Tigerbeetle is in the database business. They operate on an entire different level of correctness expectations compared to something like Bun. The correctness guarantees they provide come first and foremost from the design, architecture and rigorous testing, not from the language they use for the implementation.

So, hopefully, the tech people involved in choosing a database for their project understand that, and do their own correctness and performance testing before making a decision. As for the business people, the tigerbeetle.com landing page doesn't mention Zig at all, although it probably will come up when they "ask AI" for a comparison. So, yeah, probably some risk. Perhaps the LLM will also point out the Jepsen report on Tigerbeetle to offset it.

> not from the language they use for the implementation.

Everybody has a plan until their compiler compiles an if-else wrong, and that's the level of instability that Zig had when I left it and Tigerbeetle committed to it.

No amount of testing is going to save you from that and it's a completely avoidable unnecessary self-inflicted problem, as a technical person that makes me seriously doubt every other decision you make on top of that. So I feel like they've already been fighting an uphill battle, where they have to proof that they have a stable system, despite an unconventional early stage foundation, and I feel like that it just got even harder.