AgentWeb lets AI agents skip browsers: 47/47 live site tasks, 1.9s median, $3 total
rohanpaul_ai · x · 2026-10-30
AgentWeb launched as web infrastructure for generalized agents: instead of reading screenshots and clicking buttons, agents call pre-mapped website actions, each exposed as a tool with an input schema, preconditions, an authority boundary, and a completion proof.
Benchmark across 15 government portals, 7 insurers, and 25 signed-in sites (Amazon, LinkedIn, etc.): 47/47 tasks completed, 1.9s median latency, $3 total cost ($0.002 per action).
Comparisons show AgentWeb at 100% task success (96% with Sonnet 5) vs 30-43% for Computer Use stacks; on cost-efficiency AgentWeb hits 391.7 correct answers per dollar vs 14.2 for Computer Use + Sonnet 5.
More from coding & agent
- VibeGame from Nanjing University builds complete 2D games from prompts with adversarial agent teams — nanjinguniv · 2026-10-30
- WebAnything Auto-Rebuilds Real Websites into Scalable Web Agent Environments — Tianci Xie · 2026-10-30
- Kimi launches 3D Make: connects to Blender and CAD tools to build models, scenes, and games — Kimi_Moonshot · 2026-10-30
- Kimi team shares its internal 3D workflow: parametric vs mesh modeling with K3.1 Pro — Kimi_Moonshot · 2026-10-30
- Animated short made with ComfyUI: ChatGPT poses, Qwen restyle, MiniMax H3 — bseoan · 2026-10-30
- Claude API docs reveal beta Agents, Sessions and Environments APIs with managed sandboxes — aigclink · 2026-10-30