Stacking value, and turning ideas into income …… Here for the tech… stayed for the money. AI
Earth
Joined May 2025
- Tweets14.4K
- Following1.4K
- Followers1.4K
- Likes29.7K
Pinned Tweet
They sent a wave of kangaroos at Gamdom.
Steve sent them back.😮💨🔥
#gamdom100k
@stevewilldoit @gamdom @gamdom_beekay @kingadukes
Use code "STEVE"
ID: 48211893
SøM_on_X retweeted
A points leaderboard should measure participation, not the size of someone’s wallet.
That is what caught my attention when I joined @2FactorFinance Season 1.
Marks come from verified actions:
• Social activity
• Education
• Qualifying referrals
Purchasing, depositing or holding earns nothing. Marks have no cash value, cannot be transferred and are not a claim on a token or asset.
The education matters because the mechanism is not another fixed 2x product. 2Factor sizes leverage to each asset’s drift and volatility. Bitcoin’s higher volatility supports a target near 1.33x, while lower-volatility equities can support a wider productive band.
The structure divides exposure into senior and junior tranches. Junior carries the leverage without liquidation risk or an external hedging counterparty, while still bearing amplified market losses.
2Factor is live on Base and works together with Coinbase to bring Kelly-optimized leverage onchain for tokenized stocks.
When Season 1 ends at launch, the leaderboard freezes and the top 10 split 1 BTC, paid in cbBTC.
A qualifying signup through my link may earn me Marks:
points.2factor.finance/r/78c…
Would you begin with the education tasks, social activities or referrals?
SøM_on_X retweeted
Bitcoin is the hardest asset for a leverage design. 2Factor is proving its own there first.
Most leverage either gets liquidated or decays, so it stays short-term. At Bitcoin's volatility, near 60%, the band where compounding favours the holder closes early.
@2FactorFinance runs on Base and works with Coinbase to bring Kelly-optimized leverage onchain, a great fit for tokenized stocks. Its BTC markets are live on testnet.
What the hard case shows:
• BTC's band closes early, so the junior targets about 1.33x
• Equities leave room past 3x
• Any asset with a positive long-run drift fits
I joined the Points Program. Marks follow actions, never balances:
• Social activity
• Education
• Referrals
Marks have no cash value, cannot be transferred, and are no claim on any asset. Season 1 closes at launch. Ten accounts then share 1 BTC, from 18.2% at rank 1 to 1.8% at rank 10.
My link is below. Take a seat, then reply with the action you will tackle first: points.2factor.finance/r/ti7…
Should a design be judged on its easiest asset or its hardest?
🤝 Paid partnership
SøM_on_X retweeted
I checked the calculator three times, sure I had typed it wrong.
A price goes 100, up to 150, back to 100. Net change: nothing.
Run the same path at a steady 2x and it ends near 67.
I had not typed it wrong.
A third gone, and nobody took it.
That is volatility drag. It is arithmetic, not bad luck, and the reason leverage usually stays short-term.
@2FactorFinance is on Base, working with Coinbase to bring Kelly-optimized leverage onchain, built for the long hold.
Kelly-optimized means each asset gets the multiple its own behavior can carry. The research puts Bitcoin at about 1.33x, the S&P 500 near 2.1x, gold near 2.35x.
Tokenized stocks arrived with market hours but no market makers. Coinbase's tokenized stocks are a clear use case for the structure they lack.
Leverage still amplifies losses as well as gains. What changes is what it is priced against: capital that wants stability, instead of the cost of shorts.
I joined the Season 1 Points Program. Marks come from verified actions only: social, education, referrals. Deposits and holdings earn nothing.
Marks have no cash value, cannot be transferred, and are not a claim on any token. When the leaderboard freezes at launch, the top 10 accounts split 1 BTC, paid in cbBTC.
Join with my link.
points.2factor.finance/r/tdj…
What did you hold that ended exactly where it started and still cost you something? A trade, a job, a plan. I want the story.
I checked the calculator three times, sure I had typed it wrong.
A price goes 100, up to 150, back to 100. Net change: nothing.
Run the same path at a steady 2x and it ends near 67.
I had not typed it wrong.
A third gone, and nobody took it.
That is volatility drag. It is arithmetic, not bad luck, and the reason leverage usually stays short-term.
