Hello Product Hunt!
Tech lead at Lokuma right here.
The core shift in v2.0: we changed the mounted “generate-once” pipeline with an actual agent loop — the type plans, writes code, inspects output, and self-corrects till it is glad. That is what makes the standard bounce really feel so dramatic.
A couple of issues beneath the hood we are pleased with:
1.Plan-first — the agent drafts a construct plan ahead of writing any code. Customers approve it ahead of execution begins.
2.Centered patching — edits use find-and-replace at the explicit modified segment, now not complete document rewrites. Sooner and extra predictable.
3.Reside preview — updates because the agent works, now not simply on the finish.
We constructed this as a result of we had been bored with AI developers that glance nice in demos however wreck the instant you customise. Lokuma 2.0 is supposed to be an actual collaborative builder.
Glad to reply to questions concerning the structure — having a look ahead in your comments!
