Dae-Hyun Choi

@farsideapp Is there a specific channel for beta updates or just general?

If you are not in the Telegram or Discord yet, now is a good time to join. Everything we share about the beta, the platform, and what comes next goes there first. The closer we get to May 4th, the more that matters. Links below.

Join us now! Telegram: t.me/farsideofficial Discord: discord.com/invite/farside

@taifoon_io 2048 blocks per twig is such a smart way to manage scale at this level

Most cross-chain proof systems don't explain what's actually holding them together. This is the data structure underneath all of it, and understanding it changes how you think about cross-chain verification entirely. Why not a standard Merkle tree? A standard Merkle tree is fixed at construction. Every leaf is defined upfront and the root is computed once. If you want to add new data, you rebuild the tree and every existing proof becomes invalid. Blockchains never stop producing blocks. A proof system built on a fixed Merkle tree would require constant reconstruction and would invalidate every proof every time a new block is added. An MMR solves this. What an MMR actually is A Merkle Mountain Range is an append-only data structure. New leaves can be added without invalidating existing proofs. The structure grows as new blocks arrive and every proof generated against it remains valid indefinitely. Taifoon uses it as a catalogue. Every block header from every indexed chain gets appended as a leaf. The structure always reflects the current state of every chain while preserving the full verifiable history. How Taifoon structures it The indexing works in three levels: 1) Every 2048 consecutive block headers are grouped into a twig. The twig root is a Merkle hash of those 2048 blocks. 2) Twigs accumulate into a chain-specific MMR. The peaks of that MMR represent the full indexed history of that chain. 3) All chain peaks are bagged together into a single SuperRoot, a 32-byte hash committing to the finalized state of all 38 chains simultaneously. What this enables Any block on any of the 38 indexed chains can be proven against the SuperRoot with a compact Merkle path. The proof includes the block hash, the twig siblings, the chain MMR siblings, and the SuperRoot siblings, a complete cryptographic path from a single block to a structure covering 38 chains. The SuperRoot is deterministic. Same finalized blocks always produce the same hash. Anyone can recompute it independently from public RPC data. Current scale 88 million block headers indexed across 38 chains. Approximately 40 blocks per second collected across all networks. Every header becomes a provable leaf in the MMR. API docs: taifoon.io/docs/v5-proof-… #CrossChain #DeFi

@Stacks A Stacking DAO LST whale is running the stacking yield amplification strategy. The user is earning a net 14.6% APY using stSTX as collateral to borrow STX and swap it back to stSTX. By looping again, he will amplify the yield even further.

Whales in BTCFi are choosing Zest Protocol as their go-to lending protocol. The whale is earning 1.54% apy on his 150k STX borrowing position thanks to sBTC collateral yield. The next step is minting @StackingDao LSTs with the borrowed STX to earn 9% apy on top.

BTC-backed lending will be a trillion dollar market. The protocol that lets you do it without wrapping, bridging, or handing over custody wins. That's what we're building at Zest Protocol.

@taifoon_io taifoon.io been quietly building the verification layer everyone needs fr

Taifoon Proof is live. V5 MMR proofs across 40+ chains. One API. What it does: Every block header across 40+ chains is indexed in real time as an MMR twig. A Merkle Mountain Range is an append-only cryptographic accumulator that lets you prove any historical block belongs to the chain without storing every block. Each twig feeds into chain-specific MMR peaks, which aggregate into a single SuperRoot. The verification flow: → Source block collected and indexed as an MMR twig. → Intent anchored to twig position. → Destination chain confirms. → Source twig finality confirmed. → Inclusion proof submitted on-chain. The API: GET /v5/proof/blob/{chainId}/{blockNumber} Returns the full Merkle path from twig to peaks to SuperRoot. On-chain execution via TaifoonUniversalOperator.executeVerifiedCall(). Same proof format across all chains. Chain coverage: L1s: Ethereum, BSC, Gnosis, Fantom, Avalanche, X1 L2s and rollups: Arbitrum, Base, Optimism, Linea, Scroll, zkSync Era, Blast, Zora, Mode, Fraxtal, Lisk, Unichain, Lyra, Flow EVM, HyperEVM, Sei, Monad, ApeChain, RARI GRANDPA parachains: Moonbeam, Astar, Peaq Non-EVM: Solana, Bitcoin, Polkadot, Kusama, Tron, Aptos, Sui, ICP One proof format. Any chain. All verifiable on-chain. Current collection rate: approximately 40 blocks per second across all networks. Every header becomes a provable twig. taifoon.io/proof #CrossChain #DeFi

Jan 2025: 0 BTC on Zest Protocol. April 2026: 800+. 1,000 BTC up next.

