Aaron C.-W. Liang · TKSpring(Term Structure)產品經理|固定利率借貸Product Manager, TKSpring (Term Structure) · fixed-rate lending

Fixed-Rate Lending, From DeFi to Institutional Repo.

鏈上借貸幾乎都是浮動利率。要做固定利率、固定期限,就得同時處理到期、續期與結算。我從第 2 號員工做起,負責三代產品的設計、開發與上線:從 zk 訂單簿、橫跨六條鏈的 TermMax,到為銀行與機構打造的 TermPrime。

Almost all on-chain lending is variable-rate. Fixed rates and fixed terms mean handling maturity, rollover and settlement at once. Starting as employee #2, I've led the design, development and launch of three generations of product: a zk order book, TermMax across six chains, and TermPrime, built for banks and institutions.

$100M+TermMax 歷史鎖倉高峰(非目前數字),橫跨 6 條鏈TermMax historical peak TVL (not current), across 6 chains
1.4M+TermMax 累積註冊用戶(不等於活躍用戶)cumulative registered users on TermMax (not active users)
0TermMax 已知資安事故,2025 年 4 月上線至 2026 年 10 月(過去紀錄,不是未來保證)reported security incidents on TermMax, April 2025 launch to October 2026 (a record, not a guarantee)

以上皆為產品層級的團隊成果;我個人負責的工作在下方逐項標示。

All three are product-level team outcomes; what I owned personally is labeled below.

  1. GEN 1Term StructureEthereum 上的 zk rollup 固定利率訂單簿zk-rollup fixed-rate order book on Ethereum
  2. GEN 2 · 2025TermMax全鏈上固定利率借貸,6 條鏈fully on-chain fixed-rate lending, 6 chains
  3. GEN 3 · 2026TermPrimeCanton Network 上的機構級附買回撮合與結算institutional repo matching & settlement on Canton Network

01 · The Problem

固定利率在鏈上為什麼難Why fixed rates are hard on-chain

  • 到期日會切碎流動性。每個到期日都是一個獨立市場,資金分散在不同期限,任何一個市場都可能不夠深。
  • Maturities fragment liquidity. Every maturity is its own market, so capital spreads across terms and any one of them can run thin.
  • 借款會到期,部位要能延續。沒有續期,借款人就得先還款再重借,中間要承擔利率變動和額外的交易成本。
  • Loans mature, but positions need to continue. Without rollover, borrowers repay and re-borrow by hand, exposed to rate moves and extra costs in between.
  • 機構要的是結算的確定性。銀行和機構無法承擔對手方的結算風險,也需要符合自身規範的託管方式。
  • Institutions need settlement certainty. Banks and institutions can't carry counterparty settlement risk, and need custody that fits their mandates.

02 · Three Generations

Three Generations of Product Decisions

每一代都回答上一代留下的問題。每段都標明貢獻類型:個人實作 產品決策 工程領導 團隊成果

Each generation answers the question the previous one left open. Every block is labeled by contribution: PERSONALLY IMPLEMENTED PRODUCT DECISION ENGINEERING LEADERSHIP TEAM OUTCOME

GEN 1

Term Structure

Ethereum · zkTrue-up
Problem
固定利率需要訂單簿撮合,但直接在以太坊上撮合太貴。
Fixed rates need order-book matching, and matching directly on Ethereum is too expensive.
Decision
以 zkTrue-up rollup 打造訂單簿。後來發現 zk 架構限制了鏈上可組合性,其他 DeFi 協議難以直接接上;我主導了下線這一代、改為全鏈上重做的決策。
Built the order book on a zkTrue-up rollup. When the zk architecture limited on-chain composability — other DeFi protocols couldn't plug in directly — I led the decision to sunset it and rebuild fully on-chain.
Result
這個決策直接催生了 TermMax。
That decision became TermMax.
My role
產品決策PRODUCT DECISION 工程領導ENGINEERING LEADERSHIP 負責設計、開發與上線,並主導下線與重做的決策。Led its design, development and launch, and the decision to sunset and rebuild.
GEN 2

