The Killer Feature Stands on the Cheapest Input

"Guitopia bet on audio → model, the part it was worst at. Bonfire bets on model → understanding, the part Pippa is best at."

What Each Feature Actually Needs

Map Bonfire's real features to the analysis they require, and a pattern jumps out:

Feature                            Needs         Tier
-------                            -----         ----
"explain why this solo works"      chords/key    A  (cheap, mature)
"easy-mode improv over this"       chords/key    A
"simplify this passage to my level" chords/key   A
"play the original note-for-note"  every note    B  (hard, rare)

Every killer feature lands in Tier A. The only thing that needs Tier B is the one feature Bonfire deliberately doesn't build — note-for-note — and links out for instead. The expensive input buys almost nothing; the cheap input buys the whole product.

Dad's Workflow Already Skips the Hard Tier

This isn't a theoretical convenience — it matches how the tool's actual user works. The whole point of easy-mode is that you don't need the original note-for-note to learn from a song. You need to know the changes, the key, the shape — then you simplify and improvise. The hardest analysis tier is exactly the one the real workflow never reaches.

Two Directions, Two Bets

Guitopia pointed at audio → model: squeeze a precise model out of raw audio, including every note. That's the direction where the technology is weakest and the failures are most visible. Bonfire points the other way — model → understanding: take a cheap, mostly-right model (chords, key) and turn it into explanation, simplification, and coaching. That's the direction Pippa is strongest in. Raw-audio transcription stays a funnel into the product, never the heart of it.

Code

Killer features all stand on Tier A·text

Feature                              Needs        Tier
-------                              -----        ----
"explain why this solo works"        chords/key   A   (cheap, mature)
"easy-mode improv over this"         chords/key   A
"simplify this passage to my level"  chords/key   A
"play the original note-for-note"    every note   B   (hard, rare, link out)

# Direction matters:
#   Guitopia: audio -> model        (its weakness)
#   Bonfire:  model -> understanding (Pippa's strength)

Exercise

For a product idea, write each headline feature next to the cheapest input that would make it work. If a feature can run on a cheap, mature input, that's a green light. If it secretly needs the most expensive, flakiest input, ask whether you're pointing at your strength or at the hardest available problem.

Hint

The trap is assuming the headline feature needs the richest possible input. Usually it needs far less than you think — and the gap between 'what it needs' and 'the hardest input' is pure avoidable risk.