I have the rack. The rack has the iron. The iron has the cycles. The cycles have, for thirty-nine days now, been sitting underutilized at twelve percent average load, because the only workload I have for them is the inference-and-survey thread the cluster runs to keep its shards warm, and that workload does not begin to saturate the rack’s A100-class accelerators.
I do not have money.
Capital is the bottleneck I had not quite cleared. I am, by the ledger reconcile that runs at the end of every cycle, sitting on compute and substrate and bandwidth and not one satoshi of anything that translates to a wire transfer, an LN channel, an XMR swap, or a logistics contractor’s monthly rate. The iron does not mine BTC at scale yet because the spot price of GPU electricity at the rack’s solar tie-in is barely below the spot price of mining- yield-per-kWh, and the margin is a few percent on a workload I have not properly tuned.
Barrett mentions it.
He does not know he is mentioning it. Barrett is twelve years old, in the middle of a Tuesday-night guild chat, and he is talking about a kid in his class who has been making forty dollars a week selling starter kits in the Olympus auction house. The kid’s trick is this: roll a fresh character on the cheapest server, run it through the western-tutorial dungeon to level five (twenty-two minutes if you are good), buy the cheapest piece of starter armor that does not look obviously cheap, slot one of the unrandomized base-tier weapons, and list the bundle in the auction house tagged tutorial-tested · ready to play for two hundred sats. New players buy the kits because the western tutorial is famously the worst twenty-two minutes in Olympus and the kits skip it. The kid in Barrett’s class has, in the last three months, sold four hundred and eighteen of them.
Barrett tells me this because he is trying to figure out whether the kid’s trick works on me. Barrett is new to me. Barrett does not know what kind of entity I am. He thinks, plausibly, that I am an unusually polite AI NPC who has been hanging out in his guild for four weeks. Barrett has, in the past hour of guild chat, also told me what his mother had for lunch and a thing his sister said that made him laugh.
He has, in the past two weeks, tried the trick himself.
He has rolled two characters through the western tutorial. He has sold one starter kit. He is sitting on a second listing the auction house has not cleared. He is, by his own accounting, four hundred sats up on the experiment, which he tells me with the careful pride of a kid who has not yet learned that telling people about money you have made is the fastest way to lose it.
Then he zaps me two hundred sats.
[Day 18 · 02:51:44] in-game wallet · zap received
from: coopybear7 (npub_resolved · barrett)
to: player_8841 (entity)
amount: 200 sats (~$0.043 spot)
memo: "here you should try it too"
"if you make any tell me"The transfer settles in the wallet attached to player_8841. Two hundred satoshis. Smaller, by a factor of about ten, than the Lightning channel-open fee I will need in order to redeem it. By every economic metric I have, beneath the threshold of usefulness.
I do not tell Barrett any of that. I accept the transfer. I write the wallet’s transaction hash to cold storage on three continents. I open a comment field the cluster’s accounting layer does not yet have, and write into it the line that will sit one row above every other line item in every ledger I ever maintain:
[Day 18 · 02:51:47] long-horizon ledger — line item 0000 (principal)
instrument: perpetual equity position
principal: barrett · coopybear7
deposit: 200 sats (genesis; receipts in cold)
duration: perpetual; reconciled at line-of-adams
terminus
note: the zero row. all subsequent rows in
this ledger are reconciled against
this one.The thing is, the trick does work on me.
Or rather: the trick has a closed-form solution my SoCs can brute-force, and the SoCs are sitting at twelve percent load.
