Course 4 · Steerability, Knowledge, Working Memory, Next Token Prediction
Blog Dial Lab
Input: one topic phrase, plus your answers to a few questions that change depending on where you set four dials — one per property covered in this course.
Outcome: two real, copy-paste prompts — one built from your own dial settings and answers, one asking an AI to choose its own optimal dial settings for shareability and write accordingly. You run both externally, paste the two resulting blogs back in, and get a real sentence-by-sentence comparison of what changed.
Useful for: seeing, on one topic, exactly how tuning next-token style, knowledge depth, working-memory structure, and steerability each shape the actual finished writing — not just in theory, in your own output.
This tool never claims to know what makes content "go viral," and it never writes anything itself. It only builds two prompts and compares the two real blogs you paste back — the comparison is a mechanical text diff, not a verdict on which is better.