Uncommunicated Expectation Is Premeditated Resentment
There is a line I keep returning to: uncommunicated expectation is premeditated resentment. If you hold an expectation of someone and never voice it, you are not being considerate — you are scheduling a future grievance. You already know, in advance, that you will resent them when they fail to meet the standard they never knew existed.
I know this. And I still find it brutally hard to act on.
The problem, for me, is that when I try to communicate expectations, I get misread — often, and in ways I could not predict. Enough repetitions of that and you start assuming everything will be misunderstood, so why bother saying it? Which is, of course, just the original trap with better lighting. I know the real bottleneck is that I am bad at communication, and I know that is a fixable skill, not a fixed trait. Knowing it and fixing it are different projects.
Interestingly, the same failure mode shows up in a completely different domain: workshops and seminars.
I have sat through enough of them to notice a pattern. Most are too general to be useful — the speaker’s real product is their own branding, and the audience’s real cost is their time. A good YouTube video beats most seminars, and it is free. The reason is structural: a live session must fit the average background of everyone in the room, and the moment you optimize for the average, the only things you can transmit are generalities. Superficial content is not a speaker failure; it is the mathematically inevitable output of designing for everyone.
A recent example: an AI and agentic AI workshop. To be fair, I am biased — I work in data and AI, so “nothing new” is my baseline complaint. But the depth of the content was genuinely thin: plan, execute, evaluate. Buzzwords you could retrieve from a search engine in seconds. What was missing was everything practical — the actual platform choices, the tech stack, which models are state of the art for agentic workflows. The most useful thing a workshop can share is one practitioner’s concrete, idiosyncratic setup; instead we got the averaged-out nothing.
Worse, they skipped the single most important caveat of the field: agentic AI hallucinates, constantly, and prompting alone does not fix it. What tames hallucination is context engineering — giving the model enough context about its environment that it can use tools to check its own work. That is the practical core of the whole discipline: provide context so the model can reach your spreadsheets, your video channel, your banking, your investment accounts, your chaotic crypto portfolio. Context is what turns a confident liar into a useful agent. Nobody mentioned it.
So the pattern repeats across scales. In relationships, unspoken expectations curate future resentment. In teaching, content averaged for everyone teaches no one. The common root is the same: communication that avoids the risk of being specific. Specificity is what gets misread; it is also the only thing worth saying. The resentment, the wasted seminar, the shallow workshop — all of it is the price of dodging that risk.
Drafted with AI assistance from my personal journal (2025-10-15), then edited by me. See On AI Assistance.