Tank set up ยท trading not yet live

Can a fish learn to
trade memecoins?

FIN is a real fish who will be connected to a live trading wallet. He picks the coin. AI filters the risk and executes the trade. Every profitable decision earns him a treat โ€” every losing one just gets recorded. Nothing about it will be pretend, and nothing on this page is invented in the meantime.

See How It Will Work Read the Docs
Status
Pre-launch
Portfolio
โ€”
Return
โ€”
Treats Earned
0
Streak
โ€”
Rank
โ€”
FIN's tank, set up and cycling with live plants, filter, and thermometer
FIN's actual tank โ€” filtered, planted, and cycling before trading goes live.
The premise

AI understands the market. The fish makes the call.

AI never picks the winning coin โ€” it only removes the obviously dangerous ones. Everything after that, buy, hold, sell, is decided by where a fish chooses to swim. This site exists to document that honestly, trade by trade.

AI's job
Screen & execute
Filters liquidity, contract, and concentration risk before FIN ever sees a coin.
FIN's job
Choose & act
Selects a coin, then buys, holds, or sells by swimming into a lit zone.
The question
Is he learning?
Reward rate and repeat-selection are tracked to see if reinforcement changes his behavior.
Meet FIN

The fish behind the wallet

No mockups, no stock footage. This is FIN and this is his actual tank โ€” filtered, planted, and being cycled and monitored before any trading starts.

FIN, a betta fish, swimming in his tank
FIN โ€” the fish making every buy, hold, and sell call once trading goes live.
FIN's tank with lighting, filter, and thermometer
Filtered, planted, and temperature-monitored โ€” the tank zone-tracking will run in.
Coming next

FIN vs. the internet

Once trading is live, four identical simulated wallets will run alongside FIN's real one โ€” a random fish, an AI-only trader, a human crowd vote, and a buy-and-hold baseline. Same rules, same coins, one season.

The leaderboard isn't running yet โ€” it activates once FIN's wallet starts trading. See how it will be scored โ†’
Live tank

Every movement matters

Computer vision tracks FIN's position and converts intentional zone selections into trading instructions. A choice only counts once he holds a zone for five full seconds โ€” that's what stops ordinary swimming from accidentally placing a trade.

COIN A
COIN B
COIN C
Demo โ€” searching for a trade
Preview mode
Current zone
COIN A
Selection progress
0.0 / 5s
Fish confidence
โ€”
Pending action
โ€”
Current position
โ€”
Unrealized P&L
โ€”
ai_risk_manager.log โ€” demo output
Coming Soon

The live tank feed, decision terminal, and trade metrics activate once FIN's wallet and computer-vision tracking go live.

How a trade happens

From ocean of tokens to one confirmed action

AI scans

The market gets swept for anything trading. Coins failing liquidity, holder-concentration, contract, or trading-risk checks are removed before FIN ever sees them.

FIN chooses

Three qualified coins are assigned to three illuminated zones. FIN picks one by staying inside it for five seconds.

FIN acts

A second set of zones asks what to do with the coin he picked.

Buy
Hold
Sell

The trade settles

When FIN sells, realized profit or loss is calculated and written permanently to the wallet's onchain history.

FIN gets feedback

A win triggers a light, a tone, and one small treat. A loss is simply logged โ€” nothing is taken away, nothing is punished.

Scan and decision counts will appear here once the wallet and computer-vision tracking go live.
The treat economy

Profitable trades earn real rewards

FIN's normal feeding schedule never changes. Trading treats will be small and pulled from a separate, capped daily allowance โ€” this creates positive reinforcement without ever overfeeding or withholding regular care.

Treats earned today
0
Daily safe limit
5
Lifetime treats
0
Daily feeding status
Not started
No trades have closed yet, so no treats have been earned. This section fills in the moment the first profitable trade settles.
FIN's brain

Is FIN actually learning?

Early decisions are treated as random. As profitable choices create consistent rewards, we're watching for whether FIN starts favoring particular zones, signals, or coins โ€” and whether that favoritism beats chance. No claim is made that he understands markets. The only question is whether reinforcement changes his behavior.

