Cory Doctorow's enshittification model of platform development can be conceptualized as a repeated two-step move:

• Companies reward you for doing things that increase your switching costs, even at a loss for them.
• Then they reconfigure themselves to take advantage of those high switching costs to get back every cent they spent and more. Much more.

There's multiple groups of people on your platform you can do this to: users, advertisers, sellers of goods or labor, contributors, etc. It's a very scale-friendly model; the more groups you can extract a margin from, the more leeway you have to get new groups onboard. It's also the business model of drug cartels, loan sharks, and other industries this post would get too long going over.

It's also central to much of the investment world's view of what a great startup looks like. Spending money at a loss to scale and then monetizing the heck out of it has two ways of working. Either you spent the money making the product better — which would result on the product, platform, or service being more enjoyable and beloved as time goes by — or you spent the money on the bait-and-switch (or "embrace, extend, and extinguish") with your users, customers, etc — resulting in an inverted-U pattern of satisfaction with the platform even if it continues growing.

An useful way to model this from a financial point of view is to remember that if you participate in any algorithmic platform the return of your capex is very much under control of the platform in ways that are by design opaque to you and much more flexible than you think. High returns at the beginning aren't an indicator of future performance but rather deliberately subsidized. It's a front-loading strategy to first get lock-in through investment in money, time, or specialization, and then maximize value extraction. You never have a good business on a platform. At best, you are allowed to have one to entice you or others to keep investing.

The accounting rule of thumb should be:

• Every cent you get from a platform is one-time revenue. Revenue is "recurring" at the whim of an organization with more data, deeper pockets, and the strategic high ground. Don't overestimate your power.
• Every cent, hour, or idea you put in a platform gets depreciated at once.

There are counterexamples. Very well-known, very visible, very few counterexamples. One of the features that makes social networks so good at this sort of model is that they only need to work well for a few people for a short bit for a lot of people to see it working a lot of time.

At a systemic scale there's probably things to be worked on in terms of regulatory frameworks. For individuals the practical approach is to think of any specific platform — be it a social network, a gig platform, a marketplace, whatever — as a highly speculative investment, not… well, not a platform. It's a platform for the owner's business model, not for you (or for anybody else for that matter; if your strategic analysis investing in a platform is that it helps you take advantage of an already locked-in group, you're forgetting that the trick in most cons is to make the mark believe they are pulling a fast one on somebody else).

Whatever sort of technological or reputational capital you build has to exist outside, because on the platform it is whatever the platform managers want it to be this week. Hence the vital importance of things like personal websites and email addresses, both yours and having other people's. In the simplest terms, having a Twitter subscriber or being highly-ranked on LinkedIn is an ephemeral sort of parasocial capital: you might have worked hard to convince people that you were worth paying attention to, and they might have chosen to, and yet whether that channel can work, or what you can do with that, depends on neither of you. Having somebody's email is very different from having somebody follow you on Twitter, Substack, or wherever: you can email them, and Elon Musk can't do anything about that. It's not a perfect substitute for a direct relationship, but it's orders of magnitude less susceptible to hyperfast reparametrization to maximize somebody else's profits. Being in someone's DMs might be more fashionable than being in their email inbox, but it's not the same.

The same goes, in the other direction, for having your own website and selling or saying whatever you want to sell or say through it. Other channels can have larger reach and easier tools, and it's tactically useful to take advantage of them, but unless you control your core infrastructure sooner or later you'll find yourself racing to the bottom when the platform looks at you and decides the time is right to move on to phase two. The great benefit of the basic protocols of the Internet, from TCP/IP and DNS to email and URLs, is that they kept switching costs low enough that the Internet could build much more value than it would have if somebody would have been able to lock it in along the way.