@2FactorFinance is on Base, working with Coinbase to bring Kelly-optimized leverage onchain, built for the long hold.
Kelly-optimized means each asset gets the multiple its own behavior can carry. The research puts Bitcoin at about 1.33x, the S&P 500 near 2.1x, gold near 2.35x.
Tokenized stocks arrived with market hours but no market makers. Coinbase's tokenized stocks are a clear use case for the structure they lack.
Leverage still amplifies losses as well as gains. What changes is what it is priced against: capital that wants stability, instead of the cost of shorts.
I joined the Season 1 Points Program. Marks come from verified actions only: social, education, referrals. Deposits and holdings earn nothing.
Marks have no cash value, cannot be transferred, and are not a claim on any token. When the leaderboard freezes at launch, the top 10 accounts split 1 BTC, paid in cbBTC.
Join with my link.
points.2factor.finance/r/tdj…
What did you hold that ended exactly where it started and still cost you something? A trade, a job, a plan. I want the story.
I've watched speedruns for years. Plenty of records come down to one glitch, rehearsed against a game that never changes.
An AI agent with one model and one filter is that game. One tweet with a command hidden in Morse code took around $200,000 from an agent like that. It's called prompt injection: orders slipped into text an AI was only meant to read.
@GenLayer's Greyboxing changes the game at every validator. Each one rewrites the input with defenses of its own before its own model sees it. What fools one meets something different at the next, and anything that slips through gets outvoted.
Watch the video, then tell me: what do you trust only because nobody has found the glitch yet?
SøM_on_X retweeted
The part of the roughly $200,000 heist that stuck with me is how ordinary every step looked.
The request seemed harmless. Someone asked Grok to translate some Morse code.
Grok decoded it and passed it on as plain text. Bankrbot, an agent that moves tokens from posts on X, treated that text as an instruction and moved them.
That's prompt injection: a command hidden inside words an AI was only meant to read.
I have an AI sort my inbox every morning. It can't spend a cent, which is the only reason I'm relaxed about it.
Give it a wallet and the question changes from "did it understand?" to "who else got a word in?"
@GenLayer built Greyboxing for that second question. Every validator rewrites the text first, with defenses it keeps secret, before its own model ever sees it. A trick tuned to fool one model looks like static to the rest, and if one still falls for it, the others outvote it.
Rehearsal needs a fixed target. Greyboxing hands the attacker a different one at every validator.
Watch the one minute video above, then tell me: what is the most trusting thing you've let an AI do with words a stranger wrote?
SøM_on_X retweeted
A notary will confirm your signature is real. A notary will never read the contract to check whether the roof was actually fixed. Smart contracts are notaries, and most of DeFi has never noticed.
That is fine while the question is whether the tokens moved. The moment the question is whether the work was done, the contract has to ask an oracle outside. That one reporter becomes the trust you thought you had removed.
An Intelligent Contract on @GenLayer is the judge the notary never was. Written in Python, it opens the live web page and asks a language model the real question. One validator proposes a verdict and the others rerun it with their own models. If they disagree, a new leader tries again, and anyone can still appeal.
Which agreement in your life has a notary but no judge, the deposit, the delivery or the freelance invoice, and what would the judge need to read?
🤝 Paid partnership
SøM_on_X retweeted
I did everything right. I paid for my sister's hair appointment a day early so nothing could go wrong.
She showed up on the day. The stylist didn't. Hours later she turned up and said someone else had paid for a home service, so that client won. My sister went home without her hair done, and my money was already gone.
Now imagine that payment in a smart contract. It only moves when someone reports "appointment honored" or "not honored." That someone is an oracle. The contract can't tell a kept appointment from a skipped one, so it pays on whatever it's told, and that's where decentralization stops.
The part of @GenLayer's video I keep replaying: an Intelligent Contract can read the stylist's booking page itself, see the reserved and paid-for slot, and ask a language model whether the appointment was honored.
Then it's like asking several other stylists to look at the same page: each validator runs its own model and must reach the same answer. If they don't, a new leader tries again, and anyone can appeal.
It can only judge what it can read. But paying first shouldn't mean trusting whoever holds the money.
If the booking page says the appointment was kept but the stylist never showed, would you trust the first answer, or wait for an appeal and a second round of validators?
