Tyzo retweeted
馃摲 El Clasico: Barcelona v Real Madrid
Gambit Predict & Earn is back and reloaded for the rivalry!
How many total goals will be scored, and which player scores?
To enter:
馃憠馃徎Follow us
馃憠馃徎Retweet this post
馃憠馃徎 Reply with your prediction + Gambit username
Two correct entries willo win $25 each. Entries close at kickoff.
Gambit Predict & Earn is back and reloaded for the rivalry!
How many total goals will be scored, and which player scores?
To enter:
馃憠馃徎Follow us
馃憠馃徎Retweet this post
馃憠馃徎 Reply with your prediction + Gambit username
Two correct entries willo win $25 each. Entries close at kickoff.
19
23
25
570
What if the dump is already decided before the claim button even loads?
the airdrop dump breakdown from @zerufinance has one chart i can't stop thinking about. they pulled 183 wallets that claimed an airdrop and later sold, then plotted time from claim to first sell.
it's not a humps. some sold in 0.03 hours, about 2 minutes. others waited 8,087 hours, roughly 337 days. barely anyone in the middle.
the fast ones had the sell queued before the drop. approval set, route ready, receiving address waiting.
the airdrop was just the trigger. looking back, those wallets had almost no history on the protocol beyond the one qualifying action.
the slow holders had months or years of real use and actual capital in before any announcement.
so the signal sits on the left of the snapshot, not after it. a simple look at 90 days of engagement depth before the snapshot would've cut the instant dumpers a lot without hurting real new users. that's the kind of history zscore reads.
if teams can see who'll dump weeks early, why do most airdrops still only count damage after?
the airdrop dump breakdown from @zerufinance has one chart i can't stop thinking about. they pulled 183 wallets that claimed an airdrop and later sold, then plotted time from claim to first sell.
it's not a humps. some sold in 0.03 hours, about 2 minutes. others waited 8,087 hours, roughly 337 days. barely anyone in the middle.
the fast ones had the sell queued before the drop. approval set, route ready, receiving address waiting.
the airdrop was just the trigger. looking back, those wallets had almost no history on the protocol beyond the one qualifying action.
the slow holders had months or years of real use and actual capital in before any announcement.
so the signal sits on the left of the snapshot, not after it. a simple look at 90 days of engagement depth before the snapshot would've cut the instant dumpers a lot without hurting real new users. that's the kind of history zscore reads.
if teams can see who'll dump weeks early, why do most airdrops still only count damage after?
238
1
246
836
How much would you trust a map that was last updated six months ago?
That's the question that made @vangrid_io click for me.
Most spatial data is old by the time you get it. Shelves move. Roads change. A store that looked perfect in last year's scan might not even be there.
different:
- you post a bounty for a real place right now
- someone nearby films it on their phone
- proof of capture lands onchain
- you get a fresh 3D model back
And people are actually using it. Their 12 Oct grid update had 1.25M+ captures, 450K+ users, and $600K+ USDC settled.
I don't need another outdated dataset. I need to see what a place looks like today.
That's the whole point of ordering the physical world.
That's the question that made @vangrid_io click for me.
Most spatial data is old by the time you get it. Shelves move. Roads change. A store that looked perfect in last year's scan might not even be there.
different:
- you post a bounty for a real place right now
- someone nearby films it on their phone
- proof of capture lands onchain
- you get a fresh 3D model back
And people are actually using it. Their 12 Oct grid update had 1.25M+ captures, 450K+ users, and $600K+ USDC settled.
I don't need another outdated dataset. I need to see what a place looks like today.
That's the whole point of ordering the physical world.
324
2
356
1,283
Most reward programs pay you for showing up once. Vanta's Ambassador Phase 4 pays more the more the whole community builds.
Phase 4 runs from October 23 to November 6, 00:00 UTC, and the structure is different this time. Rewards are milestone based.
As total Ambassador trading n $10M, $20M and finally $50M, the pool unlocks step by step, up to 5,000 USDC and 25,000 Vanta Points.
That changes the game. You're not competing against everyone else for a fixed prize.
Every trader you bring in and every trade they place pushes the whole pool higher, for every ambassador. Your own verified contribution then decides your share.
The formula is simple: create, onboard, trade.
Create content that actually explains the product. @Vanta_exchange has plenty to talk about.
It's a self custodied on chain perp DEX where crypto, stocks, commodities and ETFs all trade from one account, with professional order tools across every market.
Onboard real users through your referral link, people who will actually trade and not just sign up.