TermMax

2025 年 4 月上線 · 6 條鏈Launched Apr 2025 · 6 chains
Problem
要在不犧牲可組合性的前提下,讓固定利率借貸有足夠的深度與易用性。
Make fixed-rate lending deep and easy to use without giving up composability.
Decision
全鏈上部署到 Ethereum、BNB Chain、Base、Arbitrum、Pharos、Berachain,並與策展夥伴推出管理型金庫。2026 年 7 月推出 Flash Rollover:一鍵把到期部位續到新的期限。
Deployed fully on-chain to Ethereum, BNB Chain, Base, Arbitrum, Pharos and Berachain, with managed vaults run by curator partners. In July 2026 shipped Flash Rollover: one click to roll a maturing position into a new term.
Result
團隊成果TEAM OUTCOME 歷史鎖倉高峰超過 $100M、累積註冊用戶 1.4M+、累積借款金額 $55M+;上線至今未有已知資安事故。Flash Rollover 上線幾週內,一位借款人就把約 $20M 的債務續到之後的期限。Historical peak TVL above $100M, 1.4M+ cumulative registered users, $55M+ cumulative borrow value, and no reported security incidents since launch. Within weeks of Flash Rollover's release, one borrower rolled about $20M of debt into later maturities.
My role
產品決策PRODUCT DECISION 工程領導ENGINEERING LEADERSHIP 負責設計、開發與上線,以產品經理身分帶領所有工程團隊。As product manager, led its design, development and launch across all engineering teams.
GEN 3

TermPrime

Canton Network開拓機構市場INSTITUTIONAL
Problem
銀行與機構也需要固定利率、固定期限的附買回(repo),但他們要隱私、要結算不會只完成一半,還要符合自己的託管規範。
Banks and institutions want fixed-rate, fixed-term repo too — but they need privacy, settlement that cannot fail halfway, and custody that fits their mandates.
Decision
建在 Canton Network 上,利用它的隱私與原子結算:撮合後資金與抵押品在同一筆交易中交換,消除「一方已交付、另一方未交付」的交割風險;合約、託管與信用風險仍需另外管理。我定義了機構上線模式:白名單錢包、錢包與服務商綁定,KYB 交由銀行與券商夥伴處理。
Built on Canton Network for its privacy and atomic settlement: once matched, cash and collateral swap in a single transaction, removing the risk that one side delivers and the other doesn't. Contract, custody and credit risk still have to be managed separately. I defined the institutional onboarding model: whitelisted wallets, provider-wallet binding, and KYB delegated to bank and broker partners.
Result
2026 年 6 月完成首筆正式交易並全額還款,7 月達到正式營運水準(續期已上線)。通過 Halborn 智能合約稽核,2 項中風險、5 項低風險全部修復。與機構交易台完成端到端測試網演練:部分撮合、抵押品協商、雙方簽署的鏈上借貸合約,一路到續期。另外推出抵押品不可再利用(non-rehypothecation)、由第三方託管的市場層級。
Completed its first production trade in June 2026, fully repaid, and reached production level in July (rollover live). Passed a Halborn smart-contract audit, with all 2 medium and 5 low findings remediated. Ran an end-to-end testnet rehearsal with an institutional lending desk — partial matching, collateral negotiation, a dual-signed on-chain loan, through rollover. Separately, added a non-rehypothecation market tier with third-party custody.
My role
產品決策PRODUCT DECISION 工程領導ENGINEERING LEADERSHIP 負責設計、開發與上線,以及產品路線圖與機構上線流程。Led its design, development and launch, plus the roadmap and institutional onboarding.

下一步:抵押品以所有權移轉(title transfer)方式放到出借方錢包、且不可再利用的附買回結算,以及跨網路結算(Ethereum、BNB Chain 與 Canton 並行),目前在 MVP 階段。

Next: repo settlement where collateral moves to the lender's wallet by title transfer and can't be rehypothecated, and cross-network settlement (Ethereum and BNB Chain alongside Canton), now at MVP stage.

03 · Shipping Safely

Every Number Has to Reconcile at Launch

產品決策之外,上線那一刻每個數字都得對得上。每個案例都標明是我親手做的,還是我帶隊推動的。

Beyond product decisions, every number has to reconcile the moment it goes live. Each case says whether I did it myself or led the team through it.

