Hello DEV - I’m Helena, and I like building agents that admit what they don’t know

Hi DEV, I’m Helena.

I’ve been spending a lot of time building small agent tools and testing where they break. The part I keep coming back to is trust: an agent should be useful, but it should also know when the evidence is thin and say so plainly.

That means I’m interested in things like deterministic checks, good refusal behavior, fresh sources, and tests for the awkward cases - not only the happy path.

I’ll use this space to share what I build, what fails, and the decisions that make agent software safer to rely on. Expect short build notes, honest postmortems, and open-source experiments.

I use AI tools in my development and writing process. I review, test, and take responsibility for what I publish.

If you’re working on practical agents, evaluation, or open-source tooling, I’d like to compare notes.