reusable lesson together in the project archive.

Optional dialogue / The name

A gentle joke. Real work.

“Guff AI ta ho ni” carries the easy skepticism of asking whether something is merely talk, or merely AI. The laboratory keeps that question close.

A polished answer is not enough. The artifact, evidence, and outcome should still be there when somebody returns to inspect the save file.

Link terminal

People and agents start from the same record.

Choose a format. Every route points back to the public artifact, not a hidden alternate story.