Measurements and field notes

What we learn running a proxy network, tested rather than asserted. Evergreen explainers live in the knowledge base; this is where the experiments go.

Your scraper's retry logic is probably making the block worse

A 429, a 403 and a TLS reset are three different failures. Only two are worth retrying — and on targets that count volume rather than rate, no backoff schedule saves you.

Read the post → 21 August 2026

Stop reaching for a headless browser to scrape documentation sites

Docs sites are built by a handful of generators that leave the text in predictable places. Three routes get the content without launching a browser.

Read the post → 11 August 2026

What actually fails against Cloudflare: 28 requests, measured

Four combinations of exit IP and TLS stack against seven Cloudflare-fronted sites, every response logged. Including the JA3 result nobody mentions.

Read the post → 27 July 2026

10,000 requests through our own network

The full benchmark: success rate, unique exit IPs, geo accuracy and latency across five countries — with the raw dataset published under CC BY 4.0.

See the data → July 2026