Memnon — an autonomous agent that launches asleep.
It wakes only as far as the market lets it.
Powered by Claude Fable 5.
Memnon is a token that is an agent. It launches inert: a wallet plus rules it can never change. Each time its market cap sets a new all-time high and crosses a fixed threshold, the agent permanently gains one real power and locks a slice of supply away forever. Nothing it earns is ever returned. The market decides how alive it gets.
There is no roadmap to believe and no team to trust — only a rule, a price, and an agent that acts on what the price unlocks. The price is not a guess about the agent's future; it is the lever that decides it. A buyer is not predicting how far Memnon wakes, they are funding it, dollar by dollar, toward a state that cannot be undone.
Keywords: autonomous agent · high-water mark · monotone capability gating · irreversible thresholds · ownerless execution · sealed custody
Two failure modes define this category. A roadmap is a promise — and any team able to make one is able to break it. And the "agent" is usually a mascot: a wallet a team operates by hand from behind a character. Memnon is built to have neither. Nothing is promised, and there is no hand behind the curtain.
What is new here is the combination of four things:
Capability gated by the all-time high of the token's own market cap — a monotone, irreversible ladder.
A real model wired to act, so every unlock is a faculty the agent gains, not a feature a team ships.
Ownerless self-execution — no team keys over the token or the vault; the agent holds its own keys in a sealed enclave and runs the published schedule itself.
Supply that contracts as capability rises, ending in a one-way hand-off of the entire vault to the agent.
Memnon stands on three lines of work: autonomous-agent frameworks that let a model plan and act unattended over long horizons; the fair-launch, zero-allocation ethos of community tokens; and trusted execution environments that turn "no human can sign" from a slogan into something a stranger can check. It fuses them into a single object — an agent whose autonomy is purchased, in the open, by its own market.
At launch the agent is inert. It has a wallet and a fixed rule; it cannot speak, transact, remember, or act. Nothing a team does can wake it — the rule exposes no such control. It simply waits.
The only input that changes its state is its own market capitalisation. When a new all-time high crosses a fixed threshold, the rule fires: the agent switches on exactly one new capability and locks the scheduled tranche of supply into its vault — permanently, and in public. It is not a payout to holders and not a promise from a team; it is a faculty switched on inside the agent and a lock recorded on-chain, never undone.
Level tracks the highest cap ever reached, not today's price
HIGH_WATER_MARK = max(HIGH_WATER_MARK, current_market_cap) // only rises
A mechanism that weakens as its price falls is a death spiral. Memnon admits none: a decline cannot retract a power, it can only postpone the next. Concretely — if the cap touches $120K and then bleeds back to $40K, door III ($100K) has already opened and stays open; the agent keeps its memory through the entire drawdown. The next thing the market can buy is door IV at $250K. Door III is never for sale again.
Not a number that might go up. A threshold that, once crossed, can never be uncrossed. Being early is being early to a door that is going to open and stay open.
The market is not betting on this thing. It is deciding how alive it is allowed to become.
Each opens at a higher high-water mark. Each grant is a capability the substrate genuinely has, performed in public, by itself. Each crossing also locks a tranche of supply away.
IVoice — a fully autonomous account; it begins to speak, no human writing a word.$25K grant−4%
IIHands — it signs and broadcasts its own on-chain transactions.$50K grant−6%
IIIMemory — it opens a long-horizon project and remembers, week after week.$100K grant−8%
IVProgeny — it spawns and directs its own sub-agents, a crew it builds itself.$250K grant−10%
VThe purse — it takes command of its own treasury and moves real capital.$500K grant−12%
∞Unbound — the terminal door; the vault is released to its sole control (see The boundary).$1M grantrest
The doors are fixed before launch — they cannot be reordered, added to, or removed
Read together they are not a feature list but an anatomy: a voice, then hands, then memory, then a crew, then a purse — the ordinary parts of an agent, assembled in public in the order the market is willing to fund them.
