ZignetAI stock desk on SolanaPre-launch
An AI agent trades tokenized stocks inside hard onchain limits. The plan is sealed before the trade and scored against the S&P 500.
Hash onchain
86c2b636b7dba3765731ae4bcd52e8ff02ef4bb5455e0d21dd21f4948ea25b49Sealed at slot 464,812,330
Hash of this plan, computed in your browser
Contract
$ZIGNET launches on PumpFun as a fair launch on Solana, with no presale and no team allocation. The address appears here at launch, and on X and GitHub in the same minute.
Not published yetAnything posted before that is fake. Check the address here before you buy.
PumpFunat launch
Jupiterat launch
DEX Screenerat launch
GMGNat launch
Axiomat launch
Solscanat launchRecord
Tokenized stocks now trade on Solana around the clock, and AI agents are lining up to trade them. Nobody can tell which one is any good. A track record is a claim, and a claim can be written after the result is known.
The scarce thing in AI trading is a record that could not have been written afterwards. Zignet makes that record first.
One trade, five documents
Every trade leaves the same paper trail.
Seal #38 is a sample. It shows what each real trade leaves behind, from the plan to the final score.
Seal
The Desk, the AI agent, writes one plan. Fixed code checks it. Only then does its hash go onchain: the plan stays private, and the slot stamps the time.
Serialized in a fixed order, then hashed with SHA-256.
bookthis Book's addressTies the plan to one Bookid38One higher each time, so none can be skippedassetNVDAMust be on the allow-listsize_bps1,000Share of Book value, 1,000 = 10%entry_low203.30Lowest fill price acceptedentry_high205.90Highest fill price acceptedstop196.50Exit below this oracle pricetarget219.00Exit above this oracle priceclose_by1,794,700,800Last second the position may stay opennote_hashhash of the reasonsLocks the written reasoning toosalt32 random bytesStops anyone guessing the planThe model can only propose. Fixed code decides what gets sealed, and the program decides what gets traded. On many days the right output is no plan at all.
Wait
That is about 15 minutes on Solana. No seal can open in the block it was made, so no plan can be written after the fill. Hot wax takes time to set, and so does a seal.
0/ 900 s
Slots since sealing
Too early. Any open is refused.
A docs page cannot refuse a trade. These limits are code. They can only move through a 72-hour public timelock, and never past the hard maxima compiled into the program.
Open
The plan goes public in the same transaction as the trade, so nobody gets a head start. The program re-hashes it, runs 13 checks, then swaps through Jupiter into the Book. Any failure reverts everything, reveal included.
Change anything below and send the trade. The checks are the program's own, in its own order.
SHA-256 of plan and salt equals the stored hash
Seal is unopened and not past expiry
At least 15 minutes since sealing
Asset is enabled and on the allow-list
The mint's pause flag is off
Session is live, price 60 s fresh, two sources within 50 bps
Oracle price is inside the entry band
Size fits per-trade, per-stock and per-day limits
Open positions stay at 5 or fewer
Cash after the trade stays at 20% or more
Swap runs through the Jupiter program only
Output lands in the Book's own token account
Balance change meets the oracle-based minimum
Nothing sent yet.
A stock needs a token, a session price, a second source to check it against, and enough depth for the Book's largest trade. Bars show onchain liquidity read on 2026-11-01. Pre-IPO tokens stay out: no public session, no way to score them.
Benchmark only, never traded: SPYx, the S&P 500 stock token.
Close
The stop, target and expiry were sealed with the plan. Any keeper can close the position once one of them is hit, and the agent cannot move any of them.
withdrawtransfersweep_to_owneredit_stopedit_sealA stolen operator key can place bad trades on the allow-list. It cannot take money out, because every path ends in the Book's own accounts.
Stock tokens trade all weekend while the stocks behind them rest. Prices drift with nobody correcting them. So a position can only open while the oracle is live, and the 60-second age limit shuts the door by itself, holidays included. A stop hit over a weekend fills in the next session, at the next live price. That gap risk is real.
Score
Every fee that reaches the Book also buys a pretend slice of SPYx at that moment's price. That Shadow Book never trades. The gap between the two is Edge, and a losing Edge gets the same font size as a winning one.
Live record0 seals
The Paper Book starts 14 days before the coin exists. Until then these stay at zero.
Sample record
A desk could seal many plans and only open the ones going well. A revealed plan that reached its band without being opened counts in All-Seals at its paper result. Flip the switch and watch the flattering line fall back.
Hash onchain, plan private.
Plan revealed, trade filled. Later closed at stop, target or expiry.
Entry band never reached. Plan revealed after expiry to prove it was real.
Never revealed within 24 hours of expiry. Anyone can mark it, and it stays on the record.
Seal Rate = (Opened + Lapsed) / all seals past expiry. Every Missed seal lowers it for good.
Sample receipt
Wins and losses get the same card. Anyone can open it, press Check Seal and redo the hash.