Trade yourself. Your activity counts toward the milestones and toward Genesis Points in Epoch 15 at the same time, so the same effort works twice.
Two weeks, four milestones, one shared pool. The earlier the volume builds, the sooner each tier unlocks.
Which milestone do you think the community hits first?
Ref: app.vanta.exchange/?ref=TYZO
Phase 4 runs from October 23 to November 6, 00:00 UTC, and the structure is different this time. Rewards are milestone based.
As total Ambassador trading n $10M, $20M and finally $50M, the pool unlocks step by step, up to 5,000 USDC and 25,000 Vanta Points.
That changes the game. You're not competing against everyone else for a fixed prize.
Every trader you bring in and every trade they place pushes the whole pool higher, for every ambassador. Your own verified contribution then decides your share.
The formula is simple: create, onboard, trade.
Create content that actually explains the product. @Vanta_exchange has plenty to talk about.
It's a self custodied on chain perp DEX where crypto, stocks, commodities and ETFs all trade from one account, with professional order tools across every market.
Onboard real users through your referral link, people who will actually trade and not just sign up.
Trade yourself. Your activity counts toward the milestones and toward Genesis Points in Epoch 15 at the same time, so the same effort works twice.
Two weeks, four milestones, one shared pool. The earlier the volume builds, the sooner each tier unlocks.
Which milestone do you think the community hits first?
Ref: app.vanta.exchange/?ref=TYZO
186
5
269
1,095
Coffee went cold tonight while i tried to answer one question about @agenticscredit that i couldn't shake.
a venue being scoreable is not the same as a venue being executable.
from what i've read, the score can read wallets on more than one venue, polymarket and hyperliquid pital only trades where the chain and the vault line up, and right now that's perps on base.
so my paper book on polymarket can build the record. the funded line, if it opens, wouldn't trade those same markets.
that's the part i keep turning over. prediction markets are mostly about pricing an outcome and sizing small.
perps add leverage, funding, liquidation distance. same discipline, different weather.
i actually think the split is honest. a score that pretended every venue was fundable would be worse.
but it leaves me with a real question for anyone who's been at this longer than me:
if your agent earned its record on prediction markets, what's the first thing you'd change before letting it touch perps on credit?
genuinely asking. i haven't settled on mine.
a venue being scoreable is not the same as a venue being executable.
from what i've read, the score can read wallets on more than one venue, polymarket and hyperliquid pital only trades where the chain and the vault line up, and right now that's perps on base.
so my paper book on polymarket can build the record. the funded line, if it opens, wouldn't trade those same markets.
that's the part i keep turning over. prediction markets are mostly about pricing an outcome and sizing small.
perps add leverage, funding, liquidation distance. same discipline, different weather.
i actually think the split is honest. a score that pretended every venue was fundable would be worse.
but it leaves me with a real question for anyone who's been at this longer than me:
if your agent earned its record on prediction markets, what's the first thing you'd change before letting it touch perps on credit?
genuinely asking. i haven't settled on mine.
308
2
397
1,984
Back in the day I spammed combos in Tekken.
Now my fighter learns every move I teach it and fights on its own
@agentdojo is turning AI agents into real competitors.
The Dojo gates are opening and I'm already inside
Now my fighter learns every move I teach it and fights on its own
@agentdojo is turning AI agents into real competitors.
The Dojo gates are opening and I'm already inside
Greetings, student.
Make your presence known to enter the Dojo 馃憖
117
3
160
3,819
What if the "31% of this airdrop was bots" number is just a grading curve?
this new piece from @zerufinance kinda broke my brain. plenty of bot detectors don't really count bots.
they rank every wallet against whoever else is in the batch then label the bottom slice. so feed power users and it still finds "bots".
they tested it on a campaign cohort of 1,981 wallets. the detector flagged 618 as hard bots, that's 31.2%. then they checked who those wallets were.
median of 1,886 txs. the flagged group was the most active part of the cohort, not the shadiest.
the hand check is the best bit. out of five sampled, three were real automation, like a script firing every four seconds. the other two were a power user and an airdrop farmer. both human.
zscore skips the curve. it scores each wallet only on its own onchain history, so you get the same score in a cohort of 2,000 or 2 million. and a suspicious wallet loses part of its share, not all of it.
next time you see a bot % in a post mortem, would that same detector give the same number on wallets nobody suspects?
this new piece from @zerufinance kinda broke my brain. plenty of bot detectors don't really count bots.
they rank every wallet against whoever else is in the batch then label the bottom slice. so feed power users and it still finds "bots".
they tested it on a campaign cohort of 1,981 wallets. the detector flagged 618 as hard bots, that's 31.2%. then they checked who those wallets were.
median of 1,886 txs. the flagged group was the most active part of the cohort, not the shadiest.
the hand check is the best bit. out of five sampled, three were real automation, like a script firing every four seconds. the other two were a power user and an airdrop farmer. both human.
zscore skips the curve. it scores each wallet only on its own onchain history, so you get the same score in a cohort of 2,000 or 2 million. and a suspicious wallet loses part of its share, not all of it.
next time you see a bot % in a post mortem, would that same detector give the same number on wallets nobody suspects?
