From our blog

Check out our latest news and updates.

What a large language model actually predicts
What a large language model actually predicts
12/09/2026 — [email protected]

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
Tokens, not words: how a model reads your text
10/09/2026 — [email protected]

Models do not see characters or words. They see tokens — and once you know how text becomes tokens, several odd behaviou...

Why temperature changes the answer, not the knowledge
Why temperature changes the answer, not the knowledge
08/09/2026 — [email protected]

Turning temperature down does not make a model more accurate. It makes it more repeatable — and confusing the two is how...

Training, fine-tuning and prompting are three different tools
Training, fine-tuning and prompting are three different tools
04/09/2026 — [email protected]

Teams reach for fine-tuning when they need context, and for prompting when they need behaviour. Knowing which problem ea...

The context window is a budget, not a memory
The context window is a budget, not a memory
02/09/2026 — [email protected]

Bigger context windows did not give models memory. They gave you a larger envelope to fill on every single request — and...

Why models hallucinate, and what actually reduces it
Why models hallucinate, and what actually reduces it
31/08/2026 — [email protected]

Hallucination is not a glitch that a better model will one day remove. It is what generation does when it has nothing to...

System prompts: the instructions the user never sees
System prompts: the instructions the user never sees
29/08/2026 — [email protected]

A system prompt sets the rules for everything that follows. Written well it is the cheapest quality improvement availabl...

Structured output beats parsing prose
Structured output beats parsing prose
27/08/2026 — [email protected]

If your code needs a value from a model, ask for JSON and validate it. Regexing an answer out of a paragraph is a bug wa...

Tool calling is an API contract the model can read
Tool calling is an API contract the model can read
21/08/2026 — [email protected]

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
Why agents fail on long tasks
19/08/2026 — [email protected]

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
Giving an agent memory without giving it amnesia
17/08/2026 — [email protected]

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 —...

Generating layouts that respect a grid
Generating layouts that respect a grid
09/08/2026 — [email protected]

Generated layouts usually look almost right, and "almost" is where the eye catches. Most of the fix is stating the rules...