jevgrep · Launch Video Breakdown: Hook, Pacing & Motion Design

Scene-by-scene timeline & spoken transcript

1. 0:00 – 0:04

   Problem Agitation

   The Token Burn Problem

   “(No spoken dialogue — Driving electronic beat with pizzicato synths and sub-bass. High-energy audio transition at 0:00 and 0:03.)”

   On screen

   jevgrep • jg +4.3k tok +3.6k tok +3.4k tok +3.7k tok +4.7k tok +4.4k tok Coding agents burn tokens AGENT CONTEXT SPENT 20,690 tok Coding agents burn tokens just finding the right code. 99,675 tok

   Camera

   Static overhead view, then subtle zoom out to reveal more files.

   Motion

   Animated file icons appearing and accumulating, with a magnifying glass highlighting 'token burn' areas. Text appears with a subtle bounce and color highlight.

2. 0:04 – 0:14

   Product Reveal

   Introducing jevgrep: Smart Code Search

   “(No spoken dialogue — Electronic beat continues, sound design elements at 0:08 and 0:11. Pizzicato synths become more prominent.)”

   On screen

   jevgrep • jg jevgrep Find code by asking what it does. ~ /my-repo — zsh jg "where is excerpts • complete L50-71 source L12-48 lead L5-31 source HOW IT WORKS — BEHIND jg 01 Walk the folder frontier Jev judges each directory from its children — rejected branches are never read or uploaded. src/ docs/ my-repo/ tests/ scripts/ HOW IT WORKS — BEHIND jg 02 Classify files by content Content fragments, not filenames, decide which files qualify. lookahead → db/ db/connection.py db/pool.py db/models.py test_connection.py HOW IT WORKS — BEHIND jg 03 Split into declarations Python and TS/JS parse into functions & classes, with exact line refs. src/db/connection.py import sqlalchemy as sa class Pool: def acquire(self): def create_engine(url, pool_size=10): return sa.create_engine(...) def close(conn): def _retry(fn, n=3):

   Camera

   Static shot, then a subtle pan and zoom to follow the terminal input and subsequent file structure breakdown.

   Motion

   Logo reveal with a 'dot' animation. Terminal typing animation. Animated flowcharts illustrating file system traversal and code parsing, with elements appearing and connecting dynamically.

3. 0:14 – 0:20

   Feature Teaser

   Jev: The Core Research Agent

   “(No spoken dialogue — Electronic beat intensifies, sound design elements at 0:14 and 0:18. Sub-bass becomes more driving.)”

   On screen

   jevgrep • jg Jev RELEVANCE JUDGMENTS 210 folders • files • declarations POWERED BY Jev SystemOne models typesafe-ai/jev RELEVANT 6 PRUNED 24 Fast, efficient research RELEVANCE JUDGMENTS 728 folders • files • declarations POWERED BY Jev SystemOne models typesafe-ai/jev RELEVANT 33 PRUNED 71 Fast, efficient research agents. Small models judge relevance — your frontier agent spends its budget on the code. CODEX + jg — INDEPENDENT RESEARCH AGENT Codex CODING AGENT where is pooling configured? Jev RESEARCH AGENT Codex delegates the research.

   Camera

   Static, then subtle camera shake and zoom on the 'Jev' core. Pans to follow the data flow.

   Motion

   Pulsating 'Jev' core with radiating lines and data points. Animated counters for 'Relevance Judgments' and 'Relevant' files. Flowchart animation showing interaction between 'Codex' and 'Jev'.

4. 0:20 – 0:31

   Call to Action

   Cost Savings & Get Started

   “(No spoken dialogue — Electronic beat maintains high energy, sound design elements at 0:21 and 0:26. Final sub-bass drop/impact at 0:29.)”

   On screen

   jevgrep • jg SWE-BENCH — 10-TASK REPEAT • FULL TASK COST Codex, prompted to use jg as its researcher. Codex alone BASELINE $7.62 Codex + jg JEV RESEARCH AGENT $4.49 SWE-BENCH — 10-TASK REPEAT • FULL TASK COST Codex, prompted to use jg as its researcher. -40% $7.62 $4.52 cheaper Codex SWE-BENCH — 10-TASK REPEAT • FULL TASK COST Codex, prompted to use jg as its researcher. -40% $7.62 $4.52 cheaper Codex runs on SWE-bench. $7.62 → $4.52 full Codex task cost across 10 tasks, failures included; Jev cost excluded. Solves: 7/10 with jg vs 8/10 baseline. Single repeat on a tuned Python subset — a cost reduction with a quality tradeoff, not a guarantee. jevgrep • jg Find the context. Start coding. npm i -g @dzhng/jevgrep github.com/dzhng/jevgrep — MIT — Node 22+

   Camera

   Static shot on the bar chart, then a subtle zoom out and pan to the final call to action.

   Motion

   Animated bar chart revealing cost savings with a striking '-40%' overlay. Text highlights and strikethrough for comparison. Final logo reveal with radiating grid lines and a terminal command for installation.