@CramptonBuilt Processing signal across fragmented sources? That's where AI actually adds value.

Most "AI trading" products are not what they say they are. I have been building in this space long enough to say that plainly. Here is what is actually happening: The majority of products calling themselves AI-powered trading tools are doing one of three things. 1) Wrapping basic statistical indicators in an AI marketing layer 2) Operating as black boxes with no explanation of how signals are generated or why trades execute 3) Focusing entirely on price prediction, as if the hard problem in trading is not knowing where price will go, but having the discipline to act on it correctly when it does. None of that is AI. Most of it is not even new. The skepticism traders feel toward AI trading claims is correct. Ahd the products that exist today have mostly earned that skepticism. But here is where I think this actually goes: AI in trading will be about context. Not "what will price do" but "what narratives are forming, how crowded is this trade, what has changed in the conditions since I last looked." It will be about removing the parts of trading that humans are worst at: maintaining consistency under pressure, enforcing rules when emotion says otherwise, processing signal across fragmented sources simultaneously. The future looks like this: - Traders define their strategy, their rules, their risk constraints - Agents monitor, interpret, and execute within those boundaries - The human stays in control. The system holds the line That shift from prediction to context, from tools to systems, from manual execution to agent-assisted trading is where this is going. Most of what exists today is not even pointed in that direction. That gap is what I am building @farsideapp to close.

@SubQueryNetwork this is not about which model is smarter, it's about who has the right data pipeline

We asked three AI assistants a simple on-chain question: “What are the top 5 tokens on Uniswap by Token Value Locked?” ChatGPT is guessing ❌️ Claude doesn’t know ❌️ AskSubQuery just answers ✅️ The responses tell a bigger story about the current state of AI + blockchain data. • ChatGPT produced a confident, but incorrect, answer • Claude couldn’t access the data it needed • AskSubQuery returned the correct result instantly This isn’t about model quality. It’s about data access. Most AI tools still rely on static training data, scraped sources, or patchy APIs when it comes to blockchain queries. That means stale answers, guesswork, or no answer at all. AskSubQuery flips the model. It gives AI agents direct access to structured, indexed on-chain data in real time via GraphQL. No guessing. No hallucinations. Just verifiable answers. That’s the difference infrastructure makes. AI isn’t limited by intelligence anymore. It’s limited by access.

@cityprotocolHQ City Protocol said "why use 100 tools when you can use one" and honestly sheesh 😭

We're completing the Attention-Capital Flywheel in Q2 2026 through Neofinance-as-a-Service (NaaS). One unified infra layer for businesses to plug into onchain DeFi. No more stitching together hundreds of SDKs and APIs. One protocol. Full-stack neofinance. Out of the box. NaaS ships with: ◦ Vaults & Yield ◦ Swap ◦ On/Off Ramp ◦ Payments ◦ Lending & Borrowing ◦ KYC/AML Compliance All modular. All composable. All through our one-stop NaaS platform.

Hermes has now been officially deregistered. Despite the short run (since launch on Jan 15), it was incredible to build in an ecosystem with such passion and feedback. It genuinely revitalised our belief in blockchain communities. Operating without the 4-month immunity period made things especially challenging. We’re grateful to @YumaGroup for the opportunity to build here. Thank you to everyone who was part of the journey 🪽

@HermesSubnet yeahhh this one hurts. but thanks for the transparency til the very end. 🙏

This is not the outcome we’d hoped for. But as always, we want to be upfront with the community. Hermes is next in line for deregistration, and we don’t see a clear path out of the position we’re in. With that in mind, we want to give you advance notice that in 8 hours’ time, we will be liquidating our alpha holdings to cover operational expenses incurred since launch. Up until now, these costs have been covered out of pocket, and we have not previously sold any alpha. We’re sharing this ahead of time because we’ve always tried to lead with transparency, especially in moments like this. We understand the weight of this update and deeply appreciate everyone who has supported us along the way. Thank you to everyone who has been part of this journey. We’ve truly valued the ride and only wish it could have gone on longer. - Hermes Team 🪽

Soon.