Find what changed.Discover what matters.Build what becomes possible.
发现什么正在改变。找到真正重要的问题。构建因此成为可能的公司。
Singlebase is a venture studio built to discover important problems whose solution space has been fundamentally changed by technology.
Singlebase 是一个 Venture Studio。
我们寻找那些因为技术变迁，其解决空间被根本改变的重要问题。
How we work
我们如何工作
AI has made building cheap, but not knowing what is worth building. We turn uncertainty into evidence, and evidence into companies.
AI 让构建变得便宜，却没有让“知道该构建什么”变得更容易。我们把不确定性变成证据，把证据变成公司。
- Signal信号
- Intelligence智能
- Decision决策
- Action行动
- Proprietary data专有数据
Proprietary data sharpens the next signal. Time should make the product smarter.
专有数据让下一次信号更清晰。时间应该让产品更聪明。
How we decide
我们如何做决定
Principles that bind the studio itself, for when projects, capital and preferences conflict.
这些原则约束工作室自己：当项目、资本与偏好发生冲突时，用它们做决定。
Start from the problem, not from AI
从问题出发，而非从 AI 出发
We look for expensive problems that are newly solvable. AI is the break point; the problem is the starting point.
寻找仍未被解决、且现在第一次可能解决的昂贵问题。AI 是断裂点，问题才是起点。
Build companies, not features
做公司，不做功能
If foundation models get 10x better and 10x cheaper, does the company get stronger or disappear? We build the former.
如果基础模型变得 10 倍更好、10 倍更便宜，这家公司会变强还是消失？我们只做前者。
Falsify first
先求证伪
Every project starts with a pre-mortem. Experiments attack the likeliest cause of death before any code is written.
每个项目先做预演失败，实验首先攻击最可能的死亡原因，先证伪，再写代码。
Kill cheaply
低成本淘汰
Failure is expected; expensive failure is not. We measure the time and capital it takes to reach the truth.
失败在预期之内，昂贵的失败不是。我们衡量的是抵达真相所需的时间与资本。
Every project leaves something behind
每个项目都必须留下东西
Data structures, evaluations, research methods, relationships. Project N+1 should be faster and cheaper because project N existed.
数据结构、评估系统、研究方法、合作关系。第 N+1 个项目应因第 N 个而更快、更便宜。
The studio builds, others scale
工作室建造，他人规模化
Once validated, a company gets a founder team of its own and outside capital. We keep enough ownership to share in the upside, and leave enough for others to build something great.
项目通过验证后，由独立创始团队与外部资本规模化。我们保留足够的所有权，也为他人留出空间。
Evidence earns capital
证据赢得资本
Capital is released in stages. Each stage retires one key uncertainty. Enthusiasm does not unlock the next dollar; evidence does.
资本分阶段释放，每一阶段消除一个关键不确定性。不是热情，而是证据解锁下一美元。
- Problem问题
- Explore探索
- Validate验证
- Spin out分拆
- Scale规模化
Where we hunt now
当前狩猎主题
Can signals that were unobserved, underused or fragmented become a compounding intelligence loop? We are especially interested in longitudinal intelligence. It is a current theme, not the studio’s identity.
过去未被观察、未被充分利用或碎片化的信号，能否变成持续复利的智能循环？我们尤其关注纵向智能，但它只是当前主题，不是工作室的永久身份。
China can be a lab, not a boundary. Geography serves the opportunity, not the other way around.
中国可以是实验室，而非边界。地理服务于机会，而不是机会服务于地理。
When principles collide
当原则冲突时
- Narrative versus evidence: choose evidence.
- Sunk cost versus truth: choose truth.
- Short-term revenue versus long-term ownership: protect ownership.
- 叙事与证据冲突：选择证据。
- 沉没成本与真相冲突：选择真相。
- 短期收入与长期所有权冲突：保护所有权。
Work with us
与我们合作
Founders, operators, domain experts: tell us about an expensive problem.
创始人、运营者、领域专家：欢迎与我们分享一个昂贵的问题。
Want to join the build? See the open role.
想加入我们？查看开放的角色。