@vybhavbhadri 2, and im slowly starving them out. the only TODOs i let stay are the ones with an owner and a date, otherwise they become permanent furniture :)
@darrenmarble been there, and the look on their face is unforgettable. once i started leading with the job-to-be-done and the buyer, the rest of the pitch finally had something to attach to.
you can build ai agents, but if your systems and data are a mess you end up automating chaos, so step 0 is getting the plumbing right.
what's your step 0 before you let agents touch anything?
wrapped up the final .env.example for ragentis. it's onboarding disguised as config.
its basically a mini spec now:
what's allowed, what's recommended in dev, what's recommended in prod.
if a new dev cloned your repo today, what's the first env var they'd have to ask you about?
when i'm about to subscribe to a tiny SaaS, i check the team.
not because i care where they went to school.
just to see if there's at least one engineer.
it's not even a "vibe coding is bad" take. i like vibe engineering, we all use shortcuts.
but if the product is shipped by someone who doesn't really understand the code they're shipping, i get nervous.
@irbaazkadri yep. in my experience good devs use ai to compress the boring parts, tests, docs, refactors, and spend the saved time on understanding and tradeoffs. the ones skipping that part just accumulate debt at ai speed.
@joherkhan agreed, the direction feels like infra. the only wedge i see for smaller teams is being closer to a specific job-to-be-done than the carrier.
i don't know the full context, but i know the product.
dify is genuinely solid. i use it daily as the ai backend for chatflow + workflow apps, and i'm building an agent strategy plugin for it.
best part is you can pick the abstraction level, no-code, low-code, then drop to code when you need it.
really respect what the team is building.