[danielsiders]

Cofounder here.

We’re surprised and flattered at the interest and discussion.

For those wondering why we shut down here’s a brief summary:

We created Flynn to be Heroku that you could run on your own infrastructure (cloud or otherwise). We started Flynn as a crowdfunded project with support mainly from other startups, we imagined it would be non-commercial and community-driven.

After the prototype it became clear that we would need a full time team to develop and maintain the project. When we couldn’t find anyone who wanted to fund the project long term we applied to YC and then raised a seed round which let us build a 1.0.

Unfortunately we were never able to raise more VC so we spent several years trying to build a business that would allow us to keep developing the project.

We spent the last four years with a skeleton team and while we were able to build a business doing “ops as a service” for other startups we ultimately couldn’t generate enough revenue to cover both development/features and the 24/7 on call team we needed. While we were able to support our paying customers we couldn’t deliver the features we wanted to build or support unpaid open source users.

We also take stability and support really seriously. We were extremely uncomfortable with the fact that people were running Flynn but didn’t know enough to run it safely. So we’d get panicked emails and github issues from users who couldn’t pay us but needed help. That’s a terrible situation for everyone involved.

The ecosystem has come a long way since we started (we were one of the first Docker powered projects and started about a year before kubernetes was announced). But we still feel like there’s a need for something like Flynn :(

Unfortunately it’s expensive to develop full-featured platforms and we could never get VCs interested in a Series A.

The 24/7 on call lifestyle got to be too much for our team, especially during COVID so with no prospects of things changing we decided to call it quits.

[forgotmypw17]

Just a hypothetical question I want to know as a FOSS dev, but someone who's never used Flynn before:

If someone came to you right now and offered an amount to keep going, no strings attached, enough to support the applications which remain hosted on your platform after those who have already lost their confidence leave, would you consider keeping it running?

What amount, in whatever currency you prefer, would make you consider it?

I've no horse in this race, but I'd like to get at least one datapoint.

This is not the first time I see a post like this, and it always makes me wonder if you could've gotten enough funds if youd gone out and said, "Look, Flynn is going to shut down if we don't get X amount minimum to support us. If this platform means anything to you, now is your chance to save it."

[danielsiders]

Interesting question. I think there are two angles:

1. Having already shut down and the team gone separate ways and onto new projects how much would you need to get back on the horse so to speak

2. Are there still things you want to accomplish and what would it take to make them succeed

Personally I’d happily take a check to do more development. There are a lot of Flynn 2.0 features that were in various stages of completion, if someone was willing to fund development to complete them up front that’s great.

However I wouldn’t want to do sales for this/run a b2b company around this again.

So if one of the companies that was using Flynn called and said they wanted to pay our development team to keep building, that’s great. But if a VC called and said “wait! Let’s finally make this company take off” there would have to be a decent sized bonus check attached. Mostly that’s just because after years on a ramen profitable startup as a founder your finances aren’t awesome so saying “yes” means turning down higher paying offers.

The biggest problem we had financially was being able to develop features before our customers needed them. We knew what to build but didn’t have enough budget to ship them. Any offer would have to come with the guarantee that we could focus on development for day 6 months before starting sales back up.

At the end of the day there’s work we left unfinished and if someone was willing to fund that in advance I’d generally be up for it. But if someone just wanted to keep the business wheels turning that’s not super attractive at this point, so it would need to be a https://levels.fyi salary rather than a startup founder salary.

[forgotmypw17]

Thank you for providing such a detailed answer.

You've gifted a lot of insight.

If you don't mind me asking, what are some features you think are most desirable for you to develop, and what would be your time budget for developing them, if it is more than 6 months?

And which title from levels.fyi do you think is most applicable to the person or people who would be working on that feature?

[danielsiders]

Happy to — it’s a great question! I want all software to be open source but we really haven’t figured out how to fund many types of open source projects yet.

- Database appliances (RDS) for major DBs

- turnkey security, especially for compliance (SOC2, HIPAA, ISO 27001)

- close the loop on development environments (vscode)

- CICD

That’s pretty close to our series A roadmap. Mythical man months aside, I’d say all of that could be production-ready in around 18 months with a few engineers per bullet point.

In terms of salaries it’s a little tricky for a number of reasons (people are willing to work for less for a startup and/or on open source plus we hired around the world), but for many of the developers I’d say equivalent to L4-5.

The important thing to remember about PaaS is multi-node vs single node are different worlds. There are great tools like dokku that were designed only for single node operation, then there are things like Flynn or k8 for many nodes. As soon as you’re doing many nodes you’re in distributed systems territory where engineers have to manage much more complete systems design and as a result are more expensive and in more demand. That’s not saying anything bad about simpler tools, they’re great and important, it’s just a lot harder to design around distributed systems where failure modes, consensus, partitions, etc are a lot more complicated. A lot of the challenges when we started were figuring out what the problems were and how to solve them rather than implementation time. In the last 5 years there has been a lot more written so less research would be needed to implement some of these today.

Our overall goal was to automate anything in the devops lifecycle that could be automated, so that’s a never ending process. Unfortunately it means that the more you scale the more technology you have to make, so Flynn for smaller companies with fewer users is less expensive to design and build than Flynn for huge companies with lots of traffic. (Again, distributed systems are hard) so knowing the scale of your prospective users is a big part of the cost equation.

