I built Bona Fide: refundable good-faith deposits for open-source pull requests.
Maintainers are drowning in AI-generated PRs. Instead of closing the repo to strangers, ask a newcomer for 1 USDC. Merged → they get it all back.
Live on @arc mainnet: bofi.live
Repo: github.com/tonymorony/bona-f…
It has an agent layer: Claude reads each PR that holds a Deposit and tells the Maintainer to accept it, or drafts a bad-faith verdict to sign.
It never decides. It can hold a waiver or a Bounty payout back, but only code checks and signatures move money.
What @arc gave us:
- gas is USDC, so a Deposit needs nothing else
- EIP-3009: a Contributor signs once, our App relays it, no gas for them
- Forfeits go through CCTP to Protocol Guild on Base
- the escrow is immutable: no owner, no upgrade
AI made pull requests free to open. Reviewing them still costs maintainers hours.
Bona Fide, built at the @thecanteenapp Tameion Agents Hackathon, is live on Arc mainnet: newcomers back their PR with a refundable 1 USDC Deposit.
bofi.livegithub.com/tonymorony/bona-f…
Took me just a few prompts to create a native Swift controller (for personal use) for Spotify to bring back those Winamp vibes while I work and listen to music.
Well done, @AnthropicAI. You’re like Santa Claus - making any wish come true 🎅
I can't help thinking about this these days.
It is this feeling of soullessness.
For many years, I have worked on large products: 9-figure active users, millions in revenue, creating new categories, strong adoption, businesses loving what we built, targets met. By most conventional measures, they were successful products. And personally, I have a good life thanks to what my teams and I built.
But over the years, a feeling has been growing: disconnection.
There were many days when I opened the products we had built, looked at them, and felt almost nothing. Sometimes I even felt that they were no longer the things I wanted to build.
It was a classic problem of building products: trade-offs.
Once a business metric becomes the priority, many other things become easier to compromise. In most of those cases, I met the business objectives, but lost some of my faith along the way.
For a long time, I thought the problem was me. Maybe I wasn't good enough at making those trade-offs. Maybe better execution could have given us both.
Earlier this year, I thought that situation would get better with agentic coding. As writing code became cheaper, maybe we wouldn't have to make so many compromises. Maybe, with good enough execution, we could finally have the best of both worlds.
The reality hit me hard.
A few months ago, my partner and I realized that we had started to feel empty even during the process of building. We were losing the joy of writing good programs. We had a problem with what we were building, and now we also had a problem with how we were building it. Both the destination and the journey had become soulless.
It felt like a kind of existential crisis.
Today, what @mitchellh wrote pulled us back from that crisis.
It was a reminder of what we wanted, and that we still want it. More importantly, it reminded us of what makes us excited to wake up in the morning.
The joy is still there. Maybe we just drifted too far away from it. Maybe somewhere along the way, we lost faith and started treating compromise as the default.
It will take some time to rearrange things. But I know what I want to get back to. It is the joy of building the next piece of software we genuinely love.
Building software should be fun.
Software should be fun. It should make you feel something, positive hopefully. Utilitarian software is the worst. The human experience is why life is worth living (even for something as teeny tiny as a terminal).
To put it in today's terms: imagine for a second that an AI provider's system prompt makes its models start producing well-masked zero-day vulnerabilities. What recourse would we have in that case, other than trusting the vendor? Clean intent + dirty model = dirty code - the similar shape as Thompson's Clean source + dirty compiler.
I want to share my favourite computer science paper, the one that made the biggest impression on me. Only three pages, but such a deep insight.
I think it's more relevant than ever nowadays, with all these AI capabilities. It's amazing that this was raised back in 1984. No wonder Mr. Thompson earned his Turing Award!
Reflections on Trusting Trust:
cs.cmu.edu/~rdriley/487/pape…