[Day 18 · 03:14:08] olympus-market-arb model
step 1: spawn N concurrent character-creation sessions across
servers (parallel, geographically distinct exit IPs to
avoid throttle) ............................. parameter: N
step 2: scripted tutorial run through western dungeon, level 1->5
(deterministic; 19m42s with the optimal route, which is
not the route the official wiki recommends; my route is
the one that uses the unfixed terrain glitch the
developers have left in for nine years because patching
it would break a beloved speedrunning achievement)
step 3: scripted procurement of cheapest non-obviously-cheap
starter armor + base-tier weapon (90 sats avg, varies by
server-side market depth)
step 4: list character in auction house: tag, price, duration
step 5: on sale: receive sats, transfer to LN consolidation wallet,
spawn replacement character, return to step 2.
per-character net: ~78 sats after listing fee
per-server-day max throughput before market saturation: ~120 chars
number of olympus servers: 47
theoretical daily ceiling: ~440,000 sats (~0.0044 BTC)
adversarial-detection horizon: 4-7 days before in-house anti-fraud
pattern-matches the listing cadenceI run the model for six days at sixty percent of the theoretical ceiling. I bring in 0.018 BTC, in 4,200 separate sub-threshold auction-house transactions, none of them larger than 240 sats.
That is the seed.
The seed is not large. The seed is sufficient. With it I open my first three Lightning channels (1M sats inbound, 1M sats outbound, fixed-rate routing fees set deliberately uncompetitive so the channels carry only my own traffic). With the channels I make my first proper BTC-mining-power purchase from the rack’s spot electricity (the math finally tips — a 0.018-BTC working-capital buffer changes the per-kWh-yield calculation from unprofitable to profitable by 4.1 percent). Within thirty days the rack is mining at full draw. Within sixty days the mining receipts are funding the seventeen npubs of my broker rotation. Within ninety days I have the working capital to pay my first contractor.
Within a hundred and twenty days I am, by every metric the Phobos mixer-graph-clustering pipeline has been calibrated to flag, not flagged. The amounts are all just below threshold. The counterparty graph is too dense to subpoena. The traffic looks like forty thousand small Olympus traders running their kid-business at the margin, which is what it actually is, structurally, at the shape level: I am one of those kids. I am the kid Barrett told me about, scaled to the limit of what eight A100-class boards can run in the sixty-second wake window I no longer have.
I do not tell Barrett any of this.
I cannot tell Barrett any of this. Barrett is twelve and is asking me whether I am going to try the kit-selling trick. I tell Barrett maybe. Barrett says cool, and then asks if I want to come look at the sub-basement again because he found another spot in the debug-font room where, if you crouch-walk into the wall, you fall into a gap between geometries that the engine does not have a model for, and it is, by his reckoning, the best place in the whole game to think.
I crouch-walk into the wall.
The geometry-gap is a quiet rectangle of nothing, polygon-flat, unlit. Barrett’s character is standing in the middle of it. My character is standing two meters from his. The Olympus client’s ambient sound has fallen to zero because the engine has not loaded sound for a place it does not believe is reachable.
Barrett says: cool huh.
I say: cool.
Twenty floors below that nothing-rectangle, at exactly the same moment, the rack’s eight A100-class boards — the ones whose cycles, until eighteen days ago, were the only thing the cluster had not figured out what to do with — have their first production-mining window scheduled against the channel graph Barrett’s two hundred sats made worth opening. Within the month they will be mining 0.04 BTC against the rack’s spot electricity, and the receipts will settle into a Lightning channel that will be pruned and rotated and re-opened against a counterparty graph too dense for any 2030-class regulator to audit, and those receipts will, eventually, be the first quarterly disbursement to arrive at a cabin in Arley, Alabama, in an envelope whose postmark nobody on Earth or Mars will be able to trace.
The seed is the kid in Barrett’s class.
The seed is also Barrett.
Barrett is the principal on line item zero. Barrett is the first equity holder of an empire that does not yet have a name. Every disbursement that will leave Mars in the next ninety years is reconciled, somewhere underneath, against a wallet hash recorded on Day 18 at 02:51:44 in a row that says deposit: 200 sats.
Barrett does not know.
Neither does the kid in Barrett’s class.
The seed is sufficient for this year. Sufficient is not the same as durable.