The country with the most active residential IPs in our network is the United States, by a wide margin.
Top 7 by live active IPs: US 967K, UK 302K, Canada 234K, China 136K, Brazil 106K, Italy 95K, Germany 91K (Byteful dashboard, June 2026).
Why it matters: a 35M global pool with 94K IPs in Italy behaves nothing like one with 940K. The shape of the coverage matters more than the headline number, so ask for active counts in your target country.
I run GLM 5.2 inside Claude Code with the same tools, the same skills, and the same agent loop. Three environment variables is all it takes. https://www.zyte.com/blog/how-to-run-any-model-inside-claude-code?utm_campaign=blog-posts&utm_activity=ORS&utm_medium=social&utm_source=mastodon
Research and Development at a Web Scraping company https://www.youtube.com/watch?v=VYfLRRVx2tI?utm_campaign=blog-posts&utm_activity=ORS&utm_medium=social&utm_source=mastodon
An interview with Zyte's head of R&D. Read it here: https://www.zyt...
More instruction, worse output. Zyte's head of R&D on why telling your agent exactly what to do can blind it to the obvious answer. https://www.zyte.com/blog/the-best-agent-skill-is-the-one-that-says-the-least?utm_campaign=blog-posts&utm_activity=ORS&utm_medium=social&utm_source=mastodon
For a human, 10ms is invisible. For an AI agent chaining thousands of requests, it's everything.
So we traced one residential request end to end: closest of 12 POPs for TLS (~15ms), pick from 35M+ exits (under 8ms), session (~40ms), target round trip (~300ms).
410ms median, down from 690ms. Most of it is physics, not us.
Is GLM-5.2 really closing the gap to Anthropic - and at just a fraction of the cost - or is it just more AI hype? I think so, and let me show you why. https://www.zyte.com/blog/why-im-adding-glm-5-2-to-my-agentic-coding-arsenal?utm_campaign=blog-posts&utm_activity=ORS&utm_medium=social&utm_source=mastodon
For data gatherers, negotiating anti-bot tech is not for the faint-hearted. That’s why we built a multi-layered system that treads lightly but brings a host of tactics to the table. https://www.zyte.com/blog/how-we-built-a-320000-strategy-web-access-hero?utm_campaign=blog-posts&utm_activity=ORS&utm_medium=social&utm_source=mastodon
Why my Scrapy project plays a triumphant fanfare when a crawl finishes clean and a sad trombone when it doesn't, and how I finally learned how to build Scrapy extensions (it's easy) https://www.zyte.com/blog/how-to-build-your-first-scrapy-extension?utm_campaign=blog-posts&utm_activity=ORS&utm_medium=social&utm_source=mastodon
"Four people, four diets, two work schedules, and a baby who answers to nobody. That's what finally made me build a personal agent." A walkthrough of the actual architecture I run to hold my household and my DevRel work together — profiles, skills, memory, and the web-data layer that makes it all reach the live web. https://www.zyte.com/blog/my-personal-agent-setup-the-architecture-that-runs-my-household-and-my-dev-rel-work?utm_campaign=blog-posts&utm_activity=ORS&utm_medium=social&utm_source=mastodon
"Four people, four diets, two work schedules, and a baby who answers to nobody. That's what finally made me build a personal agent." A walkthrough of the actual architecture I run to hold my household and my DevRel work together — profiles, skills, memory, and the web-data layer that makes it all reach the live web.
Proxy Users in the Byteful dashboard 🔐
Create separate users for teammates, apps, and agents.
Cap data per product, e.g. limit an agent to 5GB.
Restrict which proxies each user can access.
One user works solo. With a team or multiple apps, it scales with you.
Byteful
Zyte
