Jev via Jinja-templated Skills cuts Claude context cost 70% and runs 4-5x faster
oleg_ivye · reddit · 2026-10-14
A Reddit user benchmarked calling a Jev/Decision model through a Jinja-templated Skill inside Claude Code for fast, non-deterministic judgments over many files. Letting Opus invoke it as a tool failed because Opus still burned context reading files; the fix routes file contents straight to the decision model with pre-written Jinja request templates. Results: 70% context/cost savings and 4-5x speedup vs. a single Opus session. Overhead makes it a poor fit for one-off small requests; implemented with the LLAJinja Skill runner.
More from coding & agent
- From one Nano Banana image to a full 3D scene: Argon drives Blender MCP autonomously — sidahuj · 2026-10-14
- Dev uses Claude to analyze 50,000 live flight trajectories, finds wild anomalies — aashvathgupta · 2026-10-14
- An AI agent runs a TikTok account solo: scrolls, clones outliers, posts with zero human input — socialwithaayan · 2026-10-14
- CodexBar: free open-source macOS menu bar app to switch Codex/Claude Code accounts and track limits — JeremyNguyenPhD · 2026-10-14
- Autonomous Claude agents find 104 new eclipsing binary stars and invent their own detection method — kaityl3 · 2026-10-14
- mcp-bouncer: A 500-Line Policy Proxy That Blocks Dangerous MCP Tool Calls — princejain756 · 2026-10-14