The FM <-> platform analogy is a great one, and one I haven't heard before. These types of "big" ideas are very seductive and capture the imagination of VCs and philanthropic orgs alike. My feeling is that we will not escape from this vicious cycle until the well runs dry, since it's kind of embarrassing for all involved to admit mistakes were made.
Excellent section on FMs which could probably be a post in of its own. One comment I would make is that many of the issues mentioned apply to any complex modeling infrastructure in EO they are just more apparent with FMs because of the scale. Many start ups in the space have taken a technology-first approach rather than a product-first approach and have ended up saddled with expensive technical debt and eventually sold for pennies on the dollar.
Hey Chris, thanks for the comment. Agreed, a lot of this isn’t FM-specific; we’ve seen the same pattern in other EO-based products. Beyond scale, I'd argue that FMs make the risks more visible, because the level of abstraction and the cost of iteration are so high, both in terms of financial investment and a form of technical debt.
I put them front and center in this piece because in 2025 FMs reached a kind of peak moment. You're right though, it could easily expand into a stand-alone piece.
Your "scale is a trap" comment has stuck with me, really.
Since then I linked this urge to build "one model for everyone and everywhere" to the old "one platform for everything" phase and the more I think about it the more I struggle to see how that becomes sustainable in the long run.
And that's before we even get to questions about equity and who can afford to build these models and infrastructure to support them.
The FM <-> platform analogy is a great one, and one I haven't heard before. These types of "big" ideas are very seductive and capture the imagination of VCs and philanthropic orgs alike. My feeling is that we will not escape from this vicious cycle until the well runs dry, since it's kind of embarrassing for all involved to admit mistakes were made.
Excellent section on FMs which could probably be a post in of its own. One comment I would make is that many of the issues mentioned apply to any complex modeling infrastructure in EO they are just more apparent with FMs because of the scale. Many start ups in the space have taken a technology-first approach rather than a product-first approach and have ended up saddled with expensive technical debt and eventually sold for pennies on the dollar.
Hey Chris, thanks for the comment. Agreed, a lot of this isn’t FM-specific; we’ve seen the same pattern in other EO-based products. Beyond scale, I'd argue that FMs make the risks more visible, because the level of abstraction and the cost of iteration are so high, both in terms of financial investment and a form of technical debt.
I put them front and center in this piece because in 2025 FMs reached a kind of peak moment. You're right though, it could easily expand into a stand-alone piece.
Your "scale is a trap" comment has stuck with me, really.
Since then I linked this urge to build "one model for everyone and everywhere" to the old "one platform for everything" phase and the more I think about it the more I struggle to see how that becomes sustainable in the long run.
And that's before we even get to questions about equity and who can afford to build these models and infrastructure to support them.