A database for
state evolution,
not just current state.
TimeSpaceDB stores every moment, every branch, every version — the temporal substrate AI-era applications were missing.
時空資料庫 — 不只記錄「現在」、而是記錄「演化過程」。 每個時刻、每個分支、每個版本，都能被查詢、重放、分岔。 是 AI 時代真正需要的「時間狀態底座」。
Get in touch →Today's databases were designed for a static world.
They answer “what is true right now?” — but AI systems, agents, audit, and reasoning workflows need to answer “what was true at version V on branch B at time T?”. That difference looks small. It isn't.
傳統資料庫只回答「現在是什麼」。但 AI 推理、Agent 決策、合規審計、回測模擬， 真正需要回答的是「在某個版本、某個分岔、某個時間點，當下知道什麼」— 這個差距、不是優化、是換資料模型。
Three primitives that change how state works.
Time-travel queries
時光回溯查詢
Replay the exact state your system saw at any past moment — for debugging, audit, compliance, or model reproducibility.
回放系統在任何過去時刻看到的精確狀態 — 給除錯、合規、稽核、模型可重現用。
Branchable state
可分岔狀態
Fork the entire database in microseconds. Run counterfactual simulations, multi-agent strategies, what-if scenarios — in isolation, in parallel.
微秒內 fork 整個資料庫。並行跑反事實模擬、多 agent 策略、 假設情境 — 互不干擾、互不污染。
Causal lineage
因果追溯
Trace any conclusion back to the observations, tool calls, and versions that produced it — full provenance for every AI decision.
把任何結論往回追到產生它的觀察、工具調用、版本 — 為每個 AI 決策提供完整可審計的因果鏈。
A long arc. Built in stages.
Three eras of computing, one data model. Every stage compounds on the same storage substrate.
三個運算時代、一個資料模型。每一階段都建立在同一套儲存底座上、不重來。
Classical foundation SHIPPED
Time-aware storage engine. Append-only by construction. The base everything else stands on.
時間感知的儲存引擎。Append-only 為基底。所有後續功能的地基。
AI-era APIs IN PROGRESS
Branch Diff · Causal Trace · Time-travel RAG · Agent Memory Versioning — the toolkit for building auditable, reproducible AI systems.
分支差異 · 因果追溯 · 時光回溯 RAG · Agent 記憶版本化 — 給開發 AI 系統的人一套可審計、可重現的工具包。
Long-horizon research FORWARD
A storage substrate this expressive opens directions current databases can't reach. We're keeping a few of those quiet for now.
當儲存底座表達能力夠強、就能走到一般資料庫到不了的方向。其中幾條、我們暫時不公開。