how-ai-thinks

9 articles · newest first

Flat illustration split in two: a person with a laptop sending a message to a distant cloud on one side, the same person with a desktop PC glowing on the other

Local model vs API: what changes when the AI runs on your machine

Lesson 9 of how the AI thinks: the same kind of model can run on a company's servers or on your own computer. What changes: privacy, cost, speed, and which models you can use.

Sam Ortega · 1h ago · 4 min

Flat illustration of a friendly robot at a workbench choosing a tool from a pegboard while a person watches with a checklist

Build

What are tools and agents? When the AI starts doing things

Lesson 8 of how the AI thinks: a model on its own can only write text. Tools let it ask your app to act, and agents decide their own next step. What that changes, including the risks.

Sam Ortega · 1h ago · 4 min

Flat illustration of a person at a desk working through a long scroll of notes before writing a short answer card

Build

What is a reasoning model? When 'thinking first' is worth it

Lesson 7 of how the AI thinks: some models write out working before they answer. What that buys you, what it costs, and when to switch it on.

Sam Ortega · 1h ago · 4 min

Flat illustration of four package boxes on a conveyor, one of them transparent and empty, with a person holding a magnifying glass

Build

Why AI invents a library that doesn't exist (and how to catch it)

Lesson 6 of how the AI thinks: code models sometimes name packages that don't exist, and attackers can register those names. The research, and the one habit that catches it.

Sam Ortega · 1h ago · 4 min

Flat illustration of a person turning a big dial while three slightly different paper outputs come out of a machine

Build

Temperature and sampling: why the same prompt gives different code

Lesson 5 of how the AI thinks: the model picks each next token from a list of likely options. Temperature decides how adventurous that pick is, and on newer models you may not get to choose.

Sam Ortega · 1h ago · 4 min

Flat illustration of a person tapping one word block while thin glowing lines connect it to a few other blocks

Build

Attention, explained visually: how the AI links your words

Lesson 4 of how the AI thinks: when the model reads "it", how does it know what "it" means? A toy picture of the idea behind every modern model.

Sam Ortega · 1h ago · 4 min

Flat illustration of a conveyor belt of coloured blocks, new ones added on the right, the oldest tipping off the left end

Build

Why the AI 'forgot' what you told it: context windows explained

Lesson 3 of how the AI thinks: the model has no memory of its own. What falls out of the window, what gets summarised, and how to keep what matters in view.

Sam Ortega · 1h ago · 4 min

Flat illustration of a person dropping coloured blocks into a nearly full transparent glass box with a fill line

Build

What is a context window? The AI's working memory

Lesson 2 of how the AI thinks: everything the model can see at once, including its own answer. Why it fills up, and why bigger isn't automatically better.

Sam Ortega · 1h ago · 4 min

Flat illustration of a person cutting a long ribbon into small coloured blocks of different widths with giant scissors

Build

What is a token? How the AI actually reads your words

Lesson 1 of how the AI thinks: the model doesn't read words, it reads tokens. What they are, roughly how big, and why your bill and limits are counted in them.

Sam Ortega · 1h ago · 4 min