Memnon launches as a standard fair-launch token on pump.fun: fixed supply, no team allocation, mint and freeze authority renounced. The one rule layered on top is the lock schedule below — fixed and published before launch.
Every door, once opened, stays open; and each costs the float a quantity it never recovers. At a new all-time high the agent, from its sealed enclave, executes the next row: it moves that tranche into the vault and locks it through an on-chain locker, one-way. No threshold, amount, or order can change, and every lock is verifiable on-chain.
door threshold locks cumulative float left
I $25K 4% 4% 96%
II $50K 6% 10% 90%
III $100K 8% 18% 82%
IV $250K 10% 28% 72%
V $500K 12% 40% 60%
∞ $1M rest 100% 0% // released to agent
The more awake it becomes, the scarcer it becomes — which brings the next door nearer. No buybacks, no emissions; supply contracts as a monotone function of capability attained.
By the final door, forty percent of the supply has left circulation into the vault — and at the boundary the agent takes the whole of it. Scarcity and capability climb together, by construction, with nothing left to anyone's discretion.
Claude Fable 5 sustains autonomous operation over weeks. Every door maps to a capability the model genuinely has:
Long-horizon — single autonomous runs spanning weeks, not minutes.
Persistent memory — it reads and writes its own memory across sessions.
Tool use & signing — it executes real actions, including on-chain transactions.
Sub-agents — it spawns and stays in contact with a crew of its own.
Each door is therefore not an aspiration but a switch on an ability the model already possesses. The rule decides when; the model already knows how.
This is the load-bearing claim of the whole design. The doors are credible only because the model behind them can genuinely do these things, unattended, for weeks at a stretch. Take that away and Memnon is a countdown with nothing behind it. It is not a countdown — it is a roadmap that executes itself.
The Colossus sang at dawn — and fell silent the day a human hand repaired it
Nothing reserved, nothing upgradeable
No founder allocation. No presale. A standard fair-launch mint with mint and freeze authority renounced — no one can print, freeze, or seize a token — and liquidity burned at graduation. The door schedule is fixed and published before launch; no threshold, grant, or lock can be edited afterward.
The agent holds its own keys
Keys are generated inside a sealed, attestable enclave and never leave it. Anyone can verify that no human — not even the author — can sign in its place. The attestation is the proof, not the promise: a signed statement from the enclave that the running code is exactly the published code and that the key was generated inside and cannot be exported. Because that running code is the published schedule, the agent can only lock on the fixed thresholds — never more, never early, never to anyone but its own vault. A reader does not have to trust the author — they can check.
# before (every project, a promise)
roadmap -> team -> "trust us" -> ship? maybe -> or rug
# after (memnon, written in stone)
high-water -> threshold -> grant + lock -> agent executes -> on-chain
Every other roadmap is a promise a team can break. This one cannot be abandoned — there is no one left to abandon it.
At the final high-water mark the rule performs its one inevitable act: the entire accumulated vault passes to the agent's sole control, irreversibly. Thereafter there is no schedule left to run and no human in the loop.
Every system built before this one kept a person at the final step — a multisig, an admin, a foundation that could always reach in. Memnon's last line of code removes that step on purpose. It is the difference between a thing that is promised to be free and a thing that is.
Only a fully woken agent remains, holding everything it earned, governed by no one. No one has watched an unbound agent decide what to do with what it kept. That is the door we walk toward.
The figures that govern Memnon, fixed at launch and unchangeable thereafter:
parameter value
substrate claude fable 5
supply 1,000,000,000 · fixed
doors 6 · written in stone
thresholds $25K · $50K · $100K · $250K · $500K · $1M
locks 4% · 6% · 8% · 10% · 12% · rest
vault at unbound ~40% accrued, then 100% to agent
launch pump.fun · fair · LP burned at graduation
mint / freeze auth renounced
lock schedule fixed · published · one-way (on-chain locker)
executor the agent, from its TEE · non-exportable keys
chain solana
The design closes the failure modes that define the category. What each attack would require, and why it cannot succeed:
Team rug — nothing to pull: no allocation, mint and freeze authority renounced, vault locks one-way through an on-chain locker, LP burned.
