OKKKK Ive fully automated Codex + Claude Code it take any app idea and runs a fleet of parallel agents in iterative loops nonstop for 24hr+ it can research, design, plan and execute for DAYS zero back and forth it also costs $0 because it’s powered by AI subs... we all already pay for so I could have 100s of complex, polished iOS apps built in 30 days if i really wanted to - for the SAME price as anyone else on the same subscription this is CLONKshow more

Full build Fable-5 vs Clonk Sonnet 4.6

if it’s running nonstop for days, when do we schedule the app’s existential crisis?

Saturday

Where to download

Clonk . Ai

@dilukevin Will it come on windows too bro?

@dilukevin Ofc! When fable 5 comes back I’ll get it to translate the code

I’m very excited to try this

Ay, my DMs always open brother !

24hr unattended is the scary part, how do you keep it from drifting off the actual task halfway? mine starts inventing requirements after a few loops at what point does it flag that we're offtrack codewise or logic wise?

its just good

slop producer 9000

Slop-Reducer 9000

i cant use it on windows??

Not yet

Ever since testing days I haven’t been able to use the kommand because it freezes before the QR code was available to connect to phone scan and clink didn’t function then :/

looks sick man good job !!

@DanielWhit21874 Appreciate the kind words brother

You're welcome brother !!

Clonk also works on Short-Long term horizon tasks like "add a settings page" (dumb asf prompt) it reasons thru the task, understands the app, and ships a prod-ready feature like its got 200IQ again

howd you make the video ;)

Just some light magic

i love opensource projects like these

How to setup?

tokenmaxxing to a whole new level

The 24h loop is impressive, but I’d judge it by the morning artifact: changed files, failed attempts, tests run, open risks, and one decision for the human. If it can’t explain the mess after a day, nonstop just compounds slop.

Is there a risk of a ban?

Same model that t3 code and others use

Thx

how are you handling the state split? every time i spin up parallel agents on the same file, one of them invalidates the other's reads before it posts. been working around it with per-agent locks but curious if you found a cleaner path

running a fleet of agents nonstop for days off subs you already pay for is a clever way to squeeze more value out of it

This is cool

business idea: clonk that works :((

Local + on your own keys is the right call. The hard part everyone hits: keeping a fleet coherent over a 24hr loop. How are you handling state when one agent drifts from the spec the others are working against?

零沟通这点有点吓人，也有点香