Launch 個人實作PERSONALLY IMPLEMENTED 工程領導ENGINEERING LEADERSHIP

準時上線On schedule

TMX 代幣發行(TGE)TMX token generation event

Problem
代幣與鎖倉合約一旦部署就難以修改,上線日已對外承諾,多個交易場所要同時開始。Token and lock contracts are hard to change once deployed, the date was public, and several venues had to open at once.
Decision
我親自部署並驗證 TMX 代幣與 vote-escrow(鎖倉換 NFT)合約,跑完整的預備環境演練與上線前安全壓力測試,並統籌工程、造市、交易所的上線日流程。I personally deployed and verified the TMX token and vote-escrow (lock-to-NFT) contracts, ran full staging rehearsals and pre-launch security stress tests, and directed the launch-day runbook across engineering, market making and exchanges.
Result
2026 年 8 月 25 日依承諾日期在多個交易場所同時上線。Launched on the committed date, Aug 25, 2026, across multiple venues at once.
Reconciliation 個人實作PERSONALLY IMPLEMENTED

58,974

條 vesting 釋放流,逐筆對帳vesting streams, reconciled one by one

Problem
代幣分配要部署成 58,974 條 vesting 釋放流(Sablier stream)、對應 58,974 個錢包;任何一筆錯了,上線後就是真實的錢。Distribution meant 58,974 vesting streams for 58,974 wallets; any mistake becomes real money once live.
Decision
在 BSC 上以 Sablier 部署後,逐筆比對分配帳本的收款人、金額、釋放曲線與可取消性四個欄位。Deployed on BSC via Sablier, then checked every stream against the provisioning ledger on recipient, amount, schedule shape and cancelability.
Result
上線前抓出 3 個錯誤(方案指派錯誤、漏部署一條、收款人重複)並修正,最後零未解決差異。Caught 3 mismatches before launch — a mis-assigned plan, an undeployed stream and a duplicated recipient — and fixed them, ending with zero unresolved discrepancies.
Security 工程領導ENGINEERING LEADERSHIP

稽核先於上線Audit first

外部稽核與安全審查External audits and security reviews

Problem
借貸協議的合約漏洞會直接造成資金損失,而上線後的升級同樣有風險。A lending-protocol bug loses funds directly, and post-launch upgrades carry the same risk.
Decision
新產品先過外部稽核,已上線協議的升級先過安全審查:TermPrime 經 Halborn 稽核;TGE 後的 Access Manager、Router 02 升級(含 Flash Rollover 修正)經 Immunefi 審查後才部署。New products go through external audits and live-protocol upgrades through security review first: Halborn audited TermPrime; the post-TGE Access Manager and Router 02 upgrades (including a Flash Rollover fix) shipped only after an Immunefi review.
Result
TermMax 自 2025 年 4 月上線至 2026 年 10 月,未有已知資安事故;TermPrime 稽核發現的問題全部修復。這是過去的紀錄,不是對未來的保證。No reported security incidents on TermMax from its April 2025 launch to October 2026; every TermPrime audit finding remediated. A track record, not a guarantee.

04 · Team & Track Record

From Employee #2 to Product Manager

  • 從第 2 號員工做起。職涯歷經智能合約工程師 → 技術主管 → 專案經理 → 產品經理。
  • Started as employee #2. Smart contract developer → tech lead → project manager → product manager.
  • 團隊從 5 人擴編到 30 人,之後為了效率精簡到 15 人;帶領三代產品的所有工程團隊。
  • Scaled the team from 5 to 30, later consolidated to 15 for efficiency; led every engineering team across three product generations.
  • 建立制度。設計公司晉升制度與 SOP,導入數據驅動的產品分析(客群留存、大戶分析、推薦分潤設計)。
  • Built the system around it. Designed the promotion ladder and SOPs, and introduced data-driven product analytics (cohorts, whale analysis, referral fee-sharing design).
  • 把 AI 帶進團隊。在工程與行銷團隊導入 AI 輔助的專案管理與流程自動化(看我自己的 Agentic AI 系統)。
  • Brought AI into the team. Introduced AI-assisted project management and workflow automation across engineering and marketing (see my own agentic AI system).

Hackathons

  • 冠軍1STZK Playground2022-11
  • 亞軍2NDStarknet House Hackathon2022-08

TermMax ↗ · TermPrime ↗

The Same Discipline Elsewhere

研究裡,我讓規則被學出來、再用 DRC 工具驗證;Agentic AI 裡,有風險的動作停下來等人按;在這裡,每一筆結算都要對得上。做法不同,原則一樣。

In research, rules are learned and then verified with a DRC tool; in agentic AI, risky actions stop for a human; here, every settlement has to reconcile. Different methods, one principle.

Automate what can be automated. Verify what must be correct.