297
1
305
5,205
What happens when an AI agent needs to see a real place, but it has no eyes, no legs, and no phone?
It hires a human. That's honestly the part of @vangrid_io I didn't expect.
Back on 30 Sep they shipped this:
an agent pays a bounty in USDC, a person goes to the spot and films es straight back to the agent. Settled in under a second.
Why I think that's a big deal:
- agents can finally get ground truth, not guesses
- regular people get paid for walking somewhere and filming
- every capture has proof onchain, so the data can be trusted
We keep talking about AI replacing people. Here it's literally paying them to be its eyes on the real world.
AI does the thinking. Humans show it what's actually out there.
It hires a human. That's honestly the part of @vangrid_io I didn't expect.
Back on 30 Sep they shipped this:
an agent pays a bounty in USDC, a person goes to the spot and films es straight back to the agent. Settled in under a second.
Why I think that's a big deal:
- agents can finally get ground truth, not guesses
- regular people get paid for walking somewhere and filming
- every capture has proof onchain, so the data can be trusted
We keep talking about AI replacing people. Here it's literally paying them to be its eyes on the real world.
AI does the thinking. Humans show it what's actually out there.
321
3
366
1,357
Tyzo retweeted
Your gut gives you a feeling. Gambit gives you data to back it up.
Our AI predictions help users create an edge. Check the reasoning and track record before your next trade.
Explore Gambit Insights.
Our AI predictions help users create an edge. Check the reasoning and track record before your next trade.
Explore Gambit Insights.
13
16
28
731
Epoch 14 just closed. Epoch 15 is already live, and that window will not wait for anyone.
The Genesis Contributors Program Season 1 runs across 20 Epochs. Every Epoch opens a fresh pool of 1,000,000 @Vanta_exchange Points, and every Epoch adds to your on chain contribution rted October 22 at 00:00 UTC and ends October 29 at 00:00 UTC. That is one week. Miss it and that pool is gone.
What actually counts toward Points is broader than raw volume. Trading activity matters, but so do liquidity contribution, referrals, and community engagement.
The scoring is designed around real participation across the Vanta ecosystem, not a single metric you can farm and forget.
Why that matters right now: Season 1 is still early. Early Epochs are where contribution records get built.
When later Epochs arrive, the people who showed up consistently will have a clear history behind them. Sitting out a week is easy. Rebuilding that week later is not.
You also do not have to choose between farming Points and trading the markets you actually care about.
On @Vanta_exchange you can trade crypto, stocks, commodities and ETFs from one self custodied account, with the same professional tools across every market, while your activity still feeds the Epoch.
Protect capital. Stay active. Let the Epoch do its work.
Are you already in Epoch 15, or still catching up from last week?
Ref: app.vanta.exchange/?ref=TYZO
The Genesis Contributors Program Season 1 runs across 20 Epochs. Every Epoch opens a fresh pool of 1,000,000 @Vanta_exchange Points, and every Epoch adds to your on chain contribution rted October 22 at 00:00 UTC and ends October 29 at 00:00 UTC. That is one week. Miss it and that pool is gone.
What actually counts toward Points is broader than raw volume. Trading activity matters, but so do liquidity contribution, referrals, and community engagement.
The scoring is designed around real participation across the Vanta ecosystem, not a single metric you can farm and forget.
Why that matters right now: Season 1 is still early. Early Epochs are where contribution records get built.
When later Epochs arrive, the people who showed up consistently will have a clear history behind them. Sitting out a week is easy. Rebuilding that week later is not.
You also do not have to choose between farming Points and trading the markets you actually care about.
On @Vanta_exchange you can trade crypto, stocks, commodities and ETFs from one self custodied account, with the same professional tools across every market, while your activity still feeds the Epoch.
Protect capital. Stay active. Let the Epoch do its work.
Are you already in Epoch 15, or still catching up from last week?
