Build Your Coffee OS

Phenomenon

You are standing in your kitchen holding a bag of beans you just bought. The bag lists the origin, the process, the roast date and three flavor words.

You could follow some recipe from the internet. If it happens to taste good, great; and if it doesn't? Many people change several things at once by feel, then several more the next day, and a month later they have occasionally brewed a great cup without being able to say what they did right that day.

There are also people who have brewed coffee for years and make a good cup every day, yet if you ask them "why is today's cup better than yesterday's?", they cannot tell you.

This last chapter of the course is about exactly that question: can you explain your cup of coffee?

Variables

The previous thirty-five chapters all come together in one loop. At each step, you are making a decision:

There is one thing not in the table that runs through all of it: the record. Without it, the loop starts from zero every day.

Experiment

A thought experiment: two people brew the same bag of beans for two weeks.

• A adjusts by feel every day, changing whatever seems off — sometimes two or three things at once — and keeps no notes.
• B writes down the recipe and the taste every day, changes only one thing at a time, and before brewing writes down "what I expect to happen".

Write down your prediction before reading on

After two weeks, who brews the better coffee? Who knows more about this bag? When the bag runs out and a new one arrives, who will dial it in faster?

Your prediction

Save prediction

Stored only in this browser on this device. Never uploaded.

A second thought experiment: yesterday you brewed three cups at different grinds. Today all you remember is "the second one seemed best".

Write down your prediction before reading on

What do you actually know? What do you think you know that you really do not?

Your prediction

Save prediction

Stored only in this browser on this device. Never uploaded.

Observation

In the first experiment, after two weeks A may have brewed the occasional better cup; but the one who will dial in a new bag faster is almost certainly B. What A knows is "which days were lucky"; what B knows is "move this variable in this direction and the taste changes like this" — a map that can be carried to the next bag.

In the second experiment, you know less than you think. "The second one was best" might be because of its grind, or because it was at just the right temperature when you drank it, because of what you had eaten, or simply because it was in the middle. As Chapter 35 showed, expectations and surroundings change what people taste Confidence: [Strong Evidence]; and comparing from memory a day later is far less reliable than comparing side by side on the spot Confidence: [Rule of Thumb]. Without having written down which grind each cup used and what you tasted, "the second was best" can hardly become knowledge you can use next time.

Model

Coffee OS: a causal map, plus a log

"Coffee OS" is not a recipe but a method for finding your own recipes. It has two parts.

The first part is the causal map — the chain this course has built up along the way:

Bean (origin, processing, roast, freshness) → grind and bed → water, ratio, temperature, time → extraction and strength (and how even they are) → taste → preference

Most of the first half of the chain was already fixed the moment you bought the bag (Chapters 1–18); the dials you can turn in your kitchen cluster in the middle (Chapters 19–34); and the last two links are results you read with your tongue and nose (Chapter 35). The knowledge graph draws this chain.

The second part is the log: for every cup, write down the numbers along this chain, then what you tasted. The causal map tells you "where it might be"; the log tells you "what actually happened this time".

What goes in a brew log

A record that does the job looks something like this:

Three things are worth noticing:

• Write the prediction first, then brew. This is what the whole course has kept doing. With the prediction written beforehand, the result can test your model instead of being "explained" into it after the fact.
• Write "description" and "liking" separately. "Acidity softer, slightly drying finish" is description; "4 / 5" is preference. The Specialty Coffee Association's new assessment method also splits description and the impression of quality into two separate forms Confidence: [Common Practice] (Chapter 35). Keep them apart and you can see which descriptions your liking actually follows.
• Write only one "next cup". The next cup changes one thing.

You can keep the record in a notebook or on your phone, or use the record form under each experiment on this site's Real Labs page; those records are stored only in this browser on this device, and you can review and export them in My Progress.

One variable at a time, and its limits

Changing one variable at a time (Chapter 32) is the engine of this method. But in a real kitchen it has some limits:

• Some variables are tied together. More coffee changes the ratio and the bed depth together; a finer grind changes the time along with it. Turn one dial, then predict what it will pull along.
• Some variables change by themselves even if you do not touch them. Beans age every day (Chapter 17). To compare two recipes, brew them side by side on the same day with the same bag Confidence: [Rule of Thumb], not one today and one next week.
• Small differences may just be noise. Brew the same recipe twice and it will not taste exactly the same. Repeat a change once before trusting its effect Confidence: [Rule of Thumb]. When you are not sure whether two cups really differ, check with Chapter 35's triangle test: can you reliably pick out the odd cup of three?

