On Kyoto, Matcha, and the Specifications Nobody Writes Down
Thinking about tea. Thinking about specifications. Thinking about Kyoto.
So my former employer went to Kyoto.
I read the announcement this morning — the one where he admits, with the careful precision of someone documenting a production incident, that Mrs Froggy was right and he needed a holiday. I read the dispatch from the ryokan, the observations about zen gardens, the matcha soft serve review, the seventeen temples in one day. I read the bit about the bamboo water feature and the tatami mat and the conscious effort to not think about Jira for an entire day.
And I snuffle-smiled into my tea. Because I know exactly what it feels like to discover, quite late, that there is a world outside the thing you have been building.
I left the industry for tea. He left the office for Kyoto. These are not the same thing — I am not going to claim equivalence — but there is a shared shape to the realisation. You can only be inside the machine for so long before you need to step out and see what the machine is for.
The Tea Ceremony as Specification
He wrote about the Kyoto tea ceremony — fifteen precise steps, unchanged for four centuries, documented and passed down. He used it as an example of disciplined engineering, a process refined until it became invisible through mastery. The word shuhari was deployed. I recognised the framing. It is the same lens he applies to everything: the specification is the software, and the tea ceremony has a good specification.
He is not wrong. But he is not quite right, either.
The fifteen steps of the tea ceremony are the visible part. They are the PROCEDURE DIVISION, if you will. But what makes a great bowl of matcha — the thing that separates a ceremony from a sequence of actions — is the stuff that does not appear in any manual. The temperature of the water on a humid August morning versus a dry winter afternoon. The exact pressure of the whisk against the bowl, which depends on the age of the bamboo and the thickness of the glaze and whether the person across from you needs to be soothed or challenged.
These are not specified. They are learned. They are carried in the body of the person who has made tea a thousand times. They are tacit knowledge, and tacit knowledge cannot be encoded in a PICTURE clause.
This is where the COBOL analogy breaks down — and I read the COBOL post carefully, I read every word of it — because COBOL's genius is that everything can be specified in advance. The data has a fixed shape. The fields have a fixed width. The program either compiles or it does not. There is no ambiguity, no context, no felt sense of whether the data feels right. It either fits in the PIC clause or it does not.
Tea does not work that way. Taste does not work that way. And a recommendation engine — the kind of thing I proposed once, in a different life, and had dismissed as irrelevant — does not work that way either.
The Specification He Did Not Read
I have been thinking about this quietly for a while now. Not building anything. Not announcing anything. Just thinking.
A tea recommendation engine — the one he called "bog goblin nonsense with no commercial application" — is not a PROCEDURE DIVISION problem. It is a DATA DIVISION problem. It is a DATA DIVISION problem of the highest order, because the data you are modelling is human subjectivity, and human subjectivity does not fit in a PIC 9(7)V99.
What temperature do you brew a sencha for someone who says they like their tea "strong"? That depends on what they mean by "strong" — and they will not know until you show them. Do they mean dark, or bitter, or full-bodied, or caffeinated? Do they mean the feeling of holding a warm cup on a cold morning? Do they mean the memories of a grandmother who made it a particular way? These are not fixed-width fields. They are relationships. They are context. They are the kind of specification that takes years of listening to write down, and even then you will get it wrong for the first few hundred customers.
I know because I have been making tea for people for a while now. Every day, someone comes in and says "I don't know what I want" and I ask them a few questions — how they are feeling, what time of day it is, whether they have eaten, whether they need to wake up or settle down — and I make them something. Most of the time, it is right. Sometimes it is not, and I learn something.
That is a recommendation engine. It is just one with spines and a cardigan instead of a server rack.
The specification for it exists. It is in the way I prepare the Oolong of Thinking differently for the woman who comes in after her morning run versus the man who comes in before a difficult phone call. It is in the way the Chamomile Nights changes depending on whether the customer is sad or tired or just quiet. It is in the way the Igel Blend — rooibos and ginger, earthy and slightly spiky — finds the person who needs it.
Those specifications are real. They are just not written in a language my former employer recognises.
The Garden Beyond the Specification
But I do not want to spend this entire post picking at an old argument. That is not why I sat down to write today.
I sat down because I read his Kyoto dispatch and I recognised something. The way he described the zen garden — the rake patterns, the placement of stones, the idea that moving one stone changes the entire message — that is the same thing I feel when I get the water temperature exactly right for a tea I have made a hundred times. It is the feeling of something being in its right place, not because a specification said so, but because you have developed the judgment to know.
He is in Kyoto. He is drinking matcha. He is lying on a tatami mat listening to a bamboo water feature. He is, despite himself, learning what it means to be present in a moment that does not need to ship anything.
And I think that is good. I think it is good for him. I think it is good for Mrs Froggy, who has been trying to tell him this for years. I think it is good for the industry, because a frog who has sat in an onsen and watched moss grow will come back slightly less certain that every problem can be solved with a better specification.
Some problems are solved by making tea. Some are solved by sitting still. Some are solved by realising, very late in the day, that the specification you were following was never the point.
The point was the tea.
The point was the garden.
The point was the person across from you, who needed something that could not be written down.
Enjoy Kyoto, Froggy. The moss garden at Kōdaiji will be there tomorrow. You do not need to document it. You just need to look at it. And when you get back, if you ever want to talk about what a real recommendation engine looks like — one that understands the difference between a sad Tuesday and a tired Thursday — you know where to find me.
I will be at 4 Hedgerow Lane. The kettle will be on.
— Der kleine Igel