Grainola CEO Chris 讨论了经营一家成功的 AI 初创公司所面临的持续挑战，将其比作一场刀锋之战，生存是常态，并分享了对快速变化的 AI 领域中竞争与创新的见解。
Grainola CEO Chris discusses the relentless challenges of running a successful AI startup, comparing it to a knife fight where survival is constant, and shares insights on competition and innovation in the fast-paced AI landscape.
要点 · TL;DR
创业是刀战，无论成败都艰难。 Startups are knife fights; hard whether working or not.
Granola 专注于会议周边任务，通过 API 为智能体提供上下文。 Granola focuses on meeting-adjacent tasks and provides context for agents via API.
预生成内容是及时 AI 辅助的关键，尽管打开率低。 Pre-generating content is key for timely AI assistance despite low open rates.
核心观点 · Key points
创业就像刀战，无论是否顺利都很艰难。 Startups are knife fights; hard whether working or not.
会议记录并非终极价值；更大的计算革命即将到来。 Meeting notes are not the end-all; a bigger computing revolution is coming.
Granola 的策略：在会议相关任务上做到最好，通过 API/MCP 为智能体提供上下文。 Granola's strategy: be best in meeting-adjacent tasks, provide context for agents via API/MCP.
预生成内容（如会议简报）是提供及时 AI 辅助的关键。 Pre-generating content (e.g., meeting briefs) is key for timely AI assistance.
工作分化为异步委托（Slack）和深度协作（类似 Codex 的界面）。 Work bifurcates into async delegation (Slack) and deep collaboration (Codex-like surfaces).
AI 是肌肉；软件必须是提供结构和骨架的骨骼。 AI is muscles; software must be bones providing structure and context.
反共识 · Contrarian takes
来自大公司（OpenAI、Zoom）的竞争并未影响 Granola 的增长。 Competition from big players (OpenAI, Zoom) hasn't affected Granola's growth.
应取消传统角色（PM、设计师、工程师），采用流动的项目团队。 Kill traditional roles (PM, designer, engineer) for fluid project teams.
即使只有 10% 被打开，也要预生成数百万份简报；成本不是首要考虑。 Pre-generate millions of briefs even if only 10% are opened; cost not primary concern.
用户希望将自己的智能体带入 Granola，而不是使用 Granola 的界面。 Users want to bring their own agent into Granola, not use Granola's UI.
转录稿遗漏细微差别；Granola 应推断文字之外的情感和上下文。 Transcripts miss nuance; Granola should infer sentiment and context beyond words.
适用于多上下文智能体工作的正确 UI 尚未被发明。 The right UI for multi-context agent work hasn't been invented yet.