Building a startup is like parenting.

You can plan all you want, but most of it is learning in the moment. Some days, you crush it. Others, you just survive.

The secret? Show up every day, even when it’s hard.

The process matters more than the plan.

Startups and family life feel like two worlds pulling you in opposite directions.

Morning time with the kids. Then, a few hours building & research. By afternoon, I’m back w/ the family, trying to be fully present.

It’s not easy, but maybe that’s the point? Figuring out how to give both your best.

Morning with the kids, a few hours implementing the ‘apply’ feature, then diving into caching research. Wrapped it all up with the rest of the day spent with family.

Building a startup and being present at home is a constant balancing act.

How do you balance deep work and family on days like this?

You don’t need to win on Day 1.

You just need to move faster than everyone else, week after week, with fewer excuses.

Right now, I’m watching our tool, Murphie, go from 'interesting' to 'indispensable.'

And it’s happening faster than I ever imagined.

It’s been unbelievable to go from 'Wouldn’t it be cool if…' to 'We’re almost at parity with some of the best tools out there.' @erodriguez_ai

Not in years. In a short couple months.

The roadmap is clear. The momentum is real. The outcome is inevitable.

You fix one thing, break three others, but each iteration gets you closer. The key is to keep shipping every day.

Refactored more code, added tests, and started locking in features for the beta launch. The codebase feels solid. Focused now on building something that delivers.

Refactored the code, made it modular. Cleaner, faster, better. Murphie’s coming together. No hacks, just solid engineering. LET'S GO 🚀

Observability in LLM apps is non-negotiable. Logs, metrics, traces - if you’re not watching what your app does, how will you know when it’s wrong?

The first step to building a great LLM app isn’t picking an API. It’s defining your workflows. Know the problem, then design the solution.

Show, Don't Tell Loop

1. User gives feedback.
2. Analyze it, improve model outputs, and ship updates.
3. Show users the difference right away.

Most people don’t care about the tech - they just want it to work. And for it to work well.

So let them see the progress. That’s how you build trust.

People don’t want to fill out surveys. They don’t want to write essays. So why do we keep pretending they will?

They will click thumbs up/down and drop one-liners.

So we’re leaning into that. Lightweight feedback that our system parses, categorizes, & acts on.

No friction = better AI.

Processing unstructured security data is the real challenge.

A tip for handling large reports:

•Split text into smaller, meaningful chunks (retain context).
•Store metadata (actor, timestamp, type).
•Embed these chunks for efficient searches.

Once structured, your data becomes your best ally.

Right now, I’m in a season of building - building a startup, raising kids, and learning how to balance ambition with the life I want to create.

Every day brings new challenges, but also new insights. It’s messy, it’s exciting, and it feels like the start of something big.