Subly - Frontier Hackathon

Project submitted✅ 🟪 Introduction Subscribe Now, Pay Never. This is the privacy-first PayFi protocol for subscription services such as Netflix, Disney+ and Spotify - powered by @solana, @Arcium, @perena and @PayPal. Here is Subly's submission👇

The show-and-tell thread for @subly_fi on Colosseum Discord has exceeded 100 comments.

🏟️ Colosseum Frontier Hackathon @subly_fi Day 5🏟️ Designed a privacy layer for x402 payments on Solana (with AI today). x402 enables AI agents to pay for APIs via HTTP 402, but all payments are fully public on-chain. Anyone can see who paid whom and how much. Design: - Clients deposit USDC into a shared vault (mixing pool). More users → stronger anonymity - A TEE manages balances off-chain and batch-settles to providers on-chain - On-chain observers only see "Vault → Provider", not individual clients - The host is untrusted. TLS terminates inside the TEE - Audit logs are encrypted (twisted ElGamal + hierarchical keys) for selective disclosure - If TEE fails, funds can be recovered via on-chain force-settlement with a dispute window - Fully compatible with existing x402 HTTP 402 flow Still refining the architecture.

🏟️Colosseum Frontier Hackathon Day 4 🏟️ I continue working on the following concepts I explored in the last hackathon with @subly_fi : - Privacy-first PayFi - Buy Now, Pay Never - Deposit once. Earn yield. Turn yield into payments. Previously, I just focused on subscription services as the payment target. However, subscription payments require on/off-ramp, and the fees end up being similar to credit card fees. So this time, I want to focus on AI agent payments that can be completed fully with stablecoin. Today, I looked into the paper “A402: Binding Cryptocurrency Payments to Service Execution for Agentic Commerce.” Atomic Service Channels for Agentic Commerce A402 introduces a trust-minimized payment protocol designed for autonomous AI agents. Here are the key highlights: 1️⃣How it Works: Exec-Pay-Deliver Atomicity - Atomic Service Channels (ASCs): Binds cryptocurrency payments directly to service execution using TEE and adaptor signatures. - The Loop: Ensures that payment is finalized if and only if the service is correctly executed and the result is delivered. - High Performance: Supports sub-second latency and thousands of requests per second by offloading interactions to off-chain channels. 2️⃣Privacy: Liquidity Vaults - Unlinkability: A TEE-based Liquidity Vault manages the lifecycle of multiple channels off-chain. - Batch Settlement: Aggregates numerous settlements into a single on-chain transaction, hiding individual interaction graphs (who is paying whom) and specific channel capacities from the public blockchain. Summary: A402 enables high-frequency, private, and secure machine-to-machine commerce with minimal on-chain costs I’ll work on the pitch deck and start building a conceptual prototype.

🏟️ Colosseum Frontier Hackathon Day 3 🏟️ I attended the Superteam Japan×Brave event today. It was a short time, but I was able to meet many builders and talk about what I’ve been working on. I haven’t fully decided what to focus on yet, so in the first week I plan to define that and update my pitch deck and landing page. I’m also researching how to achieve payment privacy while building a privacy-first PayFi for agents. In particular, I’ve been looking into x402 × privacy solutions with the help of AI. I found some interesting papers, repositories and web sites. I’ll share them in a thread. I’ll keep researching and work on defining the architecture for a prototype.

🏟️ Colosseum Frontier Hackathon Day 2🏟️ I attended the #TEAMZ event yesterday and today. I learned a lot about the latest in AI and Web3. There were many discussions around how crypto fits well with areas including stablecoins, RWA, AI agent payments, and AI agent identity. For the hackathon, I’ve been deep diving into challenges in AI agent payments: - Authorization: Defining what an agent is allowed to pay for is not straightforward. - Accountability: It’s unclear who is responsible for a payment — the user, the agent, or the service. - Budget Control: Agents can generate many small payments, so clear limits are needed. - Intent Verification: Agents may make wrong, duplicate, or malicious payments. - Security: Giving agents access to keys or payment credentials is risky. - Privacy: Payment activity can reveal sensitive user behavior. - Failure Handling: Payments can fail and require proper recovery. - Compliance: KYC, AML, and responsibility need to be clearly defined. - Connection to Existing Payment Rails: Integration with cards or banks is still difficult. - UX: Full automation feels risky, so controlled automation is preferred. Especially for Budget Control, Privacy, and Payment Rails, this aligns closely with what I’ve been exploring in @subly_fi . Next, I’ll evaluate solutions in more detail, including feasibility.

🏟️Colosseum Frontier Hackathon Day 1 Today I was thinking about whether @subly_fi concept — a privacy-first UX where you can earn yield just by depositing and use that yield for payments — can also be applied to AI agents. If the yield is low, the amount available for payments from yield becomes small. To achieve “Pay Never” through yield-based payments, you would need to deposit a large amount. But at the same time, taking too much risk is dangerous. As part of exploring the right yield layer, I looked into @reflectmoney today. Given the situation with Drift, it’s difficult right now to evaluate whether DeFi protocols that partially rely on Drift are safe to adopt. I also researched whether AI agent payments using the x402 protocol can be done in a privacy-preserving way. It seems like there isn’t a facilitator that supports this yet, so I need to think more about how to make it possible. Today I started from the solution side because I was curious, but I need to clearly define the problem first.

I’m considering adapting @subly_fi concept for AI agents. My original goal was to make subscription payments more efficient, and that hasn’t changed. From a technical perspective, I’m confident that the design is feasible. However, integrating services like PayPal payouts and MoonPay requires setting up a company outside my country and completing KYC, which is difficult to do right now due to cost and operational constraints. I’m not giving up on improving subscription payments. That said, there are still very few subscription services that directly accept stablecoin payments. To better demonstrate the value of paying through yield, I’m considering focusing first on AI agents, where stablecoin payments are already more common, rather than building a product that depends on corporate KYC from the beginning.

hackathon szn

I'm building in the Solana Frontier Hackathon, powered by @Colosseum! arena.colosseum.org/hackathon/soci…

happy 6th birthday @solana!!🎂💜🎉

🔥🔥🔥

1 day left to take over the top of the week spotlight!

@Arcium Congrats ☂️☂️☂️

Arcium Mainnet Alpha is Live.

<encrypted> time is now. we accelerate ☂️

I love @Arcium ☂️ I love <encrypted> ☂️

Solanaの「プライバシー」を、読み解く。 次回のCommunity Callでは、Solanaで最も注目を浴びている領域の一つ、秘匿化技術の「今」を探ります。 Cypherpunk Hackathon Arciumサイドトラック特別賞 @subly_fi の @0xyuki3 氏を講師にお迎え。 📅 1/28(水) 20:30 📍 Superteam Japan Discord

@buildermaps_io @Arcium @arpaofficial @iEx_ec @zama @privasea @nillion @silentdao_ @PhalaNetwork @LitProtocol Thanks for including Subly!!

WOW! 🔥 We mapped the privacy ecosystem and are amazed at how much it has expanded. Which project in the privacy ecosystem excites you most? Comment below 👇

@MubarakOnchain GMGM

Gm @subly_fi