Skip to content
AI Token Coach
Get the Burn Brief

Friday, 16 October 2026

AI Token Coach.

Editorial intelligence for serious AI users.

Make Every Token Count.

A daily desk on context, models, agents, harnesses and routing. Not a token-saving app.

Open Burn RadarWhat are AI tokens?
News DeskWhat changed, and what it does to your burn
Full desk
Tell the desk what you use and how you pay. Burn is read against your stack, not an average.

Burn Radar

All radar items

Slower than the News Desk. A radar call reads several signals together and says what to change.

Four places your capacity moves.

Context, models, harnesses and routing. Every item on the desk sits under one of these. Follow a lever to see its Burn Radar items, canonical pages and guides together.

/radar

Context Craft

Shape prompts, files, and memory so models see what matters and little else.

/models

Model Efficiency

Match models to jobs so the expensive option is used when it earns its keep.

/harnesses

Agent and Harness Efficiency

A harness runs the loop and can spawn agents inside it. Tune loops, tools, and retries so they finish useful work instead of spinning.

/routing

Router Lab

Test routing rules that send each task to the right model at the right cost of capacity.

Where a token actually goes.

One task, five places its capacity can leak or land well. Scroll, and follow the coin from prompt to output.

PROMPT>_CACHE

  1. 01

    Your task

    A prompt, a pasted file, a whole thread. Every word in it is already a token, and every turn sends the lot again. Context is the first place capacity leaks.

Models

All models

Harnesses

All harnesses

Not currently verified by AI Token Coach

Hermes

Coding Agent

Nous Research

The Burn Brief. Daily editorial, plus techniques you can use today.

One short note on making capacity turn into useful work, then three or four concrete moves on context, models, agents, or routing. Free. Built for practitioners.

Free. Daily on weekdays. Unsubscribe anytime.