@theisaacmed Think about it.
Frontier labs need to replace human workflows with their models with 99.9% reliability
How much revenue do they unlock for achieving this? Look at growth TTM.
Theyβve trained on most public data.
IPOs soon. They need an edge.
You have it. For now.
Sell.
@geraldpppppp@halarewich@walden_yan I think to emulate Fusion some modifications need to happen to let the advisor tool call and be more active with corrections - experimenting a bit with this today
@geraldpppppp@halarewich@walden_yan Yes this happened to me a few times actually - I had to manually steer it.
I'm hoping the advisor setup prevents this going forward π
@iambchoor@walden_yan @devindesktop I use it as more of a workhorse - not sure I'd trust it exclusively for ambitious work
you might be able to pair a codex/claude sub with the advisor system in OMP to get a fusion-lite version
in fact, i'll try this when my fable/astra limits reset π
@halarewich@walden_yan seems to kick in around 800-900 requests/hour
ive been able to run ~4-5 concurrent goal agents using OMP harness without hitting rate limits
@iambchoor@walden_yan@devindesktop use SWE-2 only instead of fusion
SWE-2 is only rate limited by requests/min and concurrency
pushing around 2B tokens/day right now through it w/ OMP harness on the $20 plan
@jforjacob ask claude how to detect traffic from ad library (no fbclid), referrer headers, & non-consumers networks/IPs - have it write a script and put it high up in your <head></head> so it loads first and redirects peepers.
claude/codex/etc can insert the code for you using shopify cli