Waldek Mastykarz

@waldekm

I help teams evaluate and improve how well AI coding agents can use their technology

Alphen aan den Rijn, NL

Waldek Mastykarz
@waldekm
9h
Funny how we obsess about the UI and taste, while agents don't care about any of it. How long until UI becomes the back stage that's there but isn't where the audience hangs out?
Waldek Mastykarz
@waldekm
10h
Your coding agent scored 100% on its eval, and every criterion passed. Then someone tried to build the code it generated...
Be honest about what your evals prove
From blog.mastykarz.nl
4
Waldek Mastykarz
@waldekm
10h
This 👇 Stay curious, chat!
Keerthana Gopalakrishnan
@keerthanpg
Oct 17
If you’re the type of person who is curious and gets dopamine hits from learning things, talking to an AI can become super addictive. This is the type of pleasure no human can give you. No human knows so many things and no human will have the patience to explain so many things,
Waldek Mastykarz
@waldekm
10h
"Does the solution follow best practices?" Your LLM judge answers without hesitation and explains its verdict convincingly. So which best practices did it check?
Your eval judge doesn't know what you mean
From blog.mastykarz.nl
Waldek Mastykarz
@waldekm
11h
You wrote your agent skill to fix what the agent kept getting wrong. What happens to those instructions when a new model stops getting it wrong?
As models get better, your agent tools should get smaller
From blog.mastykarz.nl