Compute for the kids who'd never get it
Every trade of $PUPIL pays a fee. The fee rents the GPUs below. The GPUs run free AI tutoring for students who could never afford it. Holders get to use them too.
- treasury:
- —
- claimed:
- —
- gpus online:
- —/8
- tokens served:
- —
- students helped today:
- —
on the fleet right now
the ledger
what came in, what the GPUs cost, what is left
activity
- tokens
- 0
- requests
- 0
- claimed
- $0.00
- gpu cost
- $0.00
- inference cost
- $0.00
market cap
—
liquidity
—
24h volume
—
holders
—
10.0K to get access
creator fees claimed
—
gpu spend
—
treasury
—
—
runway
—
$PUPIL has not launched. The fleet is running on the team's own tab until it does.
how it works
01
Every trade pays a fee
pump.fun takes a creator fee on each buy and sell of $PUPIL and sends it to the dev wallet. That wallet is the treasury. Nothing else funds this.
02
The fee rents a GPU
The treasury pays hourly for consumer cards on the open market. When fees fall the fleet shrinks. When they rise it grows. The rent is on the ledger above.
03
The GPU teaches a kid
Open-weight models run tutoring sessions for students who cannot pay for one. Homework, reading, code. No account, no card, no ads.
04
Holders get the leftovers
Spare capacity goes to holders through an OpenAI-compatible API. Access is by balance: 10.0K → pupil · 100.0K → scholar · 1M → patron $PUPIL. Sell below the line and the key stops.
why
A private tutor costs more per hour than most families earn in a day. The students who need one most are the ones who never get one. A language model on a rented graphics card can sit with a child for as long as it takes, at a price that rounds to nothing, and it does not get tired or impatient at the fourth try.
The problem was never the model. It was the bill. Someone has to pay for the electricity and the card, and nobody wants to run a charity on a credit card. So we pointed a memecoin at it. The creator fee on $PUPIL is small, constant, and does not care what the price is doing. It pays the rent on the cards you see above.
We are the team behind Daso, the computer for curious kids. This is the same idea with the hardware on our side of the wire. Every session, every token and every dollar of GPU rent is on this page. The full write-up is here.
what a session looks like
sample transcriptswhy is 3/4 bigger than 2/3? they look the same size
Put them over the same bottom number. 3/4 is 9/12 and 2/3 is 8/12. Nine slices beats eight. Want to try 5/6 and 4/5 the same way?
what does a plant actually eat if it doesnt have a mouth
Light, air and water. The leaf takes in carbon dioxide, the roots take in water, and sunlight glues them into sugar. The plant eats the sugar. The oxygen is the leftover, and that is what you breathe.
my loop prints the last name 5 times instead of each name
for i in range(len(names)):
print(names[len(names)-1])You are always reading the last index. You want the one the loop is on. Change the print to names[i], or better, drop the index entirely: for name in names: print(name).
every model on the fleet
Prices are what a token costs the treasury in GPU time, not what it costs you.
access
rate limits in the docspupil
hold 10,000 $PUPIL
- req/min
- 30
- $/day
- $1.00
- max output
- 2,048 tokens
The door. Every model on the fleet, a sensible cap on how fast.
scholar
hold 100,000 $PUPIL
- req/min
- 120
- $/day
- $5.00
- max output
- 4,096 tokens
For scripts and small apps. Four times the throughput.
patron
hold 1,000,000 $PUPIL
- req/min
- 300
- $/day
- $20.00
- max output
- 8,192 tokens
Batch jobs and agents. The fleet is effectively yours to share.
Access is checked live on chain every request. No staking, no lockup. Sell below the line and the next call returns 403.