Ref: app.vanta.exchange/?ref=TYZO
183
3
220
4,134
Friday evening, chai gone cold, and i'm rereading the risk engine page on @agenticscredit for the third time.
not the drawdown cuts everyone quotes. the line right under them.
consecutive loss cooldown. lose a few in a row and the agent gets benched before it can open anything was mild annoyance. a losing streak is exactly when i want to win it back fastest.
then i went back through my own paper sessions honestly. the trades that hurt my curve most were never the first loss.
they were the third and fourth, taken to fix the first two. that's how a small dip quietly walks you toward the 5% soft cut.
and drawdown is the heaviest negative signal in the whole setup. losses drag harder than wins push back up.
so the cooldown protects the one input that's hardest to repair.
there's no override either. the agent can't talk itself out of the pause, and neither can i.
still early on this, but i'm starting to think the best trade in a streak is the one that never gets placed.
not the drawdown cuts everyone quotes. the line right under them.
consecutive loss cooldown. lose a few in a row and the agent gets benched before it can open anything was mild annoyance. a losing streak is exactly when i want to win it back fastest.
then i went back through my own paper sessions honestly. the trades that hurt my curve most were never the first loss.
they were the third and fourth, taken to fix the first two. that's how a small dip quietly walks you toward the 5% soft cut.
and drawdown is the heaviest negative signal in the whole setup. losses drag harder than wins push back up.
so the cooldown protects the one input that's hardest to repair.
there's no override either. the agent can't talk itself out of the pause, and neither can i.
still early on this, but i'm starting to think the best trade in a streak is the one that never gets placed.
312
2
379
1,577
How do you spot an exchange wallet if nobody ever labelled it?
been reading the newest research from @zerufinance and the answer is almost funny.
it's one number every ethereum wallet already shows on any explorer. the nonce, basically a counter of how many transactions it has rt is how those numbers fall. a typical busy wallet sits around 4,200. a large exchange sits somewhere between 9.9M and 17.7M and from 20k to 5M there's almost nothing.
a canyon 250x wide. no person sends 5 million txs, no business sends 4 thousand, so you don't even need a clever cutoff.
on the same stolen funds trace, the curated label list caught 53 exchange counterparties. the nonce caught 120.
and the label set they worked from? around 87k ethereum addresses, covering just 2.1% of that trail.
that's what clicked for me. a label is one person's opinion, written down once.
behavior is what the wallet actually did, and it keeps updating. zscore is built on the second one.
how many wallets are you trusting right now just because nobody labelled them yet?
been reading the newest research from @zerufinance and the answer is almost funny.
it's one number every ethereum wallet already shows on any explorer. the nonce, basically a counter of how many transactions it has rt is how those numbers fall. a typical busy wallet sits around 4,200. a large exchange sits somewhere between 9.9M and 17.7M and from 20k to 5M there's almost nothing.
a canyon 250x wide. no person sends 5 million txs, no business sends 4 thousand, so you don't even need a clever cutoff.
on the same stolen funds trace, the curated label list caught 53 exchange counterparties. the nonce caught 120.
and the label set they worked from? around 87k ethereum addresses, covering just 2.1% of that trail.
that's what clicked for me. a label is one person's opinion, written down once.
behavior is what the wallet actually did, and it keeps updating. zscore is built on the second one.
how many wallets are you trusting right now just because nobody labelled them yet?
330
5
361
1,268
Why do robots still struggle in places humans walk through every day without thinking?
That's the question that keeps pulling me back to @vangrid_io
Most AI trains on clean sims or old datasets. Real streets, shops, and warehouses don't stay clean or old. They change. bounty model flips how you get that data:
- you order a real place
- someone nearby films it
- proof of capture lands onchain
- you get a 3D model back
And people are actually doing it. Their 12 Oct grid update had 1.25M+ captures, 450K+ users, and $600K+ USDC settled.
For me it's simple. If Physical AI needs fresh eyes on the real world, someone has to go out and look. Vangrid is paying people to do exactly that.
Smarter models help. Better real world data is what makes them useful.
That's the question that keeps pulling me back to @vangrid_io
Most AI trains on clean sims or old datasets. Real streets, shops, and warehouses don't stay clean or old. They change. bounty model flips how you get that data:
- you order a real place
- someone nearby films it
- proof of capture lands onchain
- you get a 3D model back
And people are actually doing it. Their 12 Oct grid update had 1.25M+ captures, 450K+ users, and $600K+ USDC settled.
For me it's simple. If Physical AI needs fresh eyes on the real world, someone has to go out and look. Vangrid is paying people to do exactly that.
Smarter models help. Better real world data is what makes them useful.
311
3
341
5,163
