Wink Pings

The Endpoint of Software Factories May Be a Spreadsheet

In the 2010s, project management tools piled up, but spreadsheets were what actually kept the work in place. Is the software factory taking the same path? Kenn Ejima believes that once Anthropic and OpenAI enter, the real competition will begin.

The term software factory has been very hot lately. On X, Kenn Ejima draws an analogy with 2010s SaaS project management tools: Linear, Asana, Notion, Trello, Monday, and Airtable appeared wave after wave, adding more and more features, yet the market did not converge. The underlying layer that truly held work together has always been the spreadsheet.

This analogy is vivid. Project management tools sell kanban boards, collaboration, and automation, but teams still end up pouring their key lists into spreadsheets. Spreadsheets are not pretty, and their permissions and workflows are crude, but their advantage is that they can be changed at any time. The more tools there are, the more they expose a fact: no single entry point can win everything, and feature bloat will continue indefinitely.

Software factories are repeating this path. As the cost of building a factory comes down, individual developers and small teams will keep launching their own versions. Kenn's judgment is that the most flexible form may still be a bare harness. A harness can be understood as the scaffolding that lets a coding agent run; tools, permissions, context, and loop control are all inside it. Packaging it as a product gives a better experience, but it also sacrifices some modifiability.

More critical is the platform. It is impossible for Anthropic and OpenAI not to touch software factories. When official products come out, users will think: to avoid some inconvenience, I'll just use the official one first. That is when the real competition begins. Individual experiments are meaningful, but one has to accept a reality: in the end, the platform players may take it away. Build light, and don't treat a temporary solution as a moat.

There are two additions in the replies. げま says that having mechanisms like factories, harnesses, and loops is not enough; one also needs the ability to evaluate the mechanisms themselves. Asmir puts it more bluntly: PM tools retain people because tickets live there; factory skills are just md files that can be copied away in a minute. Kevin Lin also mentions that the more tools there are, the less important features are than how to adapt to existing work; pursuing universality will make things complex.

Taken together, these responses show that the hard part of software factories is not building another orchestrator. The hard part is evaluation, migration costs, and the platform's default entry point. Spreadsheets can be the boss behind the scenes because they are flexible enough that anyone can modify them. For software factories to reach that stage, they too must first answer: why can't users leave it?

发布时间: 2026-11-16 00:26