Yeah —- the fake it till you make it, ghost-in-the-machine strategy has these as extreme outcomes. AI companies whose devices are in fact human-level intelligent because they are being remote-managed by teleworkers, etc.

Delve (YC W24) was a great example of this going badly wrong really recently, but honestly I think the cavalcade of startups using a fork of someone else's open source wrapper around someone else's cloud AI are not far off.

After all, if you have no real secret sauce at all because your supplier's cloud LLM is basically capable of everything you're claiming as a feature with just the thinnest amount of tool-calling, should anyone be investing in you at all, even just to acquihire?