Dan Ubilla

Leading Developer Productivity @asana. Writing about AI dev tools & platform orgs. From VA → NYC. Formerly @wearevts, @espn, @3dvia

[Dan Ubilla @dubilla]

@frankmikesmith Nothing like shipping to learn what your users want. Nice work!

[Dan Ubilla @dubilla]

@princeserafino6 @Sam_Vecenie McAdoo 70s Clips is a pull

[Dan Ubilla @dubilla]

@noahzender Well, here’s what I’m doing tonight. This rocks. Thanks!

[Dan Ubilla @dubilla]

@noahzender What are the Obsidian plugins that help you set this up?

[Dan Ubilla @dubilla]

@LarryWilliamsTI A lot of talk about Kornet, deservedly. But that close out by Wemby is the kind of thankless play that changes outcomes

[Dan Ubilla @dubilla]

@AZuckermanVA Honestly, could be a Teen Vogue headline from 2016 (complimentary)

[Dan Ubilla @dubilla]

@allgarbled Nothing is verifiable from 30,000 ft. Part of the software engineering job is making more of the domain verifiable.

[Dan Ubilla @dubilla]

@samsaffron @bcherny Contaminates how?

[Dan Ubilla @dubilla]

@jxnlco @jacob_posel Chad?

[Dan Ubilla @dubilla]

@nateberkopec 🎯 Always "Make the thing which makes the thing". It was before all of this. It will be after.

[Dan Ubilla @dubilla]

@AthenePromachos @jonathanmb32 @Srirachachau Yeah, I wonder if this felt more prominent for those who grew up with the early seasons, trusted the parodies, and by season 10 had enough current cultural understanding to question them

[Dan Ubilla @dubilla]

@VicVijayakumar Loss aversion always comes through

[Dan Ubilla @dubilla]

@HornetsBeliever @rBillSimmonsPod Ok, I’m back in

[Dan Ubilla @dubilla]

@benhylak Follow the incentive structure

[Dan Ubilla @dubilla]

I wrote up the strategy, mental model, and mocking process. Really fun! https://t.co/wwHPVXO7AB

[Dan Ubilla @dubilla]

I built a small weekly review agent to summarize my week. It takes input from a few specialist outputs and feeds an agent that runs Sunday evenings.

But the seven day feedback loop was brutal. So I started looking into evals.

[Dan Ubilla @dubilla]

The fun part about working with this strategy was the mocking. I mocked various failure modes with JSON fixtures, then adjusted the main agent. Fixtures like: a sub-agent returns an error, a sub-agent returns an empty set, one input dominates the others.

[Dan Ubilla @dubilla]

@spencernusbaum_ We did it! https://t.co/DrSV3gX6qh

[Spencer Nusbaum @spencernusbaum_]

The Nats have lost 15 games in a row in which they had a chance to get to .500. Does it matter?

That depends on who you ask. More in here, free and unlocked to read:
https://t.co/9kCOqkfMu6

[Dan Ubilla @dubilla]

@brynary I might even go the other way around where tech debt, somehow, is more easily defined for an agent than functionality