Date: 25 May 2026
This was a small personal roasting session on the Gene Café CBR-301, aimed mainly at coffees for pour-over and Fellow Aiden brewing. The intention was not to chase the darkest or fullest roast, but to look for sweetness, enough body, and useful roast references for future planning.
I also recorded sound during these roasts where possible, partly for the ongoing crack activity monitor idea. As always, any crack markers or audible popping notes here are treated as practical observations, not confirmed scientific first-crack data.
Session Summary
| Roast | Coffee | Start | Input | Output | Weight Loss | Endpoint | Working Result |
|---|---|---|---|---|---|---|---|
| 117 | Colombia Regional Jess, Huila washed blend | 25°C | 248.3 g | 208.6 g | 16.0% | 11:52 | Upper-edge sweet/full personal filter roast |
| 118 | REDBER Colombia Swiss Water Decaf | 65–67°C | 249.0 g | 212.1 g | 14.8% | 11:08 | Successful decaf sweetness anchor |
| 119 | Honduras Pacas Washed HO2520 | 62–65°C | 250.0 g | 218.6 g | 12.6% | 10:18 | Too light / low weight-loss reference |
| 120 | Honduras Pacas Washed HO2520, end of bag | 64°C | 245.5 g | 212.1 g | 13.6% | 11:05 | Improved but still low weight-loss reference |
The Aim: Sweetness for Pour-Over
The working target for this session was sweetness rather than maximum development. For the washed coffees, I was broadly thinking in the region of 15.0–15.6% weight loss. For the Swiss Water decaf, the target was slightly lower, roughly 14.6–15.2%, because decaf often behaves differently in the roast and can look darker than it actually drinks.
These numbers are not universal rules. They are simply reference bands for this machine, this project, and my own brewing preferences.
Roast 117: Colombia Regional Jess
The first roast was a cold-start Colombia Regional Jess, a Huila washed blend. The day was warm, and the roaster started at about 25°C. The roast moved quickly early on.
- Input: 248.3 g
- Output: 208.6 g
- Weight loss: 16.0%
- Endpoint: 11:52
- 200°C: 3:43
- 230°C: 6:09
- 245°C: about 7:40–7:45
- 250°C: 8:35
- App crack marker: 10:01
- Audible crack note: good crack activity around 10:25
This landed slightly above the intended sweetness band at 16.0% weight loss. I would describe it as an upper-edge personal filter roast rather than a mistake. It may produce a fuller, rounder cup with more developed sweetness, but it could also lose a little clarity depending on how it drinks.
The useful future note is simple: for a similar warm-day cold Colombia roast, I would probably shorten slightly if aiming for the middle of the sweetness range. Something closer to 11:40–11:45 may be a better next test, unless this roast cups especially well.
Roast 118: REDBER Colombia Swiss Water Decaf
Roast 118 was the REDBER Colombia Swiss Water decaf. This was a warm-start roast and it moved very quickly through the early milestones. Based on the live data, I shortened the endpoint from the original plan.
- Input: 249.0 g
- Output: 212.1 g
- Weight loss: 14.8%
- Endpoint: 11:08
- Start temperature: about 65–67°C
- 200°C: 2:48
- 230°C: 4:34
- 245°C: 5:49
- 250°C: 6:23
- App crack markers: none recorded
This was probably the cleanest success of the session. Despite the very fast early pace, the final weight loss landed at 14.8%, which sits nicely inside the intended sweetness range for this decaf.
I would now treat this as a useful warm-start REDBER SWP decaf anchor. For similar starts around 65–70°C, a total roast time around 11:05–11:10 looks like a sensible reference point, subject of course to cup result.
Roast 119: Honduras Pacas Washed HO2520
Roast 119 was where the session became more interesting. This Honduras Pacas Washed coffee looked very fast by the Gene Café app milestones, reaching 245°C and 250°C much earlier than expected.
- Input: 250.0 g
- Output: 218.6 g
- Weight loss: 12.6%
- Endpoint: 10:18
- Start temperature: about 62–65°C
- 200°C: about 2:38
- 230°C: about 4:17
- 245°C: 5:33
- 250°C: 6:02
- App crack marker: 9:28
The live decision was to shorten the roast because the milestones looked extremely fast. That decision made sense in the moment, especially after some recent Honduras roasts had landed high in weight loss.
But the scales told a different story. At only 12.6% weight loss, this roast was clearly much lighter than intended. The wrong button press near the end added only a few seconds, so that was not the real issue. The roast was simply too short for this coffee.
The important lesson: early 245°C and 250°C app milestones did not translate into high weight loss for this Honduras Pacas. This is exactly why weight loss remains such an important validation number.
Roast 120: Honduras Pacas Washed HO2520, End of Bag
Roast 120 was the same Honduras Pacas Washed coffee, this time an end-of-bag roast at a slightly smaller batch size. Because Roast 119 landed very low, the endpoint was extended materially.
- Input: 245.5 g
- Output: 212.1 g
- Weight loss: 13.6%
- Endpoint: 11:05
- Start temperature: 64°C
- 200°C: 2:36
- 230°C: 4:12
- 245°C: 5:28
- 250°C: 5:56
- App crack marker: 9:30
This was a better result than Roast 119, but still below the intended sweetness band. Extending from 10:18 to 11:05 increased weight loss by about one percentage point, from 12.6% to 13.6%.
That confirms the direction was right, but the correction was not large enough. If roasting this same coffee again from a similar warm start, I would now consider something closer to 11:50–12:05 for a sweetness-focused pour-over roast.
Main Learning From the Four Roasts
This session reinforced a few useful points.
- The Colombia roast showed the machine was lively today. Roast 117 reached the upper band quickly and landed at 16.0% weight loss.
- The decaf correction worked very well. Roast 118 landed at 14.8% and is now a useful warm-start decaf sweetness reference.
- The Honduras Pacas did not behave like the previous Honduras Wanted coffee. Despite fast milestones, Roasts 119 and 120 landed low in weight loss.
- Fast app milestones are useful warnings, but they are not proof of roast development. The scales and the cup still have to validate the result.
- Crack markers and audible pops are useful notes, not absolute decision points. They help build context, especially for future sound analysis, but I still do not want to chase single pops.
Range Finder / Roast Planning Implication
For the evolving Gene Café Roast Report and Range Finder work, this session is a very useful reminder.
The tool should not simply assume:
Early 245°C means the roast must be stopped early.
That was a reasonable live caution for Roast 119, but the final weight loss showed that this coffee needed more time. So the prediction model needs to consider not just machine pace, but also coffee identity, batch size, start state, previous same-coffee results, and final weight loss.
In plain English: the app graph helps, but the scales still get a vote.
Working Labels for the Dataset
- Roast 117: upper-edge Colombia sweetness/fullness reference
- Roast 118: successful REDBER SWP decaf warm-start sweetness anchor
- Roast 119: low-WL / underdeveloped Honduras Pacas warm reference
- Roast 120: improved but still low-WL Honduras Pacas end-of-bag reference
The cup results will decide how useful each roast really is, but from a learning point of view this was a valuable session. Not because every roast hit the target, but because the misses were clear and measurable.
That is exactly the kind of data this project needs.