[forgotmypw17]

For me as a totally uninitiated, the database appliances sounds like the lowest hanging fruit with the fewest unknown unknowns and the biggest payoff, is that correct?

At what you described, for 6 months, that's about 200K USD per dev, so it would cost about 600K USD to develop the database appliances "superfeature"?

Do you think this is a realistic figure? And how realistic do you think it is to get that much money together from all the users?

[danielsiders]

It’s not too far off. Here’s how I’d look at it:

$180k base salary is appropriate but most of those people are expecting stock options as well (compare to Levels) so the fully loaded price* is closer to 300-500k/year x 3 people, so $1-1.5mm

We’ve already done most of the conceptual and architectural work for Postgres, MySQL, and mongo (and redis but not highly available). Much of that was inspired by the Manatee project at Joyent. So (though I haven’t looked at this I a while and the DBs may have changed in a way that becomes more challenging) this is probably a case of put money in, get value out.

However I can assure you you aren’t going to get even 600k out of the community.

We worked really hard to get the first 120k and twisted a lot of wrists to get there. (Of course if you know someone who wants to write 600k checks for open source stuff please have them call me)

The 1-1.5mm number is about what a standard YC company gets at demo day. So that’s the path I’d recommend. Add in another 500k for overhead and biz dev people and that’s a pretty solid startup. Honestly if we had just focused on that we would probably have got a series A.

One of the great things about open source is that you create 100x+ the amount of value that you capture. The dark side is that no one wants to contribute cash as users to make that happen (I’m thinking more of companies than individuals - lots of individuals will give a few dollars)

So, unfortunately, and as always, it’s easier to raise VC than sponsorships for open source.

That being said I strongly encourage anyone who’s interested to pick up where we left off. Either as a community project or a funded startup this would be a huge benefit to the ecosystem.

* of course you can find people who are cheaper and passionate about the project to do it for less, maybe much less. But if someone called me and said for what price can you guarantee this could be done, I’d say 1-1.5mm USD.

[Titanous]

Other cofounder here. I just wanted to add that we still have some t-shirts and stickers left. So if you want some defunct project/startup swag, we'd love to send it to you! https://shop.flynn.io

[jordanthoms]

Still wear my flynn shirt from 2015 when you kindly did a YC practice interview with us! Thanks for that BTW, y'all captured the intensity of it very well :)

[danielsiders]

You’re welcome! Hope you’re doing well!

(For those who don’t know, all the partners in a YC interview ask questions of everyone simultaneously— this is complicated at the best of times and totally nuts when you’re trying to answer technical questions about distributed systems and questions about how you’ll build the business, but it’s a great mental workout!)

[jordanthoms]

Thanks, we are doing well - never did get into YC but it was a valuable process every time we applied - gave us time when we're not occupied with the day-to-day operating to think about our strategy.

We're back in NZ and focusing solely on the education market which really accelerated our growth. Not sure when we'll be able to travel to the US again but would be great to catch up when that happens!

[austinjp]

Out of curiosity, why do that? To see how people function under pressure?

[danielsiders]

I think that’s part of it but it’s also the natural consequence of having several people on one side of the table trying to find out everything they can from the founders in a very short period of time and being genuinely interested if you and what you’re building.

Think speed dating plus lots of coffee.

[tylerlh]

Nice! Picked up a shirt and some stickers :) Sorry to see you guys go. Flynn was one of the first things that made me legitimately excited about the future of Docker. Thanks for all your great work.

[trishankkarthik]

Sorry to hear the bad news, John and Daniel. Thanks for all your great work, especially in promoting TUF. Hope you guys have found greener pastures now.

[forgotmypw17]

I posted a question for the co-founders, it's next to your comment. Would you mind taking a look, please?

https://news.ycombinator.com/item?id=26298288

[yourapostasy]

24x7x365 is an expensive support option, and not all customers need/want it. I found that if I budget staffing for a manager and 7 employees, create "slots" for access at different tiers (segment by business hours in different time zone ranges, response time, and support contract duration), price access to 24x7x365 slots accordingly, the price-first customers will sort themselves out. Good support that performs properly as a function of the continuous development cycle is staffed by highly-technical and high-soft-touch skill sets, and it is expensive to find and retain those people. Don't shortchange them or yourself, and charge the market accordingly.

3-7 years is about the maximum I would expect an unsustainably-structured support organization to last before people burn out, so you had a good run.

[a1o]

Just wanted to say thank you for answering the questions here! You gave a lot of insights! You could be a great mentor!

[anurag]

Sorry to hear about the lack of VC interest. I'm the founder at Render (https://render.com) and would love to chat and compare notes. Email in profile.

[mahathu]

technically illitare here, pardon the stupid question. I hear so much buzz around k8s/docker/AWS/GCP/Azure and all that but simply don't understand what they actually do. if you go with the extra effort to run your product on you own infrastructure, what do you need Heroku/Flynn et al for? isn't the idea of PaaS to take care of server/database/hardware management etc for you, so you can focus on developing your product?

[bg0]

Do you or anyone you know still provide "ops as a service"? Would love to talk to someone who does this.