Brendan Lynskey · Portfolio home · Interactive sites
Anthropic Computer Use, OpenAI Operator, Gemini Mariner, plus browser-only agents (Browser-Use, Playwright). Action schemas, screen vs DOM grounding, sandboxing, evals.
Live index: https://brendanjameslynskey.github.io/LLM_Hub_Computer_Use/
| # | Title | Status | Description |
|---|---|---|---|
| 01 | Computer-Use Architectures | live | Anthropic Computer Use, OpenAI Operator, Gemini Mariner; action schemas (click x,y vs accessibility-tree vs CSS selectors); screenshots vs DOM vs accessibility tree; containers and VMs; OSWorld, WebVoyager, AgentBench. |
| 02 | Browser Agent Patterns | live | Browser-Use, Playwright agents, Steel; DOM extraction strategies; screen vs DOM trade-offs; WebArena and Mind2Web evals; captcha & anti-bot; stateful sessions; pricing; failure recovery. |
Related site: Agent Security Explained (code): how to keep agents that read untrusted pages and drive a computer safe, defences only: trust boundaries and labels, untrusted input, the lethal trifecta, least privilege, sandboxing and egress control (chapter 5), plan-then-execute, approvals and the security–utility frontier, each chapter an animation computed by the policy module of Agent_Loop_Sim.
Part of the LLMs hub — an index of presentation series for AI/LLM engineers.