Jev Ultrafast: The Sub-10-Second Web Agent Architecture

Description and use

Jev Ultrafast replaces pixel input with DOM snapshots and predicts actions in a single request, cutting task execution time to 7 s and reducing token cost.

Problem

Autonomous web agents suffer from excessive latency and high token costs when using vision-based browsing, making real‑world web tasks slow and expensive.

Outcome

Jev Ultrafast reduces task execution time to sub‑10 s, cuts token usage, and lowers browser protocol calls by ~90 % compared to vision‑first agents.

Workflow

• Take atomic DOM snapshot of visible controls
• Generate element table
• Send single LLM request for operation and speculative targets
• If operation is TYPE_TEXT, invoke lightweight LLM for text
• Validate target node visibility and freshness
• Execute browser action

Evidence

Autonomous web agents represent one of the most promising frontiers of modern artificial intelligence. However, developers who have deployed browser agents in production quickly encounter a severe constraint: excessive latency and prohibitive token costs.

Source

By substituting raw pixel analysis with structured DOM snapshots, speculative multi-head decisions, and decoupled text generation, Jev executes complex web tasks—such as finding flights on Google Flights—in 7.1 seconds.

Source

Google Flights (Zürich to London): Complete task execution in 7.07 seconds (including goal parsing, airport selection, date configuration, and flight list verification).

Source

Chrome DevTools Protocol calls plummeted from 1,092 down to 101— a 90.7% decrease in browser traffic.

Source