[jeroenhd]

Building apps for other platforms, be they Shoppify or the App Store, always comes with the risk of getting booted off or Sherlocked.

You can pray that the platform's customers will be upset if your product gets killed ("Shoppify broke our checkout screen" is not exactly bringing in new users) but in the end you need an exit strategy as a company.

In this case, I do think legal action would've succeeded, but it would probably be a long, expensive, painful lawsuit, that's probably too much for this company to bear. You're also not guaranteed that the judge will make the losing party pay for your defence even if you do win, so it could easily be quite a Pyrrhic victory.

With the new DMA coming into effect soon, I think businesses like these will stand a much better chance. The restrictions put onto gatekeepers by the EU can introduce significant risks to platforms being scummy to smaller developers.

[caycep]

One question in my mind is - how does TSMC structure their "platform" such that not-sherlocking-customers is incentivized into their business model?

[dtech]

They don't have to. Building a semiconductor fab is such an enormous investment and requires so much specific knowledge that it's not an attractive option, orders of magnitude harder than copying some software.

A similar question would be how Airbus prevents airlines from building their own airplanes.

[politician]

As I understand it from my time at AMD, TSMC provides what is essentially an SDK for specifying how to use their process for a given node. Clients translate (compile) their higher level specifications into this framework for fabrication, and own responsibility for characterizing (testing) the resulting silicon.

This separation of responsibilities presents TSMC with a difficult reverse engineering challenge if they were interested in violating their NDA.

[philipkglass]

I don't think that TSMC is really in a position to Sherlock their customers:

- Buyers of leading edge fab services have patented features in their chip designs which TSMC would violate by simple cloning.

- If TSMC wanted to profitably manufacture something like "a modern GPU, but avoiding existing patents" they'd have to expand massively beyond their current core competency of fabrication. They'd need experienced chip designers, software engineers for developing drivers and APIs, retail partners... it's a much different business.

The key thing that makes "Sherlocking" easy is that it's just imitating third party software with first party software, from a first party that is already good at making software.

[chiefalchemist]

> "but in the end you need an exit strategy as a company."

Simple. Reach out to the platform and ask to be acquired. You might not get top dollar, but it beats having the plug pulled. While he acknowledges he didn't expect the gravy to be pouring forever, he also didn't have an exit mapped. Exit is one of those things that is in the text book definition of what is an entrepreneur.