Today, a browser can render anything, and that's why it feels like the safe default. But let's be precise about what we are actually launching. A browser is not a library you call. We're talking about an entire operating system that you boot. Because today, a browser brings a lot of stuff, like a process scheduler, its own memory allocator, audio-video codecs, a lot of IPs, and of course, its network stack. Chromium today is around 35 million lines of code, which is in the same order of magnitude as the Linux kernel.

The most expensive line of code in a scraping task is the one that launches the browser. Not because it's slow, but because of what it costs you on every single page. And a scraping task is not one page, it's often hundred thousands a day. What you spend loading one, you spend again every time, until it stops being a technical details, and it becomes how many machines you need to rent.

Fresh from the hive: Auto-Mode is live. The problem? Every scraping target behaves differently. One blocks standard proxies. Another needs JS rendering. A third requires premium infrastructure. Auto-Mode solves this. Add mode=auto and it tests configs automatically, starting simple and escalating only when needed. Stops at first success. Set max_cost to cap how far it goes. Check sb-auto-cost to see what worked. Manual mode stays for full control. Auto-Mode is for when you want extraction to just work.