I’m starting the day with a horrible cup of coffee from my machine. It’s so bad that I don’t even bother to finish it. I stare into the yucky cup during my meal and it stares back. I’ve never been more motivated. It’s on.

First breakthrough

I’m pulling shot after shot, with varying parameters and success. One or two shots seem to go in the right direction. The twenty or so other ones are undrinkable. I start to spit all coffee to avoid permanent damage to my cardio-vascular systems.

I’m getting nowhere, so I take a break to think. I’m having a discussion with ChatGPT, that tells me to vary my parameters in certain ways. That’s what I’ve been doing since 2 hours, to no avail.

Back to basics it is. I’m re-watching a video of an expert dialing in a coffee. They explain how to change the taste with each parameter. It’s interesting, but another detail catches my eye: They have a very clear preparation routine.

The coffee leaves the grinder in a little cup. They shake it after grinding to disperse clumps. The little cup fits on the filter. They cover the cup with the filter, turn it around, and in a swinging motion they distribute the coffee in the filter. The distribution is already even and flat. I didn’t care about this detail at all.

The tamping is done with a thick stainless steel tamper. My thin plastic tamper from the Gaggia looks very sad in comparison.

The porter-filter sits in the hot machine during the grinding. After its preparation, they pull a little bit of water from the machine, insert the porter-filter and start the shot. I do none of that. I wash my porter-filter with water after every shot, cooling it down considerably. Then I let it lie next to the machine while I prepared the coffee. And of course I didn’t stabilize the boiler temperature with some water flow.

I try to follow the steps in the video exactly as shown. I pull one shot. It’s as bad as the other ones before. I note the flow time and yield. Second shot, meticulously prepared like the first one. And here’s the first break-through: It was as bad as the first one. With the same timing, flow rate, everything. I was so happy I nearly drunk the shot.

Second breakthrough

My first victory tastes sweet, contrary to the coffee I’m making. But the day is not over yet. As my caffeine-infused state of euphoria fades a bit I set myself a new goal: Being able to change the taste of my shots.

I follow the instructions I get from ChatGPT and some gurus on YouTube. I manage to change one aspect of the coffee, say its acidity. But I always also change other aspects involuntarily. Time for another thinking break.

My mental model so far has been simple. I have three inputs:

1. Dose: Amount of coffee that I put into my filter, in grams.
2. Grind setting. A continuous scale from fine to coarse.
3. Yield: Amount of water I let through my coffee to extract it. This is usually measured as a ratio of coffee in. So a yield of 1:2 means X grams of dose => 2*X grams of liquid coffee out.

These inputs influence the output, my espresso shot. I’m looking to influence:

1. Acidity
2. Sweetness
3. Bitterness
4. Mouth-feel
5. Extraction time
6. …

But, the input parameters influence all output parameters at the same time. One example: If I want less acidity, one suggestion is to grind finer. This works, but adds extraction time and bitterness. I’m unable to predict the impact of input changes.

I need a better mental model of what I’m doing. Here’s what I came up with: I have a system where I’m trying to influence a process, the coffee extraction. The extraction process generates my outputs in a predictable way.

How does coffee extraction work?

Apparently one could divide it in three phases:

1. Early phase: Acidity.
2. Middle phase: Sweetness, body.
3. Late phase: Bitterness.

The first two phases have a length in seconds, while the late phase takes all the rest of the overall extraction time. So extraction time becomes part of the process, not an output.

How do my parameters influence the extraction process?

1. Dose: Higher dose means more resistance from the puck. This increases the overall extraction time, but lowers the extraction per part of coffee. I think of it as stretching the extraction, leaving the relative times of the phases intact.
2. Grind setting: Finer grind setting means more resistance from the puck, so longer extraction time. But since the parts get smaller, there’s more surface exposed to water so each part goes through the phases of the extraction faster. I imagine it as increasing overall extraction time, diminishing the relative share of the middle and early phases of the extraction.
3. Yield: Shorter yield shortens the time in which extraction can happen. So it shortens extraction time. This time though, it doesn’t change the absolute timings of the early and middle phase, removing only the late phase. I imagine it as cutting the extraction from behind. I could cut the whole late phase to produce a shot that is hardly any bitter. Or make an overly acidic shot more balanced by adding some late phase bitterness.

This mental model allows me to predict the consequences of an input change to the output. I try it out and it works. The dopamine rush is wild.

Good coffee

I have a vague idea of the kind of coffee I want to produce: Very sweet and balanced, neither very acidic nor very bitter.

I’m trying to get there, applying my new mental model. But I hit a wall. No matter what I try, I can’t get my shot to be less acidic without making it quite bitter at the same time. I accept this defeat and finish my coffee day.