本文由 AI 分析生成
建立時間： 2026-05-30 來源： https://x.com/HaoliYin/status/2045184552155664490
Summary
A reflection from two years as an MTS at a research startup arguing that startup research is a distinct discipline — not a compromise between academic research and engineering, but a fundamentally different practice with its own structure. The framework: frame the long-term goal as a research vision (uncertain path), while framing the current sprint as an engineering problem (well-defined, time-boxed). This dual-timescale approach separates noise (single sprint failure) from signal (multiple sprint failures in the same direction = vision adjustment needed).
研究創業公司兩年 MTS 的心得：新創研究不是學術研究與工程的折衷，而是有自身規則的獨立實踐。核心框架：長期目標用研究視野框架（路徑不確定），當前衝刺用工程問題框架（明確定義、限時）。
Key Points
- The “spectrum” framing is wrong: startup research is not a tradeoff between academic freedom and engineering deadlines — it operates on two simultaneous timescales
- Long-term vision: resourced like a research bet (commit budget against uncertain payoff); communicated to customers/executives as the “story”
- Current sprint: resourced like an engineering project (specific people, specific time, measurable sub-problem); communicated to team as concrete goal
- Single sprint failure = information → re-plan next sprint with new data; vision doesn’t change
- Multiple sprint failures in similar ways = signal → re-examine the vision itself
- Confusing the two registers is a common failure: sprint goals communicated as vision = team feels perpetually failing; vision communicated as sprint goal = customer thinks you have no ambition
Insights
The framing trick is described as “not semantic” because it actually changes resource allocation, communication strategy, and pivot criteria. This is a non-obvious claim worth examining: the act of labeling something as “research vision” vs “sprint goal” genuinely constrains how people reason about it. Researchers stop over-reacting to single sprint failures when the vision is explicitly separated; engineers stop under-planning when the sprint is explicitly bounded. The observation that senior people at research startups are those who “hold both framings simultaneously” is a useful heuristic for identifying high-ceiling contributors.
Connections
Raw Excerpt
The most useful thing I’ve learned in two years as a Member of Technical Staff at a research startup is that startup research is not a watered-down version of academic research. It’s a different practice with different rules.