Gene Café CBR-301 Roasting Guide, What Has Stopped Changing

Start here

If you’re new, start with:

Introduction

This guide exists to capture what has stopped changing.

Milestones are recorded as patterns stabilise. The first appears below after 20 roasts.

It isn’t a complete manual, and it isn’t meant to be read end to end. It’s a place where patterns that repeat often enough, across coffees, roasts, and mistakes, are written down once they feel stable.

Everything here started life elsewhere — usually as uncertainty.

Coffee doesn’t give up its answers easily.

The work isn’t about controlling flavour directly. It’s about setting conditions… and paying attention to what comes back.

How this guide is built

Most of the work on this site happens in the Notes.

Notes are where individual roasts are logged, questions are explored, and assumptions are tested. Over time, some of those questions stop feeling urgent. When that happens, the understanding they leave behind is distilled and recorded here.

If something isn’t in the guide yet, it hasn’t earned the confidence to be.

How to use this guide

This guide is not prescriptive.

Nothing here should be treated as a rule, a guarantee, or a best practice. It reflects what has worked consistently enough on a single setup, with a limited but growing range of coffees.

Some sections may feel obvious. Others may feel incomplete. Both are fine.

If you just want to roast and drink coffee, you don’t need to read any of this.

What you’ll find here over time

As patterns repeat, this guide will gradually include:

How decisions are made when signals are ambiguous
What tends to matter more, and less, than expected
How different coffees respond to similar treatment
Simple reference approaches for repeatable, drinkable results

Nothing is added quickly.
Nothing is fixed permanently.

Crack audibility on naturals (what stopped changing)

Across multiple natural coffees roasted on the same Gene Café system, first crack proved to be coffee-dependent rather than system-dependent.

Brazilian naturals consistently produced quiet or smeared crack behaviour, even when development was sufficient. This made time, colour, aroma, and weight loss more reliable decision anchors.

An Ethiopian natural later demonstrated that clear, rolling crack could return without any system change, confirming that airflow and exhaust were not suppressing crack universally.

A subsequent Honduras Catuai natural further refined this understanding by producing a clean, audible crack at the lowest development level recorded. This showed that crack audibility does not imply deeper roast development.

Taken together, these roasts establish that:

First crack is descriptive, not decisive
On naturals, especially under active exhaust, structural cues are more reliable than sound

Milestones

After 20+ Roasts
Gene Café CBR-301 — Milestone Synthesis

As of Roast 21 (19 full roasts and 2 sample roasts), this page records what appears to have stabilised.

This is not a conclusion.
It is a checkpoint.

Scope

Included in this milestone:

19 full 250g roasts
2 small sample roasts
Washed and natural processes
Central America, Brazil, Ethiopia

Machine baseline:

Gene Café CBR-301
250g batch
250°C set
Fan 2 → 3 around 4:00
No preheat

All roasts logged via JSON.
All weight loss calculated at cool initiation.

What has stopped changing

Thermal stability

Climbs heat predictably
Does not stall at first crack
Maintains positive development momentum
Produces repeatable total roast times

The machine is stable.

Airflow inflection

Drying feels less stable (without fan step)
Chaff clearing is slower
Roast progression stabilises (with fan step)
Crack clarity improves
Development feels more controlled

This step is now part of the working baseline.

Weight loss as structural anchor

More reliable than crack sound
More reliable than total roast time
More reliable than visual colour

14–15% WL → balanced medium structure
~16% WL → fuller structure
~17%+ WL → upper guardrail

Weight loss is currently the most dependable structural reference.

Crack audibility is coffee-dependent

Washed: audible and reliable
Naturals: often muted and unreliable

Crack is informative. It is not authoritative.

Coffee-specific structural behaviour

Structural band is coffee-dependent, not development-time dependent.

Development seconds alone do not determine roast depth.
Pre-crack energy and bean response materially influence outcome.

Repeatability confirmed

Endpoint timing defines structure
Variation is not random
The machine behaves predictably

What has not stabilised

Moving a coffee between structural bands
Drum speed variation
Temperature modulation
Multi-variable experiments

That restraint is intentional.

System vs coffee

System behaviour appears stable.
Coffee behaviour varies within that system.

Guardrail status

No guardrail redefined
Patterns observed, not declared

Any change requires:

At least 3 consistent roasts
Across at least 2 coffees

Download

Gene Café CBR-301 – What I Wish I Knew Before Roast 1 Download

Safe starting points
Airflow awareness
Total roast time (TRT)
Weight loss guidance
Common early mistakes

Closing reflection

The system is stable enough…
to begin asking better questions.

Twenty roasts did not create certainty.
They created clarity.

And for now… that is enough.