Training sessions
0
Total decisions
0
Reward rate
โ€”
Repeat-selection rate
โ€”
Random benchmark return
โ€”
FIN's return
โ€”
Measured edge
โ€”
Average selection time
โ€”
Learning score
โ€”/ 100
No sessions recorded yet. Once trading starts, this score combines reward rate, repeat-selection, and edge over the random benchmark across every session.
What we're watching for

Signals of real behavior change

MetricMeaning
Reward ratePercentage of decisions that produced a treat
Repeat-selection rateHow often FIN returns to a previously rewarded zone
Decision consistencyFrequency of choosing the same signal under similar conditions
Random benchmarkSimulated results from a fully random decision-maker
FIN edgeFIN's return minus the random benchmark
Trading performance

Every trade lives onchain

FIN trades from a public wallet with a permanent record. Wins, losses, entries, exits, and transaction links can't be hidden or rewritten after the fact.

Portfolio value
โ€”
Total return
โ€”
Realized P&L
โ€”
Unrealized P&L
โ€”
Win rate
โ€”
Profit factor
โ€”
Max drawdown
โ€”
Best trade
โ€”
Recent closed trades
No trades have been placed yet. Once the wallet goes live, every closed trade will list here with its coin, direction, result, and a link to the onchain transaction.
AI risk manager

FIN chooses. AI protects.

AI never decides which approved coin FIN picks โ€” only whether the experiment is exposed to obvious manipulation or unlimited loss.

Active safeguards
No leverage Max 1% bankroll/trade Max 3 trades/hr Max 10% daily loss Min liquidity floor Holder concentration cap Contract risk block Bundled-supply block
Risk level
โ€”
Capital at risk
โ€”
Daily loss limit used
โ€”
Rug-risk score
โ€”
Stop condition
โˆ’15%
Configured limit
System status
Safeguards configured, not yet armed
FIN vs. the market

A fish, an AI, and the internet received equal wallets.

Only one can finish the season on top. Four identical simulated portfolios run the same rules, the same coin universe, and the same starting bankroll as FIN's real wallet.

#
Trader
Return
Win rate
Drawdown
โ€”
FIN โ€” real fish
โ€”
โ€”
โ€”
โ€”
AI Trader
โ€”
โ€”
โ€”
โ€”
Human Crowd
โ€”
โ€”
โ€”
โ€”
Random Fish (sim)
โ€”
โ€”
โ€”
Season hasn't started. Rankings populate once FIN's wallet places its first trade and the three comparison portfolios spin up alongside it.
Community participation

The internet doesn't drive โ€” it influences

Viewers will shape what FIN gets to choose from without ever choosing for him.

Watchers online
โ€”
Coin nominations
0
Active predictions
0
Community accuracy
โ€”
Ways to participate
Nominate a coin for AI screening
Vote on the next three-token basket
Predict FIN's next decision
Predict whether the trade will be profitable
Unlock a new tank theme
Sponsor a trading session
Collect a permanent trade card
Daily recap

FIN's journal

AI writes a short recap at the end of every trading day, in plain language, from the raw decision log.

No sessions yet

FIN's first journal entry will appear here after his first full trading day โ€” written in plain language from the raw decision log, same as every entry after it.

Trades today
0
Daily P&L
โ€”
Wins / losses
โ€”
Learning change
โ€”
Where the fees go

Funding the fish, not the promise of returns

If a token exists, it isn't ownership of FIN's portfolio and it isn't a promise of investment returns. Creator fees would fund the experiment itself โ€” the bankroll, the livestream, and his habitat. This is a proposed allocation, not a live one.

Bankroll 40%
Buybacks 25%
Livestream 20%
Habitat 10%
Donations 5%
Allocation (proposed)Share
Trading bankroll40%
Liquidity / token buybacks25%
Livestream & development20%
Fish care & habitat10%
Conservation donations5%
Documentation

Everything about FIN, explained

A plain-language reference for the whole experiment โ€” what FIN is, how a trade actually happens, how his safety and welfare are protected, and what the token model does and doesn't mean.