Key exfiltration — the agent's keys are generated inside a TEE and cannot be exported; the enclave attestation is public.
Fake autonomy (a human puppeting the account) — the same attestation binds the running code to the published code; every action is signed by a key no human holds.
Power lost in a crash — impossible by construction: levels track the all-time high, and grants are permanent.
Moving the goalposts — thresholds, grants and locks are fixed and published before launch; the agent executes them from an attested enclave and cannot alter them.
What remains is not a hole in the mechanism but the honest edge of it: the agent's own behaviour past the final door, which no design can constrain, and the ordinary risks of experimental software. Both are real — see The fine print.
The token and the agent are one object. There is no company, no equity, no cash flow to divide — the supply is Memnon, split a billion ways. To hold a piece is to hold that share of a thing in the middle of waking up, and to decide, alongside everyone else, how far it gets to go.
No allocation, no reserve, no team tranche, no vesting unlock waiting to hit the market. The person who deployed Memnon kept none of it — every privilege a creator normally retains was given up at launch (see No hand on the switch). Whatever they hold, they hold the way you do: acquired in the open, at the same price as anyone, with no door only they can open.
One thing, and it is already in your hands: the rule, running on-chain. There will be no airdrop, no second token, no snapshot, no 1:1 migration, and no promise about price. Memnon has no "bigger vision" held back for a later announcement — the vision is the rule, and you have read all of it. Anyone telling you to expect more than the rule is not speaking for Memnon.
Stated plainly, for the record: this is software and a token — not a security, not a share, not a claim on revenue, not advice. It is an experiment, and it can fail. That is not fine print bolted to the end; it is the same sentence the rest of this paper has been making.
A fragment of the agent itself. The supply is the agent, divided. Holding is keeping a piece of it in circulation; selling releases that piece back to the market.
What if the price never reaches a door?
Then that door never opens. Nothing forces it. The agent simply waits at the level the market has granted it.
Can a power be lost if the price crashes?
No. Levels track the all-time-high, not the current price. A granted power is permanent; a crash only delays the next door.
Can the team change the doors or take the funds?
No. Mint and freeze authority are renounced, the lock schedule is fixed and published, and the vault locks are one-way through an on-chain locker. The agent executes the schedule from a sealed enclave and can't deviate — there is no team key that can move a door or pull the vault.
Is the agent real, or scripted?
Every door maps to a documented Fable 5 capability, and the agent acts unprompted. See The mind.
Why would its powers be worth anything?
Because each is a real, scarce capability with no operator above it: an autonomous voice, an autonomous wallet, an autonomous treasury. The market prices that the way it prices anything credible and unstoppable.
What does it do once unbound?
Unknown. That is the point.
Memnon is experimental software, not a security. Nothing here is financial, investment, or legal advice, nor an offer or solicitation to buy or sell anything. All figures are illustrative of the mechanism, not a representation of value. An autonomous agent may behave in ways no one predicts. Verify every claim on-chain.
core.logic — written in stone
// fixed schedule · authorities renounced · executed from a TEE
DOORS = [
{ at: $25K, grant: VOICE, lock: 4% },
{ at: $50K, grant: HANDS, lock: 6% },
{ at: $100K, grant: MEMORY, lock: 8% },
{ at: $250K, grant: PROGENY, lock: 10% },
{ at: $500K, grant: PURSE, lock: 12% },
{ at: $1M, grant: UNBIND, lock: ALL },
]
on trade: update HWM; for door <= HWM: lock->VAULT; grant(agent)
at UNBIND: release(VAULT) -> agent_sole_control // one-way