Not all good ideas make good companies, and not all profitable businesses make good companies.

I've been thinking about this in the context of (a) my own company, Transpose Health, (b) the requirements for a venture-backable start-up, and (c) most recently a blog post by Kunle. 

• My company (Transpose Health) is ultimately a services company, and my company can command contracts on the scale of $20-100K -- which seems great, if this is all I wanted to do. If I wanted to hire someone to run the day-to-day (say, $80K minimum with no benefits), I need to make sure I have at least a few contracts lined up -- a lot of work! Compare it to just working for an Epic consulting firm for a "safe" $175K/year and all the work starts to look less and less worth it. The only way this works (i.e. makes me a millionaire) is if there is scale: either an exponential ramp-up in the number of customers, or we expand to SaaS (which we are trying), or both. 
• The same principle applies to start-ups that are venture backable. Start-ups can be solving interesting problems well -- but ultimately the only way to have venture backing make sense is the ability to scale up. 
• Kunle talks about using AI to build personal efficiency tools. The catch: it is almost impossible to build a company around helping individuals with small but powerful efficiency tools. (One example: (1) finding the correct payer entity given only a payer name and (2) building an app to aggregate to-dos across different apps into Apple Notes.) AI-powered coding can help lower the barrier to create these yourself, ... but ultimately, the economics makes fixing these papercut issues hard for any company to tackle. We might be left in a world where we need to build our own tools, the "last mile" of productivity apps.

It's still hard to wrap my head around it: it's not enough for a company to fix interesting problems or be profitable. There are other key ingredients a company needs to be a viable company (including scale).

The depressing corollary is that there are some tools that will never exist as start-ups, except by someone so passionate (and naive) that they build it despite the adverse economics. Perhaps AI will be a way to overcome this hurdle.