9 micro-topics in AI Systems, each with the evidence that proves you have it and the written reason behind every link. Nothing outside this domain sits under it — you can start here. Downstream, it holds up 7 domains.
9 topics · depth 0–3 of 16 · 9 internal links · 0 in · 11 out
Load-bearing links first. Each one says what in this domain rests on what outside it, and why.
Nothing outside this domain sits under it. It is an entry point — you can start here.
In prerequisite order. Each idea names the artifact that closes it, the market demand that put it on the map, and what it stands on.
Depth 0 · conceptual · unassisted · M14 §14.1
Places each term inside the next
Depth 1 · conceptual · unassisted · M14 §14.2
Separates supervised, unsupervised and reinforcement learning
Distinguish AI, ML, deep learning and LLMs
Paradigms are how the systems in the hierarchy learn.
Depth 1 · conceptual · unassisted · M14 §14.4
Knows text becomes integer ids, not words
Distinguish AI, ML, deep learning and LLMs
Tokenization is how an LLM receives input.
Depth 2 · conceptual · unassisted · M14 §14.2
Knows the model is not learning between calls
In-context learning is contrasted against training.
Depth 2 · conceptual · scaffolded · M14 §14.3
Knows attention, parallelism and scaling
The architecture is what made the scaling work.
Depth 2 · conceptual · unassisted · M14 §14.4
Knows the hard limit and what fills it
The window is measured in tokens.
Depth 3 · meta · unassisted · M14 §14.4
Converts usage into cost per request
Cost is a function of tokens in and out.
Depth 3 · meta · unassisted · M14 §14.5
Compares families and their tradeoffs
Models differ by window and capability.
Model choice is mostly a cost and latency decision.
Depth 3 · conceptual · unassisted · M14 §14.5
Knows the same input may not give the same output
Sampling is why the same prompt varies.
Everything downstream that names an idea here as a prerequisite, grouped by where it lives.
4 links
Configure a model client rests on Navigate the model landscape
You configure a specific model.
Tune generation parameters rests on Reason about nondeterministic output
Sampling parameters control the variation.
Maintain chat context rests on Explain the context window
History consumes the window.
Enforce spend caps in code rests on Compute token economics
A cap requires knowing what a call costs.
2 links
Treat the context window as a budget rests on Explain the context window
A budget requires a limit.
Treat the context window as a budget rests on Compute token economics
The budget has money and latency on its axes.
1 link
Recognize silent LLM failure rests on Reason about nondeterministic output
Variation is why fluent output can be wrong.
1 link
Track and control AI cost rests on Compute token economics
You must know the unit cost to attribute spend.
1 link
Identify sources of bias rests on Compare learning paradigms
Bias enters through training data.
1 link
Draw the model boundary rests on Reason about nondeterministic output
The boundary exists because model output varies.
1 link
Explain embeddings rests on Explain tokenization
Embeddings operate on tokenized text.
Retool. Retrain. Relaunch.
375 ideas. 17 weeks. No tuition, ever.
Vets Who Code is a veteran-run 501(c)(3). The accelerator is free, remote, and we don’t take a share of your first paycheck.
Free · Remote · 17 weeks · EIN 86-2122804