I sold my DevTool. ft Paul Anthony Williams from ittybit
StreamPot reached Hacker News, paying customers and a steady stream of new signups, then one heavy user exposed how far a single video-processing server was from a scalable product. Jack Bridger explains why he shut it down — and how a friendly competitor bought it for £20,000 in stock options.
This time Paul Anthony Williams, founder of ittybit, turns the microphone around on Jack. StreamPot was Jack's open-source FFmpeg API: a side project that found real demand, became difficult to maintain, and was eventually acquired by the company most naturally placed to serve its users.
They unpack the technical mistakes, the attempted co-founder relationship, whether a deeply technical product should have raised money, what the small acquisition was actually worth, and why treating competitors as future collaborators made the outcome possible.
The Q&A below is experimental — extracted from the episode transcript with AI. If you spot a mistake, let us know at jack@scalingdevtools.com.
Why did you start StreamPot?
While building a podcast-clipping tool, I expected the AI prompting to be difficult but found cloud video processing much harder. The common recommendations involved assembling FFmpeg on Lambda or EC2, while existing services focused on hosting rather than arbitrary processing. There were many people asking the same question and no simple answer, so we built an open-source FFmpeg API with a Fastify server, database and queues, initially running on one Hetzner machine.
What happened after the launch?
The phrase 'open-source FFmpeg API' was catnip for Hacker News. Somebody submitted it, it reached the front page, and a paid plan soon got my first SaaS customer. Even after I stopped promoting it, people kept signing up and paying. Then a user sent a high volume of video through the single server, the queue stopped emptying, and I realized the system could not become reliable quickly enough. I closed signups and eventually moved the workload toward Railway, but the project lost momentum.
Why is video infrastructure so unforgiving?
Video can be orders of magnitude larger than text workloads and exhaust memory quickly, especially in serverless environments. At scale, ordinary inputs reveal endless edge cases: variable frame rates, groups of pictures that are not consistent, a ProRes upload, or a large iPhone recording can block an entire pipeline. ittybit has spent years making its system robust and still encounters formats it has never seen. The minimum viable version of a video product already needs serious engineering.
What went wrong in the co-founder relationship?
We began with a talented young engineer working one day a week as a paid contractor. When he said he wanted to be a founder, I changed the relationship too quickly. We had different expectations about what being a co-founder meant, and turning a working contractor relationship into a weighty, hard-to-reverse title did not fix that mismatch. The lesson is to recognize that co-founder choice has many dimensions and that waiting another three or six months would not have damaged what was working.
Was the failure technical or go-to-market?
The launch was the main marketing action and it worked well enough to keep sending people long afterward. The failure was largely technical. I inherited infrastructure I had not built, tried to learn it while users depended on it, and underestimated the operational gap between provisioning another machine and actually observing and maintaining a scalable service. You need to be honest about the team's strengths and ask what happens if the person who understands a critical system leaves.
Should you have raised money?
Possibly. When a product requires expensive infrastructure and more than one specialist engineer before it can reliably solve the problem, outside capital can be appropriate. I did not feel motivated to build the enormous company venture capital generally expects, and after the co-founder relationship ended I did not want to pursue it alone. Paul argues that a smaller pre-seed or angel round could still have funded the technical work without necessarily committing the company to a billion-dollar outcome.
What made StreamPot useful to ittybit?
It had about five or six paying users, roughly 350 accounts, a waitlist of around 200 and another 30 or 40 people joining that month. The valuable part was not the raw size but the intent: recent signups had explicitly asked for a service that gave developers control over cloud video encoding, which closely matched ittybit's ideal customer. The code contained useful lessons too, and our existing relationship meant Paul understood both the product and the problem.
What did the acquisition look like?
I was drafting shutdown emails when I messaged Paul to ask whether there was a better outcome. We spoke that evening and eventually agreed on £20,000 — roughly $25,000 at the time — in ittybit stock options. Paul compared the value of the active, well-matched audience with what the company might spend sponsoring a podcast, YouTuber or other creator to reach similar people. For a product I otherwise planned to close, it felt fair.
Why be friendly with competitors?
Adjacent companies usually have more in common than not, and today's competitor may later acquire you, be acquired by you, or simply be the right home for a customer you cannot serve. Paul openly recommends alternatives when ittybit is not a fit because those customers would churn anyway and honesty builds trust. If a market is so small that every relationship is zero-sum, it may not support a large company in the first place. The relationship we built before StreamPot closed is what made this deal possible.
Links & references
- Jack Bridger ↗
- StreamPot ↗
- StreamPot on GitHub ↗
- StreamPot closure announcement ↗
- Paul Anthony Williams ↗
- ittybit ↗
- FFmpeg ↗
- Hetzner ↗
This episode is brought to you by WorkOS.