After the correct replacement heater was installed in my Gene Café CBR-301, I did not want to assume the machine had simply gone “back to normal.”

That would have been tempting.

The machine was repaired. The heater had been changed. The roast behaviour before and after repair had not felt the same. And after 100 roasts, the one thing I have learned is that confidence does not come from assuming the roaster is doing what I expect.

It comes from checking.

So this next phase of the project started with a simple question:

What does this machine do now?

Not in theory.
Not according to old expectations.
Not according to one roast.

But across a small set of controlled, boring, useful roasts.

This note is my current verdict after the first four roasts on the restored heater.

It is not a final conclusion.
It is not an engineering diagnosis.
It is not an official statement about the Gene Café CBR-301.

It is a practical field note from my machine, my setup, and my roast records.

The short version

The restored-heater machine now looks usable, repeatable, and more energetic than the previous repaired phase.

The Brazil cold-start roasts landed close enough together to give me a credible cold baseline again.

The Guatemala warm-start roasts showed something equally important: this machine now carries a lot of momentum when warm. A warm start at around 66–67°C pushed the roast much harder than I would have expected if I had relied on old timing assumptions.

The biggest lesson so far is simple:

The machine appears to be behaving consistently again, but warm starts need much shorter expectations.

That is probably the most useful verdict at this stage.

Not “fixed forever.”
Not “back to the original machine.”
Not “solved.”

Just:

Repeatable enough to work with, energetic enough to respect, and different enough that old assumptions are not automatically safe.

The four roasts

The first four restored-heater roasts were:

Roast	Coffee	Start state	Start temp	End time	Weight loss
Roast 99	Brazil Yellow Bourbon Natural	Cold	approx. 24°C	approx. 13:13	14.9%
Roast 100	Brazil Yellow Bourbon Natural	Cold	22°C	13:13	15.3%
Roast 101	Guatemala Caturra Washed	Warm	66°C	12:51	18.4%
Roast 102	Guatemala Caturra Washed	Warm	67°C	11:46	16.4%

Roast 100 was a 250 g Brazil Yellow Bourbon Natural roast ending at 13:13 with 15.3% weight loss. Roast 101 was a 250 g Guatemala Caturra Washed warm-start roast ending at 12:51 with 18.4% weight loss. Roast 102 used the same Guatemala coffee and a near-identical warm start, but was stopped at 11:46 and landed at 16.4% weight loss.

That table is the story.

The cold Brazil roasts were close.
The warm Guatemala roasts were powerful.
The corrected second Guatemala roast proved that changing the endpoint still gives meaningful control, even if the machine itself has strong thermal momentum.

What I was trying to find out

This was not about finding the perfect roast profile.

It was about rebuilding trust.

After a repair, part change, or any sudden shift in roast behaviour, I think there is a real risk of asking the wrong question. The wrong question is:

“What recipe should I use now?”

The better question is:

“What does the machine do now under simple conditions?”

So I wanted to check four things:

Does the machine produce a credible cold-start result with a familiar coffee?
Can that result be repeated in the same general zone?
What happens when the machine starts warm?
Are the early milestones useful enough to guide live decisions?

After four roasts, I think the answer to all four is yes — with caution.

Observed

1. The cold Brazil roasts were close enough to matter

Roast 99 and Roast 100 were both Brazil Yellow Bourbon Natural cold-start roasts.

They were not identical, but they landed in the same practical neighbourhood.

Roast 100 started at 22°C, ran to 13:13, and finished with 15.3% weight loss. That is slightly higher than Roast 99, but close enough to tell me that Roast 99 was probably not a fluke.

This matters because one roast can always mislead.

Two similar cold-start roasts on the same coffee do not prove everything, but they are much more useful than one isolated result.

The practical message from the Brazil pair is:

The restored-heater machine has a credible cold-start anchor again.

That is a big deal.

It means I am not trying to recalibrate from noise. There is a repeatable reference point.

2. The warm Guatemala roast overshot hard

Roast 101 changed the mood of the test.

It started warm at 66°C, used Guatemala Caturra Washed, ran to 12:51, and landed at 18.4% weight loss.

That is high.

Not slightly high.
High enough that I would treat it as an overshoot or high-development reference rather than a normal target for my usual filter preferences.

The important part is not just the final number. It is how the roast got there.

Roast 101 reached:

Milestone	Time
200°C	2:43
230°C	4:55
240°C	5:58
245°C	6:41
249°C	7:26
250°C	7:30

That is fast.

