Designer & developer. Helping improve bitcoin design with many others at https://bitcoin.design . I write a weekly update at https://gbks.substack.com . ✌️
Notes (20)
AI sucks at time estimation, maybe in an opposite way that people do. Says 11-14 hours, and is done in 10 minutes 🤷.
Arké now has a (very rough) mobile app that is best friends forever with its desktop companion. Thanks to new nostr:nprofile1qyxhwumn8ghj7mn0wvhxcmmvqy28wumn8ghj7un9d3shjtnyv9kh2uewd9hsqg8dktefjxjev6wjyyy8xlgmh0ktraj6jtfy5zmy7pd52472qef9dy5cmhc8 Bark FFI Swift bindings. Here's a shaky setup-and-link demo:
https://www.youtube.com/watch?v=AVLt4o_GdpQ
What do you think?
This is an example with a rich BIP-21 URI (like what you might get from clicking a link at a checkout)). Instantly parsed and evaluated. Payment meta data is easy to see. Payment options (addresses) are matched against existing contacts, networks verified, balances checked, fees evaluated, and the best option highlighted. Just hit send and done. What do you make of that?
I recorded a hands-on 30-minute demo on how I use AI for coding for Arké (MacOS app, XCode). Curious to learn from others, if you have any tips. https://www.youtube.com/watch?v=rcxkYZTtOTM
Arké update 4, about native contacts, contact auto-assignment, a new console, and lots of little layout and convenience things. Still a major construction site, still progress. https://community.second.tech/t/arke-macos-prototype/156/9
I saw some people comment on the BIP-177 by default thing. Something I plan to get around to at some point and add options to choose from. Just not priority right now.
Arké now also has a consolé. Very handy for doing some manual operations during development. Took 3 chats to build this with Claude in Xcode, mostly because I knew exactly what parts needed to be touched and also set up just a little bit of the skeleton to give Claude a more narrow scope. Such a breeze to build with.
As I've been working on the Ark prototype, I tried to feel out the economic viability of the system, which can be a make or break. One part of is that every ASP is its own separate microcosm. That seems manageable for bitcoin payments, but adds exponential complexity for the advanced stuff in Arkade (lending, swaps, etc). And it seems like a centralizing force that will put intense pressure on ASPs and also invite regulatory scrutiny.
- Stablecoins are locked into individual ASPs - how many of those will Tether/etc support?
- Will ASPs need to provide liquidity for non-bitcoin assets also?
- What if a liquidation cascade depletes ASP liquidity? What if rounds are too slow?
- Is there unilateral exit for non-bitcoin assets?
- If your app only implements the payments part, are you still exposed to the other dynamics?
Do I have an answer, or understand the complexities of all this? Certainly not. If you do, congrats, please share. Intuitively, I prefer simpler systems. How do you think about these things?
In my Ark prototype, I added this super handy X-Ray screen. Let's you inspect all the underlying data that drives the wallet. Looks like I need to refresh my VTXOs...
I tinkered with an Ark wallet prototype (second.tech implementation) to get a feel for the squishy parts of the protocol. Here's a demo video and some questions that came up. Maybe some of you have answers? https://community.second.tech/t/arke-macos-prototype/156
Big thing I appreciate about AI assisted coding is that you can keep your brain focused on the architectural level. It really takes care of the small stuff.
Pretty amazing what you can do in a day with €0.5 worth of Claude Sonnet 4 credits via Github Copilot in Visual Studio Code.
Man do I wish that these workflows were also available for design tasks.