You Can't Make Good Decisions About Things You Don't Understand

There is a specific kind of expensive mistake that smart people make.

They make decisions about things they don't understand well enough, using language they haven't fully decoded, in domains where the vocabulary itself is doing work they haven't noticed.

It happens in finance: the founder who signs a term sheet without understanding liquidation preferences or participating preferred. It happens in technology: the operator who commits to a platform architecture without understanding what "serverless" actually means for their cost model at scale. It happens in AI: the executive who signs an enterprise AI contract without understanding the difference between training and inference, or why context window size matters to their use case.

In each case, the cost isn't just the decision. It's the compounding: the wrong hire who stays too long because you didn't know the interview question that would have surfaced the gap. The wrong architecture that has to be rebuilt at five times the cost. The wrong vendor relationship that locks you in while the market moves.

Vocabulary is not a superficial concern. It is load-bearing infrastructure for thinking clearly.

The problem with how we learn terminology

Most people acquire business and technology vocabulary the same way they acquire most knowledge: piecemeal, when they need it, in the specific context where they encountered it.

This produces knowledge that works in a narrow band and breaks at the edges. You understand "churn rate" in the context of your own SaaS metrics but not in the context of evaluating an acquisition target. You understand "fine-tuning" as something AI companies do but not the operational and cost implications for your own deployment decisions. You understand "MCP" because you saw it in a newsletter but not why it changes the architectural calculus for building on AI.

The gaps are invisible from the inside. You don't know what you don't know. And the questions you don't know to ask are exactly the ones that would have changed the outcome.

Why plain English matters more than people admit

There is a persistent misconception that complex vocabulary reflects complex knowledge. It often reflects the opposite.

The person who can explain RAG in two sentences: "combining AI with real data sources so the model answers from your documents rather than guessing from its training": understands it better than the person who reaches for technical terminology before they've checked whether the concept is clear.

Jargon is efficient when both parties share the definition. It's a fog machine when they don't.

The problem in fast-moving fields like AI is that terminology is being coined faster than understanding can follow. New terms are used in presentations and articles by people who don't fully understand them, to audiences who definitely don't, and everyone smiles and nods and files the term away for use in the next meeting. The vocabulary circulates without the knowledge.

I've been in boardrooms where decisions about AI investment were made by people who were confusing training and inference costs: a distinction that affects the financial model by orders of magnitude. I've watched founders commit to platforms based on a misunderstanding of what "open weights" means for their ability to control their own product. I've seen marketing decks built on the term "agentic AI" by teams who couldn't explain what an agent loop was.

These are not stupid people. They are people who moved fast in a field where the vocabulary wasn't established before the investment decisions were being made.

What 101121 is

101121 is the plain English resource I wished existed when I was figuring this out.

500+ terms across business, startup, technology, and AI. Not exhaustive. Not academic. Specifically selected for the vocabulary that comes up in decisions that matter.

Every definition is written with one question as the test: does a smart, busy person now understand this well enough to use it correctly in a decision? Not to write a paper on it. Not to pass an exam. To use it correctly when it counts.

The AI section is the most recently expanded: 150 terms covering the full vocabulary from foundation models and attention mechanisms to agents, benchmarks, compute infrastructure, and AI safety. These are the terms showing up in board discussions, vendor proposals, hiring decisions, and product strategy right now. The definitions don't assume prior technical knowledge. They assume intelligence and willingness to understand.

The startup section covers fundraising mechanics (from pre-seed through sovereign capital), company structure and governance, growth metrics, sales methodology, product development, and leadership. These are the terms that separate founders who understand the game from founders who are learning it at full cost.

The value of a shared vocabulary

One practical use of a glossary that isn't talked about enough: alignment.

When a leadership team has shared, precise definitions for the terms they use, discussions are faster and more honest. Arguments about "product-market fit" or "customer acquisition cost" or "hallucination risk" don't stall because different people are using the same words to mean different things. The energy goes into the actual decision rather than the vocabulary negotiation.

This is worth more than it sounds. I've watched leadership teams spend half a strategy meeting resolving a disagreement that turned out to be semantic. Same conclusion, different words for it. Neither party had been precise enough to notice.

A shared vocabulary is not a luxury. It's operational infrastructure.

The invitation

101121 is free. No account required for the glossary.

If there are terms you've encountered that aren't there, or definitions you think could be sharper: the contact is open. This is a resource that should improve continuously, and the people best placed to improve it are the ones using it in decisions.

Start at [leonhardwick.com/101121](/101121).

---

*101121 is a free plain English glossary of business, startup, and AI terms. Part of the [leonhardwick.com](/) resource library.*