built @lustyfinance & @stellar_turkey amb. | 23’ machine
Joined January 2019
- Tweets2.9K
- Following947
- Followers837
- Likes13K
Shh… Lethea is live 👻
Crosschain swaps in and out of Stellar mainnet from 14 chains with a single deposit.
Flip on Confidential and the route never shows up on the explorer.
lethea.xyz
@NEARProtocol 🤝 @StellarOrg
Yesterday marked the best crypto event in Turkey.
Hosted by @riseinweb3 and @BuildOnStellar at Feriye Palace, the night kicked off with a 36 hours hackathon and ended in an unforgettable celebration.
Fueled by that energy, we are pushing to take @lustyfinance even further.
tired of hitting claude design's weekly limit?
built rockdesign so you don't have to.
→ runs in your terminal
→ uses your existing claude plan (no api needed)
→ unlimited designs
→ fully open source
upload, configure, ship. that's it.
repo in the reply 👇
here you go ↓
github.com/utkurock/rockdesi…
stars, feedback, prs all welcome
some designs made with rockdesign ↓
Utku Kaya retweeted
JUST IN: Bolt CEO Ryan Breslow fired the company’s entire HR team because they were “creating problems that didn’t exist.”
Utku Kaya retweeted
it's only been one day since I shipped drand-relay and the @BuildOnStellar community already started building on top of it!🎉🎉
We didn’t make a regular coinflip on Blind.
Instead, we experimented with Stellar’s newly released trustless randomness relay infrastructure built by @0xkaankacar in a different way.
Users can open positions with:
- Random token
- Random side
- Random leverage
As a nod to randomness itself, the platform can also spin between:
- 10 different themes
- 4 different token charts
- r:r vault
Blind doesn’t require any wallet.
The system automatically creates 2 wallets for you and instantly starts you with 10K XLM test tokens.
If you want, you can also export the generated testnet wallets later. Everything stays local.
Check the replies.
(1/2)
We didn’t make a regular coinflip on Blind.
Instead, we experimented with Stellar’s newly released trustless randomness relay infrastructure built by @0xkaankacar in a different way.
Users can open positions with:
- Random token
- Random side
- Random leverage
As a nod to randomness itself, the platform can also spin between:
- 10 different themes
- 4 different token charts
- r:r vault
Blind doesn’t require any wallet.
The system automatically creates 2 wallets for you and instantly starts you with 10K XLM test tokens.
If you want, you can also export the generated testnet wallets later. Everything stays local.
Check the replies.
(1/2)
Try Now:
🔗 blind.utkukaya.co
____________
Drand Relay:
🔗 github.com/kaankacar/Drand-R…
Blind:
🔗 github.com/utkurock/blind
(2/2)
Introducing wall402 — an open-source x402 paywall gateway on X Layer.
AI agents and users pay per API call with zero-gas USDG. Get AI trading signals; token security scans, wallet intelligence reports and smart money alerts.
All powered by real onchain data. Built with 7 Onchain OS & Uniswap skills. @XLayerOfficial #BuildX #XLayerHackathon
THE $1,000,000 POLYMARKET STRATEGY
Some time ago, I built "5rcks .fun" and shared it with you.
The system allocates 5–10% of its bankroll per trade and splits positioning based on three analytical models:
- BTC Technical Analysis
- Orderbook Analysis
- Odds Analysis
Since the initial post, the simulation system went completely REKT once.
It then restarted with $1,000 and, over time, delivered strong performance, eventually surpassing the $1M mark.
Of course, this is purely a simulation. Achieving the same level of profitability in real market conditions would have been impossible.
The reasons are clear: Insufficient liquidity, Volatile odds and Slippage at larger position sizes.
However, if the system is applied using smaller, controlled position sizes, the outlook becomes more realistic.
Example scenario:
- 56% win rate
- 1,280 total predictions
- $80–$100 per trade
Results:
- 711 winning predictions × $180 payout (purchased at $90 with ~50% implied odds)
- 569 losing predictions × $90 loss
Total outcome:
- $127,980 gross revenue
- $76,770 net profit
Because the system sizes trades proportionally to bankroll, both gains and risks compound as capital grows. This creates the potential for larger profits; but also significantly larger drawdowns.
In short:
If you have a real edge on Polymarket and apply it through a disciplined automated system, building a scalable model may be possible.
This was purely an experimental automation test and does not guarantee profits.
There is a structural inefficiency in Polymarket 5-minute BTC markets.
Most people try to predict price.
Winners read the odds.
@Polymarket resolves markets using @chainlink oracle data. We track live price from Binance.
Here is the issue:
Chainlink public feed is delayed. In some 5-minute windows, there can be a $200–300 gap between Binance price action and the resolve price.
So even if you predict direction correctly, you can still lose.
The real edge isn’t price direction.
It’s detecting where the odds are mispriced.
That’s why I built a two-layer system:
I. Ensemble Signal Engine
– 40% Binance technicals (RSI, EMA 9/21, VWAP, momentum)
– 35% Polymarket orderbook (depth, spread, whale walls)
– 25% Mispricing detection
No trade if edge <5%.
Position sizing via Kelly.
II. Last 30-second strategy
If one side hits 90%+ probability:
$0.90 → $1.00 resolve ≈ 11% trade.
Small. Repeatable.
Currently running in simulation mode:
$1,000 virtual capital, real market resolution.
You can track strategy performance live.
See win rate, PnL, Sharpe, equity curve.
Build and backtest your own strategy.
