Average AI prompt length over time

Name: Average AI prompt length over time
Availability: InStock
Author: welder

by welder·Apr 22, 2026·3 points·0 comments

Visit Project View on HN

AI Analysis

●●SolidSolve My ProblemSlick

Tracks prompt length and edit rates across Cursor, Copilot, and Claude Code.

Strengths

•Quantifies AI code quality via follow-up edit rates, not just usage.
•Leverages existing IDE plugin base for data collection without new agent installs.
•Compares individual prompt depth against team averages for context optimization.

Weaknesses

•Requires manual config file editing to backfill historical AI data properly.
•Another dashboard for managers to check, adding to existing tool fatigue.

Post Description

Hi HN, I built WakaTime 13 years ago before AI. Things have changed a lot since then, and the time you spend typing in your IDE isn't as valuable as it used to be...

That's why I built a new WakaTime dashboard specifically for AI metrics. It tracks things like:

* Lines of code - AI vs. Human

* Average Prompt Length over time - How much context do you give AI

* Follow-up Edit Rate - How many times do you have to edit/fix the code AI generated

* Token usage over time - Input and Output tokens

It does this for most popular agents (Cursor, Claude, Codex, Copilot, Gemini, etc.)

If you already use WakaTime plugins, you should have AI data since last week. If not, install one of the WakaTime official plugins (https://wakatime.com/plugins) then edit your ~/.wakatime/wakatime-internal.cfg to change the ai_heartbeats_last_parsed_at to some date in the past and it will backfill your AI stats since that date. By default it only backfills a few minutes from plugin install.

Let me know what u think!