Hey! Here's a feature I haven't seen in any coding agent or IDE so far and I think it could be a real differentiator for Sinew.
The idea: run two AI models in parallel on the same task (e.g. Claude + GPT auditing the same codebase), then either:
Option A — Agentic debate loop: both models share findings and challenge each other iteratively until consensus (with a turn limit to avoid infinite loops)
Option B — 3rd-agent comparator: both models run independently, a third agent merges and synthesizes the two outputs
Why it matters:
I've personally run audits on the same codebase with Claude and GPT separately, they consistently caught different issues. Combining them would produce a far more thorough output. It also reduces hallucinations significantly: one model can flag what the other missed or got wrong.
On token cost: yes, this runs two providers simultaneously. But for full audits, critical feature implementations or complex architecture decisions, many users would willingly pay the premium to get a result that doesn't need 3 re-iterations. Sort of a "power mode" toggle or smthg.
I don’t really know how you can manage to do that (ur the agentic engineer in the room lol) and I’m not sure how it works technically. It’s a bit too deep for me with my current knowledge, but I’d love to hear your thoughts on this kind of idea around AI debating stuff.
This feature could make alot of people switch If you find a way to make it work effectively.
Open to talk about it tho.
Hey! Here's a feature I haven't seen in any coding agent or IDE so far and I think it could be a real differentiator for Sinew.
The idea: run two AI models in parallel on the same task (e.g. Claude + GPT auditing the same codebase), then either:
Option A — Agentic debate loop: both models share findings and challenge each other iteratively until consensus (with a turn limit to avoid infinite loops)
Option B — 3rd-agent comparator: both models run independently, a third agent merges and synthesizes the two outputs
Why it matters:
I've personally run audits on the same codebase with Claude and GPT separately, they consistently caught different issues. Combining them would produce a far more thorough output. It also reduces hallucinations significantly: one model can flag what the other missed or got wrong.
On token cost: yes, this runs two providers simultaneously. But for full audits, critical feature implementations or complex architecture decisions, many users would willingly pay the premium to get a result that doesn't need 3 re-iterations. Sort of a "power mode" toggle or smthg.
I don’t really know how you can manage to do that (ur the agentic engineer in the room lol) and I’m not sure how it works technically. It’s a bit too deep for me with my current knowledge, but I’d love to hear your thoughts on this kind of idea around AI debating stuff.
This feature could make alot of people switch If you find a way to make it work effectively.
Open to talk about it tho.