Skip to content

Latest commit

 

History

History
87 lines (58 loc) · 2.11 KB

File metadata and controls

87 lines (58 loc) · 2.11 KB

ContentForge Calibration — Blind Taste Test Results

Started: 2026-04-05 Target: 20 submissions across 3+ platforms Accuracy target: ≥70% correct top-5 ranking before Product Hunt launch


Current Read

No aggregate performance claim should be made yet.

Until the table below has real submissions, ContentForge should be described as:

  • deterministic
  • explainable
  • under active calibration

That is the honest state of the engine today.


Method

The blind taste test is meant to answer one question:

Can the heuristic engine rank historically better performing posts above historically worse ones without seeing the original metrics first?

Minimum standard before stronger public claims:

  1. At least 5 real submissions
  2. At least 3 platforms represented
  3. Clear notes on misses, not just wins

Calibration assets:

  • docs/calibration_dataset_template.csv
  • docs/calibration_dataset_template.json
  • scripts/calibrate_content.py
  • docs/calibration_examples.json

Launch feedback notes:

  • docs/reddit-launch-notes.md

The Reddit launch notes are qualitative market signal only. They are useful for positioning and UX decisions, but they do not count as calibration proof.


Results

# Participant Platform Posts Top 5 Correct Accuracy Date

Aggregate Stats

Metric Value
Total submissions 0
Platforms covered 0
Overall accuracy
Ready for PH? Not yet

Per-Platform Breakdown

Platform Submissions Avg Accuracy Weakest Signal
Twitter 0
LinkedIn 0
Instagram 0
TikTok 0
Other 0

Recalibration Log

When a submission reveals a miscalibrated signal, log it here:

Date Platform Signal Before Weight After Weight Accuracy Change

Updated automatically as submissions come in via Discussion #4.