Notes on ai tools, written after living with it rather than reading about it.

There is plenty written about habits for fact-checking and little you can act on. This piece stays concrete: every action can be done on the study…

If you only read one paragraph, read this: handling private data depends on how the situation is set up, not on how you feel at the time. Arrange the…

Advice on specifying output format tends to split in two: impossibly complex, or sold as one click. Reality sits between. It needs some patience and…

If you only read one paragraph, read this: strategy across several turns depends on how the situation is set up, not on how you feel at the time.…

Let us set the limits of building a prompt library first: it reliably improves devices that stay pleasant to use, and it does not deliver overnight.…

If you have already tried asking for a comparison table and dropped it, the fault is usually in the setup: a starting point set too high, an…

Advice on asking for an outline tends to split in two: impossibly complex, or sold as one click. Reality sits between. It needs some patience and…

Let us set the limits of adjusting tone for email first: it reliably improves devices that stay pleasant to use, and it does not deliver overnight.…

If you have already tried a route for learning new tools and dropped it, the fault is usually in the setup: a starting point set too high, an…

If you only read one paragraph, read this: spotting confident nonsense depends on how the situation is set up, not on how you feel at the time.…

Advice on automating a small workflow tends to split in two: impossibly complex, or sold as one click. Reality sits between. It needs some patience…

There is plenty written about costing token usage and little you can act on. This piece stays concrete: every action can be done on the commute,…

If you have already tried tasks worth handing over and dropped it, the fault is usually in the setup: a starting point set too high, an environment…

Let us set the limits of tasks to keep yourself first: it reliably improves files found in three seconds, and it does not deliver overnight. Anything…

Most people do not fail at writing prompts that specify the task from misunderstanding it. They fail because understanding is not doing. What follows…

Let us set the limits of where rewriting crosses a line first: it reliably improves files found in three seconds, and it does not deliver overnight.…

If you have already tried summarising a long document and dropped it, the fault is usually in the setup: a starting point set too high, an…

Most people do not fail at analysing a table from misunderstanding it. They fail because understanding is not doing. What follows is the doing part:…

Most people want editing a translation to make things better. The workable version of that goal is steadier: a floor you can hit on a bad day.…

Let us set the limits of turning recordings into notes first: it reliably improves backups that survive a power cut, and it does not deliver…