Oliver Hughes

Answered Streaming responses: how do you validate a schema you have not finished receiving?

Adding to the above — measure the failure rate before you optimise anything. I spent two weeks on prompt tuning for a step that turned out to be failing 0.4% of the time, which was well inside what the downstream process already tolerated. Instrument first. The expensive problem is rarely the one that feels most annoying.

Answered Best approach for prompt-injection defence in a user-facing app?

You do not need 200 labelled examples to start. Take 20 real inputs, run both prompts, and look at the outputs side by side yourself. Twenty is enough to catch a regression that matters, and you will build the labelled set as a by-product of the reviewing. Blind yourself to which output came from which prompt. Otherwise you will find the new one better, because you wrote it.

Shared a bide Turn a rough request into a task spec a coding agent can run

Convert a one-line feature request into an agent-ready spec with testable acceptance criteria, file boundaries, verification commands and stop conditions.

Shared a bide Follow-up email after an interview that adds real value

Writes a short post-interview thank-you email that references the conversation, repairs a fumbled answer, adds one useful thing, and schedules a check-in.

Shared a bide How Do You See Me? A Grounded Read of My Self-Description

The popular 'how does AI see me' reflection, done properly: evidence-based observations from your own words, confidence labels, and no horoscope flattery.