Now that fly are pivoting, should I trust them for hosting?

Doesn’t sound like it

To be blunt, my experience is that you shouldn’t be trusting them as your host to begin with. We moved our app off Fly to AWS Fargate earlier this year. We were so tired of Fly incidents causing issues for our customers. We wasted so much time debugging intermittent cluster issues. Their Redis integration with Upstash has severe performance regressions on a regular basis. Fly is fine for to get started or for your hobby app. It’s not fit for heavy production use.

My experience with them taught me a very valuable lesson. The real value AWS provides is rock solid networking. Everything Just Works. And, as the article pointed out, tell your LLM of choice to write some IaC to spin up your environment. The barrier to entry is mostly gone. The PaaS moat has mostly been filled in.

Now, all that being said, I’m going to go try sprites as a hosting option for low effort, unimportant, vibe coded internal tooling.

Have to agree here: Fly has been a huge disappointment in hosting. I’ve had 1 decent support email over the 10 or so I’ve had to raise over a few years cause they cannot keep their infrastructure running properly.

Howdy

Disclaimer: I work there

Actually machines API is too part of Compute for Agents. We are seeing an increasing number of folks running some agent that will look for markdown docs, launch one or more apps and figure out everything our machines API can do so that’s not going to go away anytime soon (we host our stuff on our PaaS too).

Sprites too are part of this since their sandbox like qualities are good for stuff like running the agent itself, hosting some automations like ntfy or even serving as a bastion server.

Happy to clarify anything further

thanks for the link.

I read it and i’m not concerned (i’m heavily dependent on their service btw). because they are making a new startup (or new product rather), with new funding. Kurt is stepping down as CEO and a new guys is being hired. all in all this is the natural way startups work generally. as far as i know how money works the servers business will sooner be sold off rather than sunsetted or broken. like if the current core offer of fly would need some major improvements then i would be pissed. but imo it’s working great, does exactly what is needed. it doesn’t need the founder imagining new ideas anymore. in fact i hope the new CEO just keeps **** running and not bringing their own brand new ideas to the table

i don’t see a reason to worry and Sprites is a great idea. gonna go try it now

only worrying note is at the very end. so Kurt will not be leading the Sprites team after all? but still, whatever. still not a problem.

Thanks for the replies everyone.

I was particularly interested in insider knowledge, but it’s useful to see public opinion too.

I partly agree, they aren’t as reliable as self-built AWS, but we’re a small product/eng team, I don’t want to spend a second on infra that I don’t have to. For us over the last few years the basic product has been reliable, we haven’t gone deep into the networking or additional services side.

That said, I’ve not seen anything here or elsewhere that indicates that Machines/hosting is actually safe long-term. (@lubien I really appreciate your response and logic, but it’s telling that no-one from Fly can officially commit to stability of machines, it’s all inferred).

I’m going to cautiously move us onto something else that’s more stable for 3-year plans. No rush, but maybe by the end of this year. for the help.

If an exec word helps we had a very similar question in our community forum about machines and stability:

Or Fly.io on X: "Turns out when you announce a big bet on "Computers for Agents," people wonder what will happen to Fly Machines. Fair. So, plainly: Machines are not winding down. There's no sunset date, and they're important enough that there's now a team whose entire job is keeping the" / X**
**

Neither public opinion nor insider knowledge matters in the end. Show me a financial report that indicates the machine product line is making money then that’s talking.

@lubien That does help a lot, thankyou.

Relevant points:

• The action of forming a new reliability team.
• The explanation of machines + sprites being very intertwined, explains a little of the backstory.

Given that, I’ll cautiously see how things go for a while.