Compared with the cold Brazil anchor, the warm Guatemala roast reached the working temperature bands much earlier and had a lot more momentum. Even though it ended earlier than Roast 100, it landed much higher in weight loss.

That is the awkward but useful bit.

The same kind of total roast time thinking was no longer safe.

3. Roast 102 tracked Roast 101 very closely

Roast 102 was originally meant to be a warm repeatability roast.

But while it was running, it became clear that it was tracking Roast 101 extremely closely. It was even slightly faster through some milestones.

Roast 102 started at 67°C, only 1°C warmer than Roast 101.

The early comparison was striking:

Milestone	Roast 101	Roast 102
200°C	2:43	2:41
230°C	4:55	4:53
240°C	5:58	5:45
245°C	6:41	6:28
249°C	7:26	7:06
250°C	7:30	7:10

That changed the decision.

Repeating the same endpoint would probably have repeated the same overshoot. So Roast 102 stopped earlier, at 11:46.

The result was 16.4% weight loss.

Still developed.
Still probably high for where I usually aim.
But much more useful than another 18%+ roast.

This is an important lesson:

The machine gave the same warning twice. The second time, I listened earlier.

That is exactly the kind of evidence-led adjustment I want this project to be built around.

4. Fan behaviour also repeated

The fan behaviour in the Guatemala pair was also useful.

Roast 101 moved from fan 2 to fan 3 at around 3:49 / 218°C. Roast 102 moved from fan 2 to fan 3 at around 3:52 / 219°C.

That is close.

This does not tell me exactly what is happening inside the machine. I am not going to pretend it does.

But as a practical user signal, it matters.

The fan event was not random-looking. It appeared in the same general zone across the two warm Guatemala roasts.

That supports the idea that the restored-heater machine is behaving in a repeatable pattern, even if that pattern is more energetic than I expected.

Working interpretation

My current interpretation is this:

The restored-heater CBR-301 is now behaving like a new practical personality of the same machine.

I do not mean that as an engineering claim.

I am not saying the machine has literally changed into a different design. I am not saying I know the cause. I am not saying this proves anything about every CBR-301.

For this project, “personality” is just a practical modelling word.

It means:

how quickly the machine climbs
when it reaches key temperature bands
how fan behaviour appears
how total roast time translates into weight loss
how cold and warm starts behave
how much I can trust older assumptions

On that basis, this restored-heater phase does appear different from the slower, lighter-landing repaired phase.

The machine now appears more energetic, especially when warm.

The Brazil cold anchors suggest it can repeat in a useful zone.

The Guatemala warm pair suggests that warm starts need to be treated with real caution.

Practical verdict

1. The restored heater appears usable and repeatable

This is the good news.

After Roast 99 and Roast 100, I am comfortable saying the machine has a credible cold-start baseline again.

The two Brazil roasts landed close enough that I can work with the machine rather than feeling like I am guessing from scratch.

That does not mean every roast will now be simple.

It means the machine is giving me patterns again.

That is enough to rebuild from.

2. The machine is not behaving like the slower repaired phase

This is probably the biggest practical shift.

In the previous repaired phase, some roasts seemed to land lighter than expected for the roast time. That made me question old timing assumptions and pushed the project into recalibration.

The restored-heater roasts do not feel like that.

They are not light-landing.

If anything, the warm Guatemala roasts show the opposite risk: the machine can move quickly, spend meaningful time in the upper band, and produce high weight loss even at shorter total roast times.

So for now, I would not use the slower repaired-phase timings as a guide.

They belong to the historical record, not the current working baseline.

3. Warm starts are powerful

This is the sharpest practical lesson from Roast 101 and Roast 102.

A warm start at 66–67°C is not a small detail.

It changed the whole roast.

The Guatemala roasts reached 200°C in around 2:41–2:43, and 230°C in around 4:53–4:55. They were already moving fast very early in the roast.

That means “warm” is not just a label.

The actual start temperature matters.

A warm start at 40°C is not the same as a warm start at 67°C. A warm start at 80°C may be different again.

For future roast planning, I need to treat start temperature as part of the roast, not background information.

4. Early milestones are becoming more useful

Time to 200°C has been useful for a while.

But these roasts reinforce that the later milestones matter too.

In Roast 102, the early tracking against Roast 101 was enough to change the plan. It was not a guess. It was a live comparison.

