I dug into my other single points of failure after the Spacemail outage

Exactly a month ago there was a Spacemail outage, and it took down more than email for anyone on the Spaceship/Spacemail platform. I wrote a blog post about it on a whim, right as it was happening, and it became my most popular post. The day after, I wrote a follow-up.

Read also: Spacemail outage: stay or go?

A month later, I finally had time to sit with something bigger than the outage itself: single points of failure.

I changed careers, becoming a solo indie founder/content creator based in Taiwan, building a handful of small products alongside a newly founded operator-led VC firm, OnPoint VC. That context is relevant, because it's exactly why this problem has been nagging at me since the outage.

I am a single point of failure

Back when I was running a large SaaS with a team, there was always backup. Someone else knew the systems. Someone else could pick up the phone.

Today I'm a solo builder, still hunting for someone to join OnPoint VC. Which means single points of failure hit differently now. I'm one of them.

If I get sick or burn out for a while, I can't just drop everything. Clients might be emailing me, not when there's a critical bug sitting somewhere waiting to become a fire.

I haven't figured out how to replace myself in an emergency, and that's probably its own post. I don't even know yet whether AI or AI agents could realistically cover my first line of defense if things went south. But that's a separate problem from the one I want to talk about here.

The stack is a single point of failure too

Because it's not just me. It's the stack.

Once I combed through what I've built, it came down to a short list:

Vercel, GitHub, Neon, Supabase.

Lean, on purpose. But lean has a cost I hadn't priced in.

The problem is Vercel specifically. Everything I've built now lives there, and Vercel has gone down before. If it goes down again, everything I have goes down with it.

Digging further, I realized I don't have proper automated backups, and I genuinely don't know what fucking happens if I lose access to Vercel.

GitHub holds all my code, sure, but Vercel also holds the environment variables and API keys that make or break several of my builds. Losing that access isn't the same as losing a repo, because I have ZERO clue what the impact would be.

This is a plain admission that I don't have the expertise or experience to build a real 3-2-1 backup setup here. Backing up a codebase and its live infrastructure isn't the same problem as backing up some photos.

So what do I do about it?

"If it's not broken, don't fix it" holds up in a lot of situations. It doesn't hold up when I have no idea whether it's broken or not.

Part of the problem is timing. I just came out of the vibe coding honeymoon phase, where shipping fast mattered more than anything else. That phase is separate from true compliance and security, and I'm only now catching up to that gap.

I'm also not confident enough to grab a live build and move it off Vercel on a whim. I don't even fucking know what the real alternatives are yet, just that they exist.

What I think I can do is start small. Try a Vercel alternative on a new test project first, somewhere nothing is riding on it. Then take an old build, something barely started or half-finished, and see what migrating it actually involves.

I want to try at least one of these this week. It's the same uncomfortable feeling I had when I was weighing whether to move off WordPress completely. That one was uncomfortable too, and I did it anyway, and I ended up much further ahead because of it.

Read also: Leaving WordPress After Twenty Years, in About a Day

There's only one way to find out if this is the same kind of move. I'd rather look at my options now than wait for another outage to force the question. Besides, I want to leave zero room for variance or error.

Because what if I miraculously acquire paying customers in my current flagship SaaS, and it goes to shit, then I am properly cooked and fucked sideways, left and right.

And honestly, this feels like exactly the kind of thing worth documenting as I go, both for myself and for anyone else building solo without a safety net figured out yet.

I have a feeling that once I start pulling on this thread, I'll find more single points of failure than just Vercel. But this is the one that startled me enough to start.