Jev Demos Added This Week

Routing wallet decisions from market signals, fees, and model cost during a 22-hour trading run

Highlighting topic-relevant passages by confidence instead of matching keywords alone

Comparing MiniMaxH3 acceleration methods across six scenes and reaching up to 9× faster generation without visible quality loss

Scanning 50,000 conversations in 20 seconds, finding 151 backlink forums, and building a placement calendar

Reading the current Claude Code task and recent context to select the best-fitting model automatically

Classifying TikTok clips in about 70 ms to repost likely viral videos and skip the rest

Mapping connections across 1,010 X posts and grouping them into 22 suspected scam syndicates

Building a blog, creating a post, and opening shared items from spoken instructions alone

Filtering 3,412 leads with 20,472 fast decisions so Grok opens only high-signal records

Completing a Google Flights search in 7.1 seconds with 10 model calls and 5.6× fewer tokens

Classifying two million RedReplier rows at about 5× the previous speed while improving quality

Evaluating every visible X post with 12 typed questions and returning keep, reply, skip, or mute decisions

Running an autonomous two-player Pac-Man match with Laya, MLX-Serve, and typed decisions

Choosing one of seven steering actions from precomputed ball trajectories while separate modules handle vision and reflexes

Searching Twitter bookmarks by meaning to return more relevant saved posts than keyword matching

Completing five official browser actions and a DONE decision locally on one RTX 5090 in 10.375 seconds

Checking uploaded resumes against explicit conditions and supporting every verdict with a quoted passage

Classifying observed trades from wallet history and graph context while keeping expensive models out of routine decisions

Rebuilding an open-source Clash Royale bot for iOS and routing its in-match control choices through typed decisions

Controlling a Flappy Bird player and saving scores in Neon Postgres for people to compare runs

Comparing local Laya and API-hosted Jev on the same movie search, returning equal results in 392 and 676 ms

Choosing whether to use a sub-agent, ask a person, stop retrying, or reuse cache before costly Picsart generation

Assembling prepared colors, copy, design, and layout parts from audience attributes without regenerating everything

Classifying page elements without filter lists, removing 16 ads, and caching the decisions for the next visit