The roast was matching the previous high-development roast very closely, so repeating the endpoint would probably have repeated the result.

That is a very practical use of the app graph and the temperature milestones.

Not as perfect control.
Not as bean temperature.
Not as a profile-roaster fantasy.

As a comparison tool.

That is where the app becomes useful.

5. Total roast time still matters, but not on its own

This is probably the clearest beginner lesson.

Roast 100 ended at 13:13 and gave 15.3% weight loss.

Roast 101 ended earlier, at 12:51, and gave 18.4% weight loss.

So total roast time matters, but it does not explain everything by itself.

Start temperature, coffee, machine momentum, time spent near the upper band, and fan behaviour all matter.

This does not mean total roast time is useless. Far from it.

It means TRT needs context.

The better question is not:

“What roast time should I use?”

The better question is:

“What roast time makes sense for this coffee, this batch size, this start temperature, and this machine state?”

That is exactly why recalibration matters.

Caution

There are a few things I do not want to overstate.

This is not proof that the machine is back to the original personality

It may be.

It may not be.

The restored heater is producing useful, energetic behaviour, but I do not yet have enough data to say it is identical to the original machine before heater failure.

For now, I am treating it as its own current phase.

That is safer.

This is not a diagnosis of the repair

I am not using these roasts to blame a part, a repair, or the manufacturer.

The useful point is much more practical:

When a machine changes, recalibrate before you experiment.

That is true whether the cause is a heater, voltage, airflow, sensor behaviour, cleaning, ambient conditions, or something else.

The cause matters, but the first job as a home roaster is to rebuild a reliable reference.

Four roasts are useful, but not final

Four roasts are enough to see patterns.

They are not enough to close the case forever.

I now have:

two useful cold Brazil anchors
two warm Guatemala roasts
one clear warm overshoot
one corrected warm result

That is a good start.

But I still need cup results. I still need more coffees. I still need to see how this behaviour holds across future roasts.

So this is a working verdict, not a final verdict.

What I would do next

Based on these four roasts, I would change my planning in three ways.

1. Keep cold and warm planning separate

I would not casually transfer cold-start timing to warm-start roasting.

That is probably the biggest practical rule.

Warm starts need their own expectations.

2. Shorten warm Guatemala targets

For Guatemala Caturra Washed on this restored-heater machine, from a warm start around 66–67°C, I would not start at 12:50 again unless I wanted a high-development result.

Roast 101 showed that 12:51 gave 18.4%.

Roast 102 showed that even 11:46 gave 16.4%.

So if I wanted a more balanced filter roast from a similar warm start, my next starting point would probably be:

around 11:15–11:30

Not because that is guaranteed.

Because that is the next sensible question.

3. Use milestones to decide whether the roast is on the same track

If a roast is matching a previous roast closely at:

200°C
230°C
245°C
fan intervention
first reach of 249–250°C

then I should assume it may land in the same neighbourhood unless I change the endpoint.

That sounds obvious, but it matters.

Roast 102 showed it clearly.

The roast was tracking Roast 101.
Roast 101 had overshot.
So Roast 102 needed a different endpoint.

That is not panic.

That is using the record.

Current verdict after four roasts

My current verdict is this:

The restored-heater Gene Café CBR-301 is behaving consistently enough to trust again, but it is energetic, especially when warm, and old roast-time assumptions are not automatically safe.

That is the practical conclusion.

The Brazil roasts gave me confidence that the machine can repeat from cold.

The Guatemala roasts warned me that warm starts have serious momentum.

Roast 101 showed the danger of carrying old timing forward.

Roast 102 showed the value of watching the early milestones and correcting the endpoint.

So I am not back to “recipe mode.”

I am back to something better:

A working baseline.

That is what I needed.

Not certainty.
Not perfection.
Not proof.

Just enough evidence to know what to check next.

The useful lesson for other Gene Café owners

If your Gene Café CBR-301 changes after repair, heater replacement, cleaning, ventilation changes, or simply starts giving results that no longer match your expectations, I would not start by chasing complex profiles.

I would do this:

Pick a known coffee.
Use a normal batch size.
Record the start temperature.
Use a simple target temperature.
Plan the total roast time.
Record time to 200°C, 230°C, and 245°C if possible.
Record fan behaviour.
Weigh the roasted coffee.
Calculate weight loss.
Taste the result.
Change one thing next time.

That is the whole method.

Not glamorous.

But it works.

And after these four roasts, I trust that method more than ever.