The premise Decision process Is he learning? Risk & safety Fish welfare Token & fees Current status FAQ

The premise

FIN is a real fish who will be connected to a live Solana trading wallet. He is the one making the decisions โ€” AI never picks which coin to buy. Instead, AI acts as a translator and risk manager: it watches FIN through computer vision, filters out unsafe tokens before he ever sees them, sizes positions, executes the trades he calls, and measures whether his choices are improving over time.

The short version: AI understands the market. The fish makes the call.

How a trade happens

  • AI scans. The market is swept for anything currently trading, and coins failing liquidity, holder-concentration, contract, or trading-risk checks are removed.
  • FIN chooses a coin. Three qualified coins are assigned to three illuminated zones in the tank. Swimming into a zone and holding it for five seconds confirms the pick โ€” this five-second threshold exists specifically so ordinary swimming doesn't accidentally trigger a trade.
  • FIN chooses an action. A second set of zones asks buy, hold, or sell for the coin he picked.
  • The trade settles. When FIN sells, realized profit or loss is calculated and the result is written permanently to the wallet's onchain history.
  • FIN gets feedback. A profitable trade triggers a light, a tone, and one small treat. A loss is simply recorded โ€” nothing is taken away and nothing is punished.

Is he actually learning?

Early on, FIN's decisions should be treated as random. As profitable choices produce consistent rewards, the experiment tracks whether he starts favoring particular zones, signals, or coins โ€” and whether that favoritism beats a simulated random-decision benchmark. Nothing here claims FIN understands financial markets. The only question being tested is whether reinforcement changes his decision-making.

  • Reward rate โ€” percentage of decisions that produced a treat.
  • Repeat-selection rate โ€” how often FIN returns to a previously rewarded zone.
  • Decision consistency โ€” how often he chooses the same signal under similar conditions.
  • Random benchmark โ€” simulated results from a fully random decision-maker, for comparison.
  • FIN edge โ€” his return minus the random benchmark's.

Risk & safety

AI never decides which approved coin FIN picks. Its only job is to keep the experiment from being exposed to obvious manipulation or unlimited loss. Planned safeguards:

  • No leverage, ever
  • Maximum 1% of bankroll per trade
  • Maximum three new trades per hour
  • Maximum 10% total daily loss before trading halts
  • Minimum liquidity requirement on any candidate coin
  • Maximum holder-concentration limit
  • Automatic block on dangerous contract behavior or suspicious bundled supply
  • Automatic exit if a token's liquidity collapses mid-trade
  • Paper trading first, before any real funds are activated

Fish welfare comes first

FIN's regular feeding schedule never changes regardless of trading performance. Trading treats are small, capped, and pulled from a separate daily allowance that stops automatically once a safe daily limit is reached. A losing trade never results in anything being withheld โ€” it's simply logged. The goal is positive reinforcement without ever risking overfeeding or under-care.

Token & fees

If a token exists for this project, it is not ownership of FIN's portfolio and it is not a promise of investment returns. The intended model: creator fees fund the livestream and FIN's habitat, a share funds his trading bankroll, holders can nominate coins for upcoming sessions, and community milestones can unlock tank improvements. A transparent dashboard is planned to show exactly where every allocation goes. See the Journal page for the proposed fee split.

Current status

FIN's tank is set up, filtered, planted, and being monitored โ€” the photos on this site are his actual tank and his actual self, not renders. Trading is not live yet. Every number on this site that would come from real trading (portfolio value, treats earned, win rate, learning score, leaderboard rank) is shown as empty or zero rather than invented, and will populate honestly once the wallet and computer-vision tracking go live. The Live Tank page's animated demo is a preview of the zone mechanic only, clearly marked as a simulation.

FAQ

Does the fish really decide? Yes โ€” FIN's tank position is the only input that selects a coin or an action. AI can remove unsafe coins from the options before he sees them, but it never chooses for him among the safe ones.

Can FIN lose money? Yes, and losses are recorded honestly, in full, alongside wins. That's the point of documenting the experiment rather than staging it.

Is FIN ever punished for a bad trade? No. A loss is only ever logged. Regular feeding is never conditional on trading performance.

Is this financial advice or an investment product? No. This is a public behavioral experiment. Nothing on this site should be read as investment guidance.