I run the free service https://countrx.app/ so i have some idea what goes into counting.
The performance as a general model is indeed really impressive and i think they might actually win compared to fine tuned models.
Their feedback loop of training on user data is incredibly strong.
I've learned that lots of accuracy results depends on threshold configs, which llms should be able to dynamically set.
Or the future will develop in llms using fine-tuned models as tools?
Inference cost and speed does still seem to be below user expectations.
But for being able to one shot with this accuracy... IMPRESSIVE
Oh man, i rented one of those! I fell in love with the west coast national parks. A month of roadtripping to all those parks. Great memories, sad to see the. Not available anymore.
They are speed limited when you floor it in Nevada tho :)
Wife is a lead motion designer, ex Bumble etc. 6+ months of no job results in NYC. Not sure what's going on honestly, all very slow. Any tips welcome :)
https://countrx.app/ is something I vibed in a month. Can people here tell?
Sure the typical gradiënt page is something to spot, but native apps i think are harder. I would love to see app store and Google Play Store stats to see how many new apps are onboarded.
Looking at distribution channels like Google Play, they added significant harder thresholds to be able to publish an app to reduce low quality new apps. Presumably due to gen ai?
Edit: Jesus guys, the point I'm trying to make is that there are probably a lot more out there that are not visible... Im not claiming i developed the holy grail vibe coding.
no the point is that there should be _more_ shovelware like your app. the fact you were able to publish shovelware doesn't mean that there's a "revolution"; the number of apps published per time doesn't seem to be going up.
Half of the images on the iOS App Store have the Gemini watermark on them (and the Google Play Store link is busted). I would assume most people would think this was built with AI.
Yeah I know is awaiting review as i mentioned in the post, the point I'm trying to make is that there are probably a bunch out there already being gen ai apps. I don't think there is a clear way to recognize them
The performance as a general model is indeed really impressive and i think they might actually win compared to fine tuned models.
Their feedback loop of training on user data is incredibly strong. I've learned that lots of accuracy results depends on threshold configs, which llms should be able to dynamically set.
Or the future will develop in llms using fine-tuned models as tools? Inference cost and speed does still seem to be below user expectations.
But for being able to one shot with this accuracy... IMPRESSIVE