Coffee | One Roast at a Time

Sample Roast 3 was a 56 g washed Colombia micro-batch that landed at 14.6% weight loss with 1:22 development after the first-crack button press. A temporary gasket fix reduced aroma leakage, and an overheat protection event introduced airflow changes mid-roast. The key learning is not “did I mishear crack,” but that hardware/airflow changes can shift sensory confidence even when the roast lands structurally in a reasonable range. The cup will determine whether this reads as clean-and-light or slightly lean.

Type: Sample Roast (non-repeatable / exploratory)
Roaster: Gene Café CBR-301
Date: 2026-02-20
Coffee: Regional Jess — Huila, Colombia — Washed blend
Batch size: 56.0 g (sample)
Set temp: 250 °C
Preheat: None
Stirring: Level 2 (throughout)
Active exhaust: Yes (baseline)

1) Measured Data (from the Gene Café app JSON)

Batch + weight loss

  • Input: 56.0 g
  • Output: 47.8 g
  • Weight loss: 14.6%

Time + first crack + development

  • Total roast time: 12:07 (727 s)
  • First crack button pressed: 10:45 (645 s)
  • Development time (button → end): 1:22 (82 s)
  • DTR: ~11.3%

Fan / airflow events (this roast’s key feature)

Fan transitions recorded in the log:

  • OFF → 2 at 0:01
  • 2 → 3 at 4:05 (245 s)
  • 3 → 2 at 7:59 (479 s)

Important context (observed in real time):
The fan dropped back because the machine had recovered from an overheat protection event. This is the kind of system behaviour that matters more on a 56 g batch than it does on 250–300 g.

Notes captured at roast time (memo)

“Roast went as normal. Crack was heard but possible it was early as no crescendo at all. Light roast in reality. Hope not under development.”

2) Context That Matters (not in the JSON, but true)

Two real-world factors likely changed the sensory cues:

A) Temporary gasket fix

I temporarily fixed the faulty gasket. That likely meant:

  • Less aroma leakage during the roast
  • A different “smell intensity” baseline than recent roasts

So when the roast “smelled lighter,” it may have been a hardware change, not a roast-development change.

B) Overheat protection event

On the Gene Café, protection events can:

  • Increase airflow temporarily
  • Alter heat retention
  • Change how aromas present
  • Change crack audibility / confidence

On a micro batch (56 g), this kind of event can be disproportionately influential.

3) Interpretation (clearly separated from measured data)

Did I “trigger development too soon”?

Based on the log: unlikely.

The first-crack button was pressed at 10:45, which is not “early” for this batch size on this system. If anything, it suggests I was cautious, not premature.

However, it’s still possible I heard early crack-like sounds before that and mentally entered “development mode” too early. The key issue here isn’t whether I pressed the button at the perfect second — it’s that sensory confidence was degraded by (a) gasket sealing and (b) protection-driven airflow behaviour.

Structural expectation from 14.6% WL + 1:22 dev

At 14.6% WL, this sits in a light-to-light-medium structural zone for a washed coffee on my system. It should not be “green” by default.

Two plausible cup outcomes exist:

Outcome A: Clean but a bit lean

  • Clear acidity
  • Clean cup
  • Possibly lighter body / shorter finish
  • Sweetness present but not “rounded”

Outcome B: Clean and surprisingly balanced

  • Good clarity
  • Enough sweetness for the roast level
  • Body acceptable given the light structure
  • Pleasant finish

The cup decides which is true.

4) What This Roast Teaches Me (the actual lesson)

This roast is less about “did I mis-hear crack?” and more about this:

When the machine behaviour changes (gasket sealing, protection airflow), my sensory anchors shift.
That can make me feel like I made the wrong call even when the numbers land in a reasonable place.

This is exactly the kind of system literacy I want to build: recognising when my confidence is affected by equipment state, not only by roast chemistry.

5) What I’ll Look For When Tasting

I’m not chasing flavour notes here. I’m checking structure.

Targets for evaluation

  • Body: thin vs rounded
  • Sweetness: present vs hollow
  • Finish: short vs lingering
  • Acidity: clean/lively vs sharp/green

Quick decision rule

  • If it tastes thin + sharp + short → I likely ended a bit early for my preference, even if it’s technically fine.
  • If it tastes clean + balanced for a light roast → this is a valid light calibration point, not a mistake.

6) Next Steps (what I’d do differently if I had more coffee)

This coffee was a 56 g sample, so it’s not realistically repeatable. But it does justify a controlled “sample mode” test when I next have any tiny batch to run.

Proposed “Sample Roast Mode” (30–70 g)

Goal: reduce protection events and reduce perceptual uncertainty.

  • Set temp: 250 °C
  • Fan: start at 3 and hold 3 (avoid mid-roast fan swings)
  • Stirring: 2 (test 3 later as a separate change)
  • Preheat: none
  • Decision anchor: time after confirmed rolling crack
  • Guardrail target: ~14.8–15.2% WL with ~90–105 s after confirmed rolling crack (context-dependent)

This is not about optimisation. It’s about stability.

7) Status

  • This is a Sample Roast (non-repeatable reference)
  • Not eligible for “Lazy Mode”
  • Value: system-behaviour learning + light structural calibration point