The notes are the working record of this project.

A note may be a single roast log, a comparison between roasts, a calibration update, a reflection on something the machine did, or a correction to earlier thinking. They are not fixed. They show the learning as it happened.

Some older notes reflect what I understood at the time. Later roasts may refine or correct those ideas.

---

Good places to start

• Roast 91 and Roast 92 — Same Plan, Different Result
• Roasts 84 to 91 — Stabilising the Current Heater
• Roast 70–75 — The Repaired Gene Café Has a Different Fingerprint
• Roast 36 & 37 — Cold vs Hot Start Behaviour

---

Key groups within the notes

Post-repair recalibration

These notes track how the machine behaved after a heater repair in 2026 and why I became more cautious about repeatability claims. Start with Roast 70–75.

Repeatability checks

These notes ask whether similar plans produce similar outcomes. The short answer is: sometimes, and the exceptions are usually informative.

Cold vs warm start behaviour

One of the most consistent variables across all roasts. Roasts 36/37 and 39/40 document this most clearly.

Reference roasts

A small number of roasts used as practical anchors for comparison. Roast 8 (Brazil Topazio) is the most frequently referenced upper-bound example.