Browsers were built for eyes
A browser that skips rendering because nobody is looking is a sensible optimisation and a quiet admission about what the web has become.
The efficiency argument
Chromium spends most of its budget producing pixels for a person: layout, compositing, animation, font rasterisation. An agent filling a form needs the document and the interaction model, not the picture. Remove the presentation layer and most of the cost goes with it.
The operational argument matters more than the compute one. Running headless browser fleets at scale is a persistent tax — version drift, memory leaks, bot detection, session management — and moving it to a hosted service is the same move that turned running your own mail server from a default into a decision.
What it concedes
If the fastest way to let software use a website is a purpose-built browser that skips rendering entirely, then the human-facing web has become an API of last resort. Every site that never shipped a real API is now being consumed through a simulation of a person who does not exist.
The pattern repeats one layer up
Muse Code launches its own sub-agents to work across large repositories in parallel, trained alongside Muse Spark 1.2 — Meta argues the coupling itself improves coding performance.
And one layer down, Naïve exposes company and infrastructure setup as an API that Cursor, Claude Code and Codex call directly, reaching 30,000 developer accounts in months without building a product surface at all.
Three companies, three layers, one design principle: build for the agent, reach the human through whatever harness they already run. The interface that wins is the one no person ever sees.
TechCrunch — Cloudflare launches Kitesurf, a browser built for AI agents → · TechCrunch — Meta launches Muse Code, an AI agent for large code bases → · TechCrunch — Naïve raises $28.5M to automate the grunt work of setting up and running a company →