Hey, basically the title, I am looking for another Residential Proxies provider as the current one started having issues with the infra, and my projects can't stay idle while they fix their issues. Anything trustworthy to recommend?
I've been at this company for five years now, most of that time as one of the core engineers, and things changed practically overnight. Management decided a few weeks ago that we're going all in on agentic AI for internal workflows, and I actually support that in principle. What I can't get behind is the process they've suddenly rolled out. The mandate came down from non-technical management with basically no warning. The new expectation: prompt an AI coding tool to generate the entire backend, agent logic, tool definitions, skills, all of it, then engineers just adjust it until the thing behaves. There's no code review requirement anymore. Nobody has to actually understand what the generated code is doing before it goes out. Security is supposedly the infra team's job, but that only covers network, auth, deployment. Nothing about the application layer, prompt injection through tool outputs, agents with permissions that are way too broad, data crossing user boundaries that shouldn't be crossed. None of that gets caught because reading the code isn't part of the process anymore. After five years of doing this the normal way, a few things are really sitting wrong with me. Nobody owns the code anymore. When something breaks in production down the line, I'm the one who's going to be reverse-engineering logic that no human, including whoever "wrote" it, ever actually understood. And personally, I've spent these years building real engineering judgment at this company. Watching that get replaced overnight with "just trust the output" feels like a gut punch. I'm completely fine with AI-assisted development where I still understand and own what ships. I'm not fine with this. So, questions for anyone who's lived through a similar shift: has anyone managed to push back successfully on a sudden requirement from non-technical managers, and what actually worked? Does the fact that this happened suddenly, after years of a completely different standard, change how seriously I should take it as a red flag? And for people in AI-heavy dev shops, where do you personally draw the line between real AI-assisted engineering and just vibecoding?
Context: I've been running a small shoe bot on and off for 4 years, did this for myself and a few friends, some Jordans and the odd Yeezy restock back when those still were relevant. Always relied on python, requests plus some selenium for the checkout. It worked fine in 2023. This year it started failing mid-drop, like I was getting 403s between add-to-cart and payment steps. Met up with one of my buddies and spent some time analyzing what's up and why. Basically, it turns out the shoe sites, at least the bigger ones running on Akamai or Shape, aren't really looking at your IP as a default, they're doing TLS fingerprinting on the handshake itself. Python's ssl module and requests have a JA3 signature which gives it away pretty easily. Didn't matter that I was rotating through a clean pool of ISP proxies, the handshake gave it away straight away.
My buddy helped me rewrite everything in Rust using reqwest with rustls, and paired it with a TLS impersonation setup so the fingerprint matched a real Chrome client. Since my knowledge of Rust wasn't that great, had to resort to my encyclopedia of a friend lol. Checkout success rate on a mid-tier release went from something like 30% to around 80% just from that one change!! I could not believe it to be honest, and we didn't even touch the proxy pool. Figured that the proxy side is a whole separate deal. Mobile proxies are still great for shoe sites simply coz carrier-grade NAT means a lot of bots can share one IP and it still looks organic, but their main issue is that they are expensive and slow as hell, I was seeing 800ms+ response times on some of the 4G proxy providers I tried. Datacenter proxies are very cheap compared to other types and also fast but they get flagged instantly. I wasted like 200 datacenter proxies at some point before switching. After some time I ended up with a mixed setup, residential proxies for the browsing and monitoring tasks, and a smaller pool of dedicated mobile proxies just for the checkout requests.
Weird part is I thought the proxy provider was the issue this whole time but it was actually my HTTP client giving me away.
Maybe someone will find this useful and learn from ones mistakes, because not always its providers fault
The sneaker stores that I was scraping are these if anyone's wondering:
Repos from Github that were used, (some are archived as of now)
Hey, perhaps someone used their services? Talking about proxies and also interested in their AI studio. Would like to hear pros and cons. From what I found out already is that their services are a bit pricey, but other than that, I would like to know the uptime, performance, stability etc.