iyernd.pythonanywhere.com

Add 17 hours of storage and 600 MW of solar and see what happens

Yea, it’s more expensive.

Talk about the new 45Z proposal lol

CATL is maintaining their massive sodium investments and starting to mix the chemistries because in the long-run that pivot starts to make sense.

That said for a US-based start-up, definitely seems like they made the right call for now

I think we'll see more interest in sodium when we actually hit the LiB floor, which definitely hasn't happened yet!

If the infra can be reused, then more innovation will happen in lithium which can ultimately be ported over to sodium.

How do F-gas GWP weigh in here? Is it worth having less performance for better refrigerants?

I asked it to make a dash dashboard in python for most of this, and then after that just ask and run and debug in a loop until I get the features

Cursor + claude_sonnet /gemini mostly.
Maybe o1 on occasion for planning specs

Hard to understand on a paper, but you can FEEL it on a dashboard.

It's a vision I had for year, to life in 2 weeks.

Across companies and organizations these kinds of realizations are happening now silently.

Energysky, early days.

I now understand what it feels like to see decreasing marginal value. That high levels of hourly matching has non-linear paths (e.g. optimal from 80% is not on the road to optimal for 90%). 100% is very very hard. How storage has more or less capacity value over a year. The impact of MACRS.

Over the past few weeks I built the tool I wish I had, to understand how renewables, clean firm, storage, and demand response interact in a grid in an interactive UI.

iyernd.pythonanywhere.com

Analysis heavy workflows are the most obvious, but optimization, games, productivity tools, visualizations, compression of the vast information flows.

This is about transforming work, not automating it away.

This could be a dramatic change to the Internet and a return to form - if small creative teams can replicate the work of behemoths, the Internet becomes a lot more interesting of a place and potentially less centralized by reducing barrier to entry dramatically.

Impossible task become trivial. Ideas that would take a year, take a day. Features ship in minutes and are highly configurable.

And this is the worst it will ever be.

That’s wild!

Sweet horror

LOL yep

Also, AI training is not exactly "base load" due to all-reduce operation training. Minute basis oscillations in power are just arbitrage opportunities waiting to be realized!

(this is part of why xAI has a megapack on site, to smooth out these oscillations)