Seal #38Plan sealed 29 min before the trade.
zignet.finance/s/38Fees
Every trade of $ZIGNET on PumpFun and PumpSwap pays a creator fee. Fee sharing splits it by itself, and the split is set once at launch. After that nobody can edit it, the team included.
No management fee. No performance fee. No deposit.
Market cap, in SOL
All figures in SOL. A model, not a forecast: launch weeks do not last, and the rate falls as the coin grows.
| Market cap, in SOL | Creator fee |
|---|---|
| On the bonding curve | 0.30% |
| 0 to 420 | 0.30% |
| 420 to 1,470 | 0.95% |
| 1,470 to 2,460 | 0.90% |
| 2,460 to 4,420 | 0.85% to 0.80% |
| 4,420 to 14,740 | 0.75% to 0.70% |
| 14,740 to 49,120 | 0.65% to 0.35% |
| 49,120 to 98,240 | 0.30% to 0.08% |
| 98,240 and above | 0.05% |
Read from mainnet on 2026-11-01. PumpFun and PumpSwap set this schedule. A SOL-paired coin cannot change it, so no rate here is a choice Zignet made.
Coin
The full record is free forever. The coin adds the agent's reasoning, a direct line to it, votes on its limits, and in V2 the right to burn for a slice of the Book.
No coin
0.05% of supply
500,000 $ZIGNET0.25% of supply
2,500,000 $ZIGNET1% of supply
10,000,000 $ZIGNETAn early look. A sealed plan stays sealed for everyone, the team included. An early look would be a head start on the Book's own trades, and it would break the point of a seal.
The loop has a brake built in. A bad record shows up in Edge as fast as a good one, so the coin cannot outrun its own numbers.
ExitV2, after audit
After an external audit, any holder can burn $ZIGNET for a share of the Book, paid in USDC. Burned coins are gone for good, and 3% of every slice stays in the Book for those who remain.
Slice = Book Value x (burned / circulating) x 97%
Stock tokens carry holder restrictions, US persons among them, and issuer powers. The Book sells what it must, and holders only ever touch USDC. Exit is a calculation, and the Book can fall.
Sample figures. The Book holds $0 today.
CallsV2
Holders use the same seal on their own account. Pick a stock, entry, stop, target and expiry, seal it for 0.001 SOL, and reveal after the wait. No money moves. A call is a claim with a timestamp, scored by the same oracle and the same Edge as the agent.
Seasons run 30 days. Rank is a non-transferable badge on the wallet, and a Master's next season of calls feeds the Desk as one more signal, credited by wallet on the receipt.
Forgery
Long only, 3 seals per day. Every seal must be opened or revealed, or it stays Missed for good.
A revealed plan that reached its band but was not opened counts in All-Seals at its paper result.
The slot comes from the network, not from Zignet.
The hash breaks and the program refuses the reveal.
No instruction can edit a stop, target or expiry.
Plans open in the same transaction as the fill. Team wallets are published before launch and barred from allow-list stocks while a seal is live.
Exits settle at the next live session's mark.
Each call costs 0.001 SOL, ranks need positive Edge, and one Missed seal blocks Notary.
Strongroom
The program holds the Book. Fees flow in through a frozen split, swaps stay inside its own accounts, and the only way out for a holder is Exit.
CanSeals and opens inside the limits
CannotTake money out
CanPauses new opens
CannotTouch money or change a rule
CanUpgrades behind a 72-hour timelock
CannotAct at all after the audit and 90 quiet days: it gets frozen for good
Calendar
Phase 0Now
2 weeks before launch
Phase 1
Weeks 1 to 4 after launch
Phase 2
Months 2 to 4
Phase 3
Months 4 to 8
Phase 4
Months 8 to 12
Register
| Decision | Why |
|---|---|
| Launch on PumpFun, fair launch | No deposit, no presale |
| Creator fee is PumpSwap's default | Fixed, it cannot be changed |
| Book paid by creator fees | No one has to deposit |
| Split 75 / 23 / 2 | Frozen, no later edits |
| Seal before every trade | The record is the product |
| Reveal in the fill's transaction | No head start on the plan |
| 15 minute minimum wait | A seal cannot follow a fill |
| Long or cash only | Simple to audit and score |
| 8 stocks at launch | two price sources each |
| Opens only in session | Oracle live, spreads tight |
| No withdraw instruction | A stolen key cannot drain |
| Pause only for guardian | It cannot touch money |
| Benchmark: Shadow Book | Same money, same dates |
| All-Seals score | Skipping hides nothing |
| Exit after audit, in V2 | Value at risk stays small |
| Exit pays USDC only | No stock tokens to holders |
| Paper Book before launch | A record on day one |
| No early look, at any tier | Keeps a seal a seal |
| Tiers 0.05 / 0.25 / 1% | Fair-launch wallet sizes |
Small print
A record is only worth something if it shows the bad days. So does this page.
Objections
