12 posts found
What a large language model actually predicts
A model does not look anything up and does not decide what is true. It estimates which token comes next. Almost everythi...
Tokens, not words: how a model reads your text
Models do not see characters or words. They see tokens — and once you know how text becomes tokens, several odd behaviou...
The context window is a budget, not a memory
Bigger context windows did not give models memory. They gave you a larger envelope to fill on every single request — and...
System prompts: the instructions the user never sees
A system prompt sets the rules for everything that follows. Written well it is the cheapest quality improvement availabl...
An agent is a loop, not a personality
Strip away the branding and an agent is a small loop: call the model, run the tool it asked for, feed the result back, r...
Tool calling is an API contract the model can read
A tool definition is documentation with a schema attached. Write it for a capable colleague who has never seen your syst...
Why agents fail on long tasks
An agent that handles five steps beautifully can fall apart at twenty. The reason is rarely reasoning — it is that the c...
Giving an agent memory without giving it amnesia
Memory is not a feature you enable. It is a set of decisions about what to keep, what to summarise and what to let go —...
When a workflow beats an agent
If you already know the steps, do not ask a model to rediscover them on every request. A fixed pipeline with model calls...
What an AI agent can actually do with a design system
Given a real design system, a model stops inventing interfaces and starts assembling them. The difference is not the mod...
Prompt caching: the cheapest speedup you are not using
If every request begins with the same two thousand tokens of instructions, you are paying full price to resend them. Ord...
The cost model of an AI feature
Per-token pricing looks trivial until you multiply by retries, conversation history and the context you resend on every...