5rcks . fun
Not financial advice.
But data changes how you see markets.
Our Avax Build Games journey has officially begun!🔺
We’re set to build and launch our 3D Shooter game from scratch in just 6 weeks.
Follow our progress and every milestone of this exciting marathon over at @skydngame.
🤝@AvaLabs
You can track real-time performance data at 5rcks.fun.
On aggregate, the “Last Seconds” strategy has been the most consistent performer.
245 Wins / 10 Losses
If we had direct access to @Polymarket live price feed for execution parity, the current 96% win rate could likely be closer to 99%.
You can join the experiment by building your own strategies.
🔗 5rcks.fun/strategy3
Create custom bots, run backtests, and don’t forget to submit them to the leaderboard.
There is a structural inefficiency in Polymarket 5-minute BTC markets.
Most people try to predict price.
Winners read the odds.
@Polymarket resolves markets using @chainlink oracle data. We track live price from Binance.
Here is the issue:
Chainlink public feed is delayed. In some 5-minute windows, there can be a $200–300 gap between Binance price action and the resolve price.
So even if you predict direction correctly, you can still lose.
The real edge isn’t price direction.
It’s detecting where the odds are mispriced.
That’s why I built a two-layer system:
I. Ensemble Signal Engine
– 40% Binance technicals (RSI, EMA 9/21, VWAP, momentum)
– 35% Polymarket orderbook (depth, spread, whale walls)
– 25% Mispricing detection
No trade if edge <5%.
Position sizing via Kelly.
II. Last 30-second strategy
If one side hits 90%+ probability:
$0.90 → $1.00 resolve ≈ 11% trade.
Small. Repeatable.
Currently running in simulation mode:
$1,000 virtual capital, real market resolution.
You can track strategy performance live.
See win rate, PnL, Sharpe, equity curve.
Build and backtest your own strategy.
5rcks . fun
Not financial advice.
But data changes how you see markets.
There is a structural inefficiency in Polymarket 5-minute BTC markets.
Most people try to predict price.
Winners read the odds.
@Polymarket resolves markets using @chainlink oracle data. We track live price from Binance.
Here is the issue:
Chainlink public feed is delayed. In some 5-minute windows, there can be a $200–300 gap between Binance price action and the resolve price.
So even if you predict direction correctly, you can still lose.
The real edge isn’t price direction.
It’s detecting where the odds are mispriced.
That’s why I built a two-layer system:
I. Ensemble Signal Engine
– 40% Binance technicals (RSI, EMA 9/21, VWAP, momentum)
– 35% Polymarket orderbook (depth, spread, whale walls)
– 25% Mispricing detection
No trade if edge <5%.
Position sizing via Kelly.
II. Last 30-second strategy
If one side hits 90%+ probability:
$0.90 → $1.00 resolve ≈ 11% trade.
Small. Repeatable.
Currently running in simulation mode:
$1,000 virtual capital, real market resolution.
You can track strategy performance live.
See win rate, PnL, Sharpe, equity curve.
Build and backtest your own strategy.
5rcks . fun
Not financial advice.
But data changes how you see markets.
One of the biggest reasons mentioned above is that this market needs to be approached probabilistically rather than purely through technical analysis.
Polymarket and Chainlink data often do not align.
data.chain.link/streams/btc-…
Looking into @nesaorg lately. A new Layer 1 focused on trusted and verifiable compute rather than the typical centralized AI API model.
The architecture is built around private execution and cryptographic verification at the base layer. No single provider sees the full model or the full query. The network proves correctness before returning results.
Notable components include:
• Equivariant Encryption
• Homomorphic Secret Sharing over encrypted embeddings
• Zero knowledge proof systems for inference correctness (KDD 2024)
• Model agnostic hybrid sharding for scaling
• MetaInf inference scheduler (COLM 2025)
Instead of the usual pattern of a centralized AI pipeline with a blockchain wrapper, this approach looks more like confidential inference and verifiable compute at scale. There is also a public model marketplace and an open participation layer for nodes and contributors.
Not making a final judgment yet. But the direction is clearly different from the common LLM infra stack. Less black box. More programmable trust.
You can find other information about Nesa in the document below.
🔗 docs.nesa.ai/nesa
Worth watching as the onchain compute ecosystem evolves.
Hello @icmdotrun community and @7VoMercy!
I recently joined you by purchasing an ICM NFT and I would like to share the journey of my own project, @Fash_AI from the very beginning.
I first presented my project in 2023 at the @solana HackerHouse. Back then, it was much more basic and only available in my local language. But the bigger challenge I faced was limited resources. Like many builders, I reached out to over 30 local VCs, and the only thing I got in return was a 1 AWS credit.
That’s when I truly understood the “RIPVC” mindset led by @pasternak's team. Because one of the biggest struggles for builders like us is this: no matter how good your project is, without strong backing, there’s only so far you can go. That’s why communities like icm.run and startup-friendly projects like @believeapp are such a huge opportunity for us.
With the revenue we managed to generate, we first went to a Solana event in Hungary and introduced our project to the local community. From there, we significantly expanded FashAI, scaling it to serve millions of new users.
We bring fashion together with both AI and social interaction. Users can get personalized style advice; explore ready-made outfit suggestions, try them on, and most importantly, connect with others in an interactive space. This allows people to take their first steps into becoming influencers; gaining followers and even earning money on the Web2 side.
On the Web3 side; we combine this massive dataset with various side projects, building a bridge that connects fashion, AI and blockchain in meaningful ways.