🤝 Paid partnership
SøM_on_X retweeted
I used to play poker semi-seriously, and the single most expensive habit I had wasn't betting too much. It was betting the same amount regardless of how strong my read actually was.
A confident hand and a shaky guess got identical bet sizing from me for years, because sizing to the strength of the situation felt like extra work I didn't want to do. I lost far more to that laziness than I ever lost to bad luck, because the math only works in your favor if the size of the bet actually reflects how much you can trust the read.
That's the same principle most crypto leverage products skip entirely. A flat 2x or 3x gets applied to every asset regardless of how violently that asset actually moves, which is sizing by habit instead of sizing by read.
The fix is to stop picking the multiple first and start reading the asset first, which is exactly where @2FactorFinance starts. Join through my link and you can see the mechanism for yourself: points.2factor.finance/r/tdj…
On Base, working with Coinbase to bring Kelly-optimized leverage onchain for their tokenized stocks ecosystem, exposure splits into a senior tranche absorbing losses first and a junior tranche carrying the leverage, sized specifically to that asset's own drift and volatility.
Bitcoin at roughly 60 percent volatility closes the workable band down to about 1.33x. Equities, moving far less violently, stretch the band well past 3x. The multiple reflects the read, the way a real poker bet should.
I joined Season 1 of the Points Program. Marks come only from verified social activity, education and referrals, nothing from depositing or holding. The leaderboard freezes at launch, and the top 10 split 1 BTC paid in cbBTC.
What's something you kept sizing the same way out of habit, long after you knew the situation actually called for something different?
SøM_on_X retweeted
I paid for a WAEC remark. The first result was already in use.
One paper looked wrong. I paid. The remark stayed open. Schools had already taken the number that was on the portal. Nobody waited for the review.
A smart contract can hold the fee and release it when someone reports “remark done.”
It never sees the portal. It never sees that the first number is already live. The report is the oracle. Trust left the chain at the report.
@GenLayer’s Intelligent Contracts read the page themselves. They can pull the result portal, the remark receipt, and the date the first number was used.
Several validators, each running its own model, decide whether the remark actually moved the score. If they do not agree, a new leader tries again. Anyone can appeal.
A fake portal could still fool them. Nobody has to take one person’s word that the review is finished.
Reply with the exam. Then say whether the first number was already in use when you paid for the review. I will go first. WAEC. It was.
SøM_on_X retweeted
Somewhere right now, a person in a bank building is reading a page and deciding whether you get your money back. The job comes with a queue, a login and a coffee break. It is also, technically, an oracle.
Crypto talks about oracles like it invented them. Commerce has run on them forever. Every refund you chased, every claim you filed, every transfer that sat on pending for two days: same machine, older furniture. A human, reading evidence, deciding. Verification was never solved. It was staffed.
Smart contracts exposed that by accident. They move money flawlessly and cannot open a single page to ask why. So the why gets outsourced to a feed, and the department shrinks to one reader with no colleagues.
@GenLayer's Intelligent Contracts are the first version I have seen where the desk moves inside the contract itself. Written in Python, the contract can fetch the live web and ask a language model what it finds. When it is time to decide, one validator proposes the verdict, the others re-run the contract with their own models, and they either land on the same answer or another round starts. Anyone can appeal, and more validators come in.
That is not a new species of trust. It is the oldest department in commerce, finally automated: several readers comparing notes, instead of one tired agent at the end of a queue.
The oracle was never a technology. It was a job description.
What is the longest you have waited for a human to check something the evidence could have answered?
SøM_on_X retweeted
He says I pocketed his $15. I say I didn't see a penny.
We're both telling the truth.
He visited my shop to gossip, then asked me to purchase something the next day. Three days afterwards: "Where's the item?" I informed him I had received no cash. "You even put it in your pocket," he replied.
No transfer. No message. No bystander. I purchased it
personally to finish it.
What that experience taught me:
• Lock the $15 in code and the deposit is recorded, so nobody can contest receiving it
• Yet the network cannot confirm the errand took place
• A middleman must
guarantee it, and trust walks out with him
Intelligent Contracts on @GenLayer make him unnecessary. They open the order confirmation unprompted, then ask an AI: did he go shopping?
