▤LLM CLICKERINFERENCE AS A SIDE HUSTLE
ALL SYSTEMS QUESTIONABLY GO
FROM YOUR BEDROOM TO THE FRONTIER

Just one more GPU.

DATACENTER / 001Local save ready
Available credits0.000.00 cr / sec
Tokens / second00.02 cr per token
GPU compute0CU / sec · 0 GB VRAM
Lifetime tokens · this run000:00 uptime
01 / INFERENCELIVE

Tiny · 125M

AUTOCOMPLETE WITH AMBITION

Click to generate. GPUs keep going without you.

Unsold tokens0worth 0.00 cr
Run a few prompts. Your first GPU costs 2.50 cr.
Per token 1 CUResearch bonus +0%
02 / YOUR DATACENTER0 GPUs

The spare room.

0 / 8 slots
YOUR FIRST GPU GOES HERE
Thermal load0.00 / 0.80 kW

Cool, quiet, suspiciously affordable.

INFERENCE STREAMIDLE

> Waiting for your first prompt_

RESEARCH & RESTART

Publish. Rebuild. Repeat.

Reset this datacenter for research papers. Each paper adds a permanent +10% compute bonus, including clicks.

0 papers First paper at 1M lifetime credits.
MILESTONES0 / 8
03 / SCALE UPTHE SILICON MARKET
Buy GPUs for automatic compute. Watch your rack space and cooling.
BUY QUANTITY

YOUR DATACENTER, YOUR BROWSER

Save & settings

Auto-saves every 10 seconds and when you leave. Keep a backup before clearing browser data. There is no account or cloud save.

How the economy works

Compute units (CU) become tokens at your active model’s CU/token cost. Bigger models pay more per token and use more compute, but improve credits per CU. You need enough total GPU VRAM to deploy them. Memory is pooled for this fictional game, not a real hardware compatibility guide.

Cooling limits usable GPU compute; it never creates debt. Adding hardware cannot reduce total GPU output, but you will get less than full performance when hot. Inventory keeps its original sale value when you switch models. Auto-sell can be disabled to bank tokens.

Offline gains use your saved hardware and active model for up to 8 hours. Rates, hardware and models are invented for gameplay. No model is actually running. Exported saves include game state only.

 CONFIRM ACTION

Start again?