DIY experiments. Every week.

Raw projects. No spam.

I Built a Tool to Predict If My Next Side Project Will Ruin My Weekend

Cover Image for I Built a Tool to Predict If My Next Side Project Will Ruin My Weekend

The Lie We All Tell About Time

When I worked at a digital agency, every new client came with a new team, a new mess of requirements, and the same damn question from the suits:

“How long will this take?”

I’d give an answer.

A good-sounding one. One that bought us time and gave them hope.

It never held up. We were always late.

I Don’t Have Time to Be Wrong Anymore

These days, it’s not project managers breathing down my neck.

  • It’s my daughter asking for park time.
  • It’s dinner with the people I love.
  • It’s me - trying to cram another stupid project into an already overloaded schedule.

I still lie to myself. “Yeah, I can probably build this in an hour.”

Nope. I’m wrong. Again. I overpromise. Again. And I pay for it. Again.

I Want an Agent That Works for Me

I want local agents. Agents that work for me, not for some SaaS startup with a roadmap and an upsell.

But for agents to help me, they need data. They need my history. They need context about how I actually work - not how I think I work.

I did this with my expenses first. (See: AI-powered expense tracker)

Now I’m doing it with my time.

My projects. My pace

CodeChrono: Track. Learn. Estimate. Repeat.

It’s called CodeChrono.

  • start begins a session

  • stop ends it and summarizes what I built using a local LLM

  • estimate lets me ask: “How long will this feature take?”

  • A dashboard shows what I’ve built, how long it took, and whether I’m speeding up or slowing down

I run it all from the terminal.

“Faster, faster, until the thrill of speed overcomes the fear of death.”

— Hunter S. Thompson

I have a bias for speed. Speed isn’t recklessness - or at least, not always.

  • It’s compressed feedback.
  • It’s rapid iteration.
  • It’s momentum.

But speed needs a mirror — something to reflect back whether you’re actually moving, or just spinning your wheels.

That’s what this tool is for.

My New Rules for Building Fast (and Staying Sane)

  • If a feature took 4 hours last time, it probably will again.

  • If a project took 2 weekends, I won’t squeeze it into one.

  • If my estimation pattern is trash, I’ll see it on the chart.

  • And if I’m getting faster, I want to know that too.

Run It Like You Mean It

🔗 GitHub Repo

python run.py start -m "build settings page"

# ...do your work...

python run.py stop

python run.py estimate -m "add payment flow"

python run.py query -m "What took the longest last week?"

streamlit run dashboard.py

Speed Needs a Mirror

I’m not trying to optimize every second of my life.

I just want to build the things I care about — as fast as I possibly can.

And now, I’ve got the data to back it up.