Hey PH, Omri here - been lurking and commenting for a bit, figured I'd actually say hi
by•
I spend most of my day running a handful of Claude Code sessions in parallel, so a lot of what I comment on here is basically me pressure-testing whether a new tool actually survives contact with a messy multi-agent workflow, or if it only looks good in the demo.
Not building anything of my own to launch right now, mostly just here for the discussions and to poke holes in the "10x faster" claims before I install something. Genuinely enjoying the vibecoding and AI agent threads especially, there's a lot of "it worked in the demo, then it fell over" stories in there that are way more useful than the marketing copy.
Happy to be here, looking forward to more of these threads.
8 views
Replies
Hey Omri, welcome properly to PH 👋
I really relate to the “worked in the demo, fell over in the real workflow” part. With AI agents, the hard test is usually messy context, multiple sessions, unclear state, and real project constraints.
The best discussions here are often from people pressure-testing tools beyond the marketing claims, so looking forward to your comments in the AI agent and vibecoding threads.
@prashant_patil14 thanks for the welcome. yeah messy context across sessions is the real killer, way more than any single model capability gap. i'll keep poking at it in vibecoding, appreciate you stopping by