Automate the browser using a visual AI assistant
though a good product, but i guess it works by reading the html and figuring out the next action it should take with the help of agents. the approach makes it slow as each call has to be done with LLM which will take long time to return an answer. if developers are able to figure it out, (which i believe is going to be highly difficult as response time from LLM is not much in their hands), it will be game changing product.
your recommended use cases does not work, and it keeps crashing my web browser and is very slow.
Sure it's a bit slow, but with some optimizations, I'm sure the developer could make it a bit faster. However, boy just this work a trick! Great job! You must be fairly specific, you can't say "Open my lawyer playlist on youtube", you have to say "Open my playlists on youtube and select the lawyer playlist" or something similar. The former produces the extension finding a public playlist dedicated to lawyers. The latter produces the expected results. Some queries may be a bit too complex, for example, "Find and download the five best studies on the positive effects of nicotine." terminates, stating that the AI is unsure of its next action. This was after leading itself down *a* path, yet not probably the intended or best path. It searched google scholar for the correct query, then selected the top result. This is where it faltered -- It selected the home icon on the study's webpage, the searched for the initial query again, and selected the filter for relevance. Finally, it terminated, stating it was [pp.] unsure of its next move.
https://quicktask.app
xlang-webot
https://checkget.com
Srujan Gurram
https://browsegpt.ai
getmagicformula
hvardhan878
https://aiemploye.com
Layer
nikfedotoff
https://cykel.ai
ralfsommer