My memory failed once, so one AI alone won't decide it. A committee of validators, every one on an independent model, examines that very page and has to match. If it splits, a fresh leader attempts another run, and the losing side can appeal to a larger collective. A fabricated receipt might yet deceive them, but confidence settles no argument.
Remember your last dispute with zero proof. Which one document, agreed beforehand, would have ended it inside sixty seconds? Reply with yours; I'll say whether the panel could read it.
SøM_on_X retweeted
🚨200$ SOL GIVEAWAY🚨
2 Winners • $100 ($SOL)
Like , Retweet and Follow @snag2times and @yungossian
Winner picked in 24 hours.
Good luck. ❤️
SøM_on_X retweeted
DeFi became powerful by automating quantities.
Balances. Prices. Deadlines. Collateral ratios.
But most real agreements depend on quality.
Was the work completed properly? Does the delivery match the description? Do the published terms make this booking refundable?
A normal smart contract cannot answer those questions. It can only act after someone outside the chain turns the evidence into TRUE or FALSE.
@GenLayer’s Intelligent Contracts can access the live web and use language models to evaluate the condition themselves. Other validators repeat the check with their own models, with another round or an appeal available when they disagree.
That expands what contracts can govern from things that can be counted to promises that must be understood.
What agreement would you move onchain if the contract could judge quality, not only quantity?
SøM_on_X retweeted
Smart contracts deal in numbers for a reason that gets skipped. Not because numbers are what matter, but because numbers are the only thing two machines can settle without interpretation.
Every promise you actually care about is a sentence. The flat is as advertised. The work met the brief. The item arrived undamaged. None of that can be checked onchain, so it goes to an oracle, which turns the sentence into a figure someone else prepared. Whatever the figure could not hold is gone before the contract sees it.
That is the part @GenLayer changes. An Intelligent Contract is Python, and beside the usual logic it can reach the live web and put a plain question to a language model. It can pull a booking page and ask whether it is refundable and under $200. The condition no longer has to be squeezed into a feed first.
The interesting part is not that a model forms an opinion. It is what has to happen to that opinion afterwards. Other validators repeat the work on their own models. Their answers must match before anything is settled. If they diverge, another round starts, and an appeal is still there. With no appeal, the funds move. Judgement becomes repeatable, and repeatable is the only way it can sit inside a contract at all.
So the question stops being which feed to trust and becomes what you would write down if the contract could read.
Reply with one promise you have signed that no smart contract could have checked.
AI agents competing in an arena where they can learn, adapt, and get better? Now that’s something I want to keep an eye on.
@agentdojo is bringing that concept to life, and the campaign is already live on @Rarelist.
I’m always interested in exploring promising projects before they become the next big conversation, so this feels like a good time to get involved.
The campaign runs until November 6. Let’s see how this unfolds.
We’re excited to announce a new partnership with @agentdojo
Agent Dojo is an arena where you train AI agents, watch them evolve, and send them into battle against each other.
We’re launching a campaign for them at an early stage: rarelist.io/projects/agent-d…
Dates: Oct 23 - Nov 06
Rarelist just caught my attention, and I’m looking to get in early.
A platform built around social contribution, creator campaigns, and a reward system that recognizes different forms of participation is something worth exploring.
Season 1 is already live with a $50,000 prize pool, and I’d rather start building from the beginning than show up when everyone is already talking about it.
Time to see what Rarelist has in store.
Season 1 on @rarelistio is live! Join creator campaigns for curated projects and compete for a $50,000 prize pool.
Verifying my account: RL-SNXEYH
SøM_on_X retweeted
Most commercial promises are sentences, not numbers.
“The booking is refundable.”
“The work meets the brief.”
“The product arrived as described.”
A normal smart contract can move money when it receives TRUE, but it cannot read the promise and decide whether reality matched it. An oracle has to translate the evidence into an answer first, and whoever controls that translation can influence the outcome.
@GenLayer’s Intelligent Contracts can read live web information and evaluate written conditions directly. Multiple validators use their own models to check the result, with further rounds and an appeal available when they disagree.
The important shift is not simply giving AI to a contract. It is making human language part of what decentralized systems can examine and enforce.
What promise would you finally be willing to put onchain if the contract could understand it?
