I've been mulling over some disparate threads that have come together in a fun little experiment.

First, I have been searching for a project that uses satellite imagery.

Second, climate-tech is interesting to me. In the current AI zeitgeist, I don't hear much being done in the space. I'm sure there are some things going on, but it must be getting drowned out. In any case, I have my concerns about what AI means for our climate future (data centres, resource usage, etc.), and if I'm going to use it, I want to find ways that could be net-positive.

Finally, I've been thinking a lot about what can be collectively achieved in markets where people are traditionally isolated.

What came up for the last point was: residential solar panels.

Installing solar on your house​

I've been through the process of getting solar panels on my house. It's generally the same as buying any large-ticket item:

1. You solicit quotes from a number of vendors
2. They present extremely similar products, at similar prices
3. You pick one

To be fair, #2 comes with a bit more from higher-end vendors. They'll actually estimate your power generation by:

1. Finding your house on Google Maps
2. Overlaying models of their panels 'on' your roof. They try to maximize how many panels can fit, and take into account things like panel orientation, sun path, and tree cover.

Financing​

A quick word about actually paying for it.

At the time we were in the market, the government had an incentive program which included a 0% interest loan on residential energy upgrades.

We didn't find solar vendors to be particularly flexible with their prices, in part I think because they wanted to get the most from the incentive program.

That said, the program is a reimbursement, so we had to pay out of pocket throughout the install, and wait for the reimbursement well after the panels went on the roof.

Collective​

Reflecting on the process, I began to wonder what it might look like if the whole neighbourhood went in on upgrades at the same time. Perhaps we could have negotiated better deals if the installers knew that they would have locked-in sales for many houses at once.

That thought quickly went to - what if we could negotiate at a municipal level? What would it even look like?

Time to fire up Claude Code.

The map!​

(takes a while to load)

About the map​

This embed was generated by a python script - here's the general pipeline description:

1. Fetch boundary — Municipality polygon from OpenStreetMap via osmnx
2. Load buildings — Microsoft Canadian Building Footprints, bbox pre-filtered then clipped to boundary
3. Fetch irradiance — Monthly GHI/DNI/DHI, temperature, and wind speed from the NASA POWER climatology API (2001-2020 averages)
4. Compute generation — For each building: project to UTM 18N for accurate m² area, determine usable roof area, transpose irradiance to plane-of-array via pvlib, apply temperature derating, calculate annual kWh
5. Classify — High / Medium / Low based on generation and roof area thresholds
6. Visualize — Folium map with color-coded markers, popups, layer toggles, legend, and a collapsible summary panel; console summary printed to stdout

It's not perfect, but wow! I wasn't expecting to get anywhere close to making something like this. If nothing else, it could inform some napkin math about the energy potential if solar gets widespread.

The Estimation​

Energy Generation
Total	4149.4 GWh/yr

Plugging this into: https://www.epa.gov/energy/greenhouse-gas-equivalencies-calculator

Dreaming​

Right now, solar installs are just around break-even, maybe a bit better, depending on your utility company, if you were part of the incentive program, if you upgraded other components, etc. etc. Hardware prices continue to go down; there's probably a floor on labour prices to install panels - but if that shifted even just a little more into the "obviously a good financial decision", maybe people could adopt en masse.

Part of that process starts with ballparking what that future could look like!

From here, some ideas start to form for me:

• modeling what emissions impact battery install saturation looks like
• modeling system/install financials could look like
• starting conversations with politicians about a green future, and how we could incentivise adoption, or negotiate collectively

But that'll be for another time.

Why does a clock tick and tock?

This was a phrase I read in a short story that stuck with me. I had an intuition for what it might be, but I wasn't sure.

The short answer is that it's sound from an escapement, with different sounds being produced on the swing and return.

This made me wonder about different ticking patterns. 60 divides cleanly by 2,3,4,5, and 6.

How weird would it be to have a clock that accented the tick every 3 beats? Or 5?

So I vibed up this part-clock / part-metronome. It's not particularly useful, but I did enjoy staring at the patterns getting in and out of sync when all clock signatures were enabled. Also, there is audio as well, if you care to leave the volume on.

Interactive demo (click to start audio):

As you might be able to see with the timestamps, trying to get into a cadence for longform writing hasn't been working.

I have a lot I'd like to write about but there is a bit of friction.

I'm hoping that allowing myself (via use of arbitrary tagging) to post shorter, less polished notes, will grease the wheels and get this going again. Onwards!

Over the last few months, I've had the opportunity to revisit some AI coding tools and see how they have evolved. My first impression was that they are now much more capable than they were even just a few months ago.

My experience is rather bimodal: it's magic when it works, and outright disillusionment when it doesn't. The good news: if you know what you're doing, you don't need it to work 100% of the time.

I recently had the shower thought about the phonetic similarity between "cell phone" and "self-own".

Naturally, there is an Urban Dictionary entry for this from 2021, so it's hardly original.

Cell Phone Self Own

1. When your phone's autocorrect absolutely savages you.

2. When you are texting in bed and accidently drop your cellphone on your face.

I posit that merely using a cell phone these days is a self-own.

A few days ago, OpenAI released Operator, their first browser-use agent. Their demo left me wondering - who is this for?

Human behavior seems bimodal when it comes to content — people largely fall into either a creation or consumption mode, with few striking a true balance.

In my roles as a coach and an engineering manager, I often find myself in situations where I need to help people navigate their feelings. Whether it's an athlete feeling frustrated with their performance or a team member feeling overwhelmed by their workload, my job is to help them find a way through it.

Often, I think of myself as a counterbalance.

Hardly a hot take, but I love working in high-trust environments. I think most people do. It's nice to be treated like an adult, to be trusted to do your job, and to be given the freedom to make decisions.

I want to live in a high-trust society. Unfortunately, experiences in various contexts have made me reflect about the importance of trust in our lives and how it can be eroded.

As a weightlifting coach, my favourite complaint from competitive athletes at the gym is "ugh, this is so heavy!".

Yeah. That's the point!