June 20, 2025
Dan, Rose, Greg, Kirsten
Last week, Kirsten shared her use of Lovable to make a voice-based ai assisted data collection component for Good Agriculture. So this week, Rose also tested it out for a application for the Bionutrient Food Association!
Rose
- We used 5 tokens… probably not enough
- Neat, because you can publish it
- Did a security review… it’s AI slop, but a lot of it worked!
Kirsten said
- if I were to make a production version off Lovable, start from scratch matching tech stack with what AI actually is best at (probably React).
- Right now, most useful to generate screenshots that are pretty accurate to reality.
Greg asked
- We are thinking about actually creating a testable (meaning, integrated with a database and usable by 10 people) so that our partners can test their assumptions. Could we actually do this?
Kristen replied
- Yes, probably, but need to learn more about Supabase probably.
- Also, people are fundraising (successfully!) on Lovable prototypes… so that’s got to mean you can produce something pretty significant.
Dan said
- well that’s great! It would really help if we could do that, because it’s quite clear that the BFA themselves have many disagreements about their own workflow.
- So being able to have them test and learn on a cheap prototype is pretty awesome.
Greg asked to Kirsten
- have you tried to just use Gemeni 2.5 or similar instead of Lovable?
- also, have you tried to create something in Lovable, push to gitlab, make changes in gitlab and then push things back to Lovable and continue to code there? Is it really ‘two way’ in that sense?
Kirsten
- Nope - not yet… maybe we should try that with our existing tech stack. Maybe we could then get useful code out of it rather than just a nice tool for making screenshots.
- No, so far haven’t really connected git and done the 2 way functionality described above.
- Also… don’t know what a token is in Lovable
Rose
- A token (it says) that one query is a credit no matter what (Kirsten was skeptical of that… didn’t seem that way).
Dan talked about fylogenesis.com, which he’s been testing with the Astera Foundation. Pretty neat - all about auto-generation of knowledge graphs from pdf’s and other sources. We gave all kinds of feedback about it - very fun!
Next meeting is July 4th… so we’re skipping that one But **Please join us after that and share / learn about this fast moving space!!!