Has anyone ventured into automating tasks using vision AI?

tidderautomate

Junior Member
Joined
Dec 4, 2024
Messages
128
Reaction score
59
I'm considering using tools like AskUI or Browser-Use to integrate into automation systems that act unpredictably. The idea is for the bot to interact with a Reddit account in a random manner, responding based on the current page's content. By combining this with models like OpenAI's o1 or other reasoning models, the goal is to create a neural bot capable of browsing Reddit as naturally as possible.
 
Reddit hates automation, but if you make it mimic real scrolling patterns and engagement, you might be onto something
 
Reddit hates automation, but if you make it mimic real scrolling patterns and engagement, you might be onto something
I have already had success with my warmup bot mimicking human interaction. I'm now advancing to the next level by adding a persona to each account, giving them a form of memory.
 
Three months ago, I had to rebuild my bot after the increase in Reddit bans, and only after two months of intense work (I can say more than 12 hours a day), I managed to get the first results. Today, I can say that my bot is complete.
I requested three quotes to fix my bot from people more experienced than me, who might have taken even less time (their estimates were also around three months), and the quotes ranged from 7K to 10K.
 
Three months ago, I had to rebuild my bot after the increase in Reddit bans, and only after two months of intense work (I can say more than 12 hours a day), I managed to get the first results. Today, I can say that my bot is complete.
I requested three quotes to fix my bot from people more experienced than me, who might have taken even less time (their estimates were also around three months), and the quotes ranged from 7K to 10K.
My boss is investing in this, which is why I continue researching how to improve the bot and automation. As Reddit is also strengthening its security, we really need to avoid complacency and keep updating our bots.
 
My boss is investing in this, which is why I continue researching how to improve the bot and automation. As Reddit is also strengthening its security, we really need to avoid complacency and keep updating our bots.
Tell your boss that before investing in the bot, they should buy proxies because without them, you won't get anywhere!
 
Tell your boss that before investing in the bot, they should buy proxies because without them, you won't get anywhere!
We have approximately 15,000 accounts, is having 200 mobile proxies a good ratio?
 
This is very helpful. What really sticks with me after reading this is the importance of stickiness of IP.

As an automator, I now little by little seeing the big picture.

This really helps a lot. Thank you for sharing.
So maybe now you understand that your 15,000 accounts are all shadow-banned?
 
I'm considering using tools like AskUI or Browser-Use to integrate into automation systems that act unpredictably. The idea is for the bot to interact with a Reddit account in a random manner, responding based on the current page's content. By combining this with models like OpenAI's o1 or other reasoning models, the goal is to create a neural bot capable of browsing Reddit as naturally as possible.
That doesn't scale; it's a waste of resources to simulate something random. Just get the browser off and that's it.
 
Three months ago, I had to rebuild my bot after the increase in Reddit bans, and only after two months of intense work (I can say more than 12 hours a day), I managed to get the first results. Today, I can say that my bot is complete.
I requested three quotes to fix my bot from people more experienced than me, who might have taken even less time (their estimates were also around three months), and the quotes ranged from 7K to 10K.
How do you automate? Are you using browser automation with Selenium/Puppeteer, or do you work directly with Reddit’s private API?

I've been automating with Reddit's private API, but a lot of my accounts keep getting banned. I suspect it's because they don’t have any browsing or scrolling history, making them look unnatural.

Now, I’m trying a hybrid approach—using SeleniumBase to automate browsing and scrolling, while handling commenting, posting, upvoting, and subscribing through the Reddit private API. Not sure if this will work.
 
I don't follow you. What do you mean by turning off the browser? Logging out of Reddit? Staying idle on Reddit?

How do you automate? Are you using browser automation with Selenium/Puppeteer, or do you work directly with Reddit’s private API?

I've been automating with Reddit's private API, but a lot of my accounts keep getting banned. I suspect it's because they don’t have any browsing or scrolling history, making them look unnatural.

Now, I’m trying a hybrid approach—using SeleniumBase to automate browsing and scrolling, while handling commenting, posting, upvoting, and subscribing through the Reddit private API. Not sure if this will work.
It should work! I made ab to and it works using Playwright, mimicking human behavior.
 
I don't follow you. What do you mean by turning off the browser? Logging out of Reddit? Staying idle on Reddit?
Reverse engineer all network calls to the Reddit server that are made from the browser, then mimic them without the browser. I know it may sound difficult, but if you want to succeed in any environment, you must be willing to do what the majority does not do and ensure that what you do is not easily copied by that majority.
 
Back
Top