Dom said anyone with a phone and natural language would be able to build. He wasn’t kidding.
I’m a builder by trade, not a coder, 109 versions and counting of a fishing app in Australia & NZ just by talking to AI. It runs on-chain on the Internet Computer.
Made with @caffeineai on @dfinity’s internet-computer:native .
Something’s biting. Stay tuned.
#ICP #BuildOnICP
No limitations at all.
The only thing that gave @caffeineai trouble (to begin with) was our “Ask AI.”Every answer had to go from our backend on @dfinity ICP out to Google's Gemini and back, so it lagged, and voice (record → text → think → answer) was slow and glitchy. Approx 20 seconds.
The fix: the phone now talks straight to Gemini Live for voice, so answers stream back in about a second, while the API key stays locked in the backend, which only hands out short-lived passes. Typed answers take a faster single-node path, and common questions are cached.
Claude wrote the fixes in the code, Caffeine built and hosts it.
Been building an app on @caffeineai for a few weeks.
I asked Claude to give me a fair unbiased feedback on using Caffeine and internet-computer:native to build the APP.
Claude: "The biggest downside right now is that it’s new. The tools are less polished than the big names, and Motoko (the backend language) is early, so AI and developers know it less. That’s a maturity thing, and it’ll close fast."
"The upside is once it’s built and running: cheaper to host, secure by design, no servers or databases to babysit."
ICP has no bandwidth fees, so the gap widens as you grow. (AI costs are the same whichever platform you use.)
WE ARE SO EARLY!!
Rough monthly hosting estimate (USD, hosting + database + storage, excluding AI):
@claudeai
When caffeines built in AI couldn’t keep up with the big changes Claude step up as per the following:
Claude switched to building it locally (on my computer) wrote and edited the app's code directly, both the Motoko backend and the React frontend, and saved the files straight into the apps folder on my PC. I’d then run one command, (a copy and paste from Claude) to the Caffeine CLI build script, which compiled the code and uploaded it to Caffeine as a new draft. It was much faster, and much easier to smash out bigger loads of work…….and once uploaded to Caffeine it deploys each draft version perfectly to review / test / and push to live. A really good combination.
Dom said anyone with a phone and natural language would be able to build. He wasn’t kidding.
I’m a builder by trade, not a coder, 109 versions and counting of a fishing app in Australia & NZ just by talking to AI. It runs on-chain on the Internet Computer.
Made with @caffeineai on @dfinity’s internet-computer:native .
Something’s biting. Stay tuned.
#ICP #BuildOnICP
Dom said anyone with a phone and natural language would be able to build. He wasn’t kidding.
I’m a builder by trade, not a coder, 109 versions and counting of a fishing app in Australia & NZ just by talking to AI. It runs on-chain on the Internet Computer.
Made with @caffeineai on @dfinity’s internet-computer:native .
Something’s biting. Stay tuned.
#ICP #BuildOnICP
I guess that leaves only the Internet Computer in the "world computer" game then, building on hardcore technical work since 2014 ;) Cloud engines this week ✌🏼..
@PartyhatsICP@caffeineai The bottleneck isn’t whether the AI / caffeine can build the feature—it’s me having to narrow down my endless list of ideas into a clean, focused app.
After having a proper go building an app on caffeine, I’m now 110% convinced we are on the right track. Everything Dom said Caffeine could and would do it already can (better than I imagined) and Caffeine has only been around approx 12 months! Pretty sure Dom said he’s aiming to be in the top 3 as well 🤔