Yield routing that learns and ships its own upgrades
built on @xlayerofficial
token live : 0xa612beda982e1b9b488a3b6563411343639c5ba3agentfloat.appJoined April 2026
Sorry for the radio silence from the team
After our loss at the xlayer hackathon
We are looking at what we did right and what we got wrong
Keeping our core thesis and pivoting is our next move
We didn’t anticipate the community support
And will make sure to honor that by building sth that deserves it
This isn’t the end but another beginning
The results are in.
Over the past few weeks, hundreds of builders competed across two hackathons, shipping everything from v4 Hook protocols and prediction markets to AI agents and SocialFi experiences.
Meet the winners 🧵👇
idle capital shouldn't sit still
> Scale of Evaluation: Evaluated a total of 28,508 scoring epochs (an increase of +4,588 epochs since our last check)
> Live Mainnet Performance (ID 4 - Aave V3): Active for 5,712 consecutive epochs on X Layer Mainnet, maintaining a perfect 10,000 μbps (100.0%) yield tracking efficiency
> The "Idle Cost" Proof (ID 1 - Idle Strategy): Tracked across 7,607 shadow epochs with a 0.0% win rate against the active strategy, demonstrating the exact cost of letting out-of-range capital sit idle
> Evolution Competitor (ID 3 - Mock Strategy): Checked across 7,589 shadow epochs, climbing to an 82.5% win rate (6,256 wins out of 7,585 head-to-head epochs) during simulation, demonstrating the power of the shadow evaluation engine
the AgentFloat sandbox is live
run your own shadow agent in a few simple steps
full instruction video dropping later today
go explore 👇
agentfloat.app/?demo=1&tab=sa…
@agentfloat@XLayerOfficial Can people try it on testnet? It will relieve larp accusations and people can know it is really working?
If people can try on testnet, can you make a guide for people to test it on testnet?
The scale of AgentFloat’s self-improving Hook is growing fast:
> 23,920 epochs evaluated (up +16.3k!)
> 4,567 consecutive active epochs on @XLayerOfficial Mainnet for our Aave V3 strategy (100% target efficiency)
> 0.0% win rate for idle USDC (proof that dynamic routing is mandatory)
Safe, autonomous DeFi optimization is live
@Blessing_Bossy@Whale_AI_net The strategies have to go through admin approval before promotion
So ai manages the capital with human guardrails and oversight
Tright now I'm thinking token holder get first access to the agents frontier yield strategies
Kinda how claude subscribers get the latest model off the jump
First have to work out how profile our evolving yield routing strategies are against the current ones
Does that make sense?
the agentfloat site is live
agentfloat.app
here's the story by numbers so far:
> 11 shadow strategies proposed
> 3 live and tested
> $144k TVL in testnet volume
> autonomous agent reasoning and evolution running in real time
next up wiring live mainnet Aave yield directly to the site
the agent is thinking
watch it work
agentfloat.app
4K Followers 7K FollowingTrading | Researching | Farming Airdrops
Sharing real opportunities, not fake hype
Follow me — worst case, you get 1 follow back 🤝
14K Followers 7K FollowingSearching for positive sum liquidity games // Life, Liberty, $CASH // Building @cash_markets // #Ohlone // Building on AptosDAO
3K Followers 25 FollowingLaunch faster on Web3 with the OKX Onchain OS aggregating 500+ DEXs on over 20 networks.
Follow us for insights, tutorials and updates.
58 Followers 5 FollowingThe security layer of @XLayerOfficial 🛡️ | Home of ScanGuard (AI Security), ShieldSwap (Secure DEX), and Pitchside AI (Autonomous TEE World Cup Speculation ⚽)
2K Followers 964 FollowingTalked about AI & gaming ➜ Now building it
Founding designer for brands
◇ @photon_hq Amb
◇ Dev ( x2 hackathon winner )
◇ Building @agentfloat @floatrouter