Fix problems first, then find preference

When dialing in a cup, problems and preferences should be handled separately, and it usually works best to deal with problems first, then look for preference Confidence: [Rule of Thumb]:

1. Is the extraction even? Sour, bitter and astringent at once with a normal time: check evenness first (Chapter 33).
2. Is the extraction about right? Sharply sour and thin: move toward more extraction; bitter and drying: move toward less (Chapter 24).
3. Is the strength right? Balanced but too weak or too harsh: change the ratio (Chapter 22).
4. Only then, preference. Brighter or softer, stronger or weaker, hotter or less hot.

Step four has no standard answer. Research has found that the strengths and extraction yields consumers like are spread over a wide range, and fall into several clearly different preference groups Confidence: [Strong Evidence]; researchers have also been able to map the sensory qualities and liking people report onto a chart whose axes are strength and extraction yield Confidence: [Strong Evidence]. That means two things: your "best" is probably not the same as someone else's; but it is not mysticism — it sits on these two dials you can turn.

Preference is discovered

Many people assume preference is innate and already known. In fact most people can only say what they like after comparing cups side by side Confidence: [Rule of Thumb]. Without a comparison, "tasty" is a vague feeling; once you have tasted four extractions of the same bean, you know which side you stand on. And your preference can change Confidence: [Preference] — the more you drink, the more likely you are to notice things you did not before. The log records that change too.

Transfer

You can take this loop anywhere:

• A new bag of beans: brew it first with a recipe you know as a baseline, taste, pick one variable, change it a small step.
• A new brewer: first work out which method family it belongs to (Chapter 25); you already know which variables to look at first.
• At a café: order and give feedback in descriptive language — "a little more body, a little less acidity".
• A friend says "my coffee is too sour": ask what they changed and what they recorded, then change one thing together.

When you can say of a cup "what I changed, what I expected, what happened, and why", this course has done its job.

Misconceptions

"There is one best recipe; just follow it." A recipe is a starting point. Beans, water, grinders and brewers all differ, and so do preferences. Research keeps finding that the strengths and extractions people like are spread widely. Someone else's recipe can save you part of the journey, but the last stretch can only be covered by your own comparisons and records.

"Memory is enough; there is no need to write anything down." Memory gets rewritten by expectations, surroundings and time. Without a record, it is hard to tell whether a change came from the variable you changed, or from the beans aging, today's water, or your mood.

"I changed it once and it got better, so the change works." One result can be noise. Repeat it, ideally side by side with the original recipe, before drawing a conclusion.

"Coffee I like must score highly, and highly scored coffee is coffee I will like." Preference and quality are two different questions (Chapter 35). Writing "description" and "liking" separately in the log is exactly so that one does not answer the other.

Real Lab

This chapter's real lab is the graduation project for this phase: One Bean, Four Cups.

Deliberately make four cups with different extractions, record what you observe, and explain why they differ. The first three cups are Chapter 20's grind experiment:

Preparation.

1. Choose one coffee and a brewing method you know (pour-over, French press or AeroPress all work), and set a baseline recipe: dose, water, temperature, technique.
2. Get four identical cups and mark the bottom of each.

Four cups. Apart from the one change listed, make each exactly like the baseline:

Before brewing, write a prediction for each cup: which will be the most sour? The most bitter? What will D taste like?

Tasting. Let the four cool to about the same temperature and have someone shuffle them. Without looking at the marks, write a description of each cup using Chapter 35's six dimensions, then give each a "liking" score. Only then reveal the marks.

Explaining. For each cup, write a sentence or two: how does it differ from A, and why? Explain it with the causal chain from this course: grind → surface area and resistance → extraction → taste; for D, the problem lies in the flow → uneven extraction.

Want to go one step further? Dilute C with a little hot water until its strength is close to A's, and taste the two side by side. Dilution changes strength but not extraction (Chapter 23): whatever difference remains comes from the extraction itself.

Can't brew four cups right now? Simulate the four cases first in the Atlas Extraction Bed and write your predictions; when you get the chance to brew real coffee, come back and compare.

When you are done, put the results in your log. Finally, pick just one variable to change in your next cup — your Coffee OS starts running from this cup.

Check Your Model

Predict. You switch to a darker-roasted coffee and keep the rest of your recipe the same. Before brewing, write down: which way do you expect this cup to move? If it comes out more bitter and drying than usual, which variable would you change first, and why?

Diagnose. A friend's log says: "Today I got a new grinder, switched to mineral water, and added two grams of coffee. Much better!" What can they conclude from this entry? What would you suggest they do next?

Explain. In five or six sentences, explain to a friend who has just started drinking coffee: what happens between a coffee cherry and "I like this cup"? What was already fixed when you bought the beans, and what can you control in the kitchen?