Elias Puurunen

The world wins thanks to UWaterloo grads We should do everything we can to incentivize them to stay

Reminder: you can make things and use them to make money. This little hardware box takes an SDI feed and encodes into SRT. It performed like an absolute champ on a recent gig.

This is awesome. And we need to do MORE of this to grow Canada's economy. We need to be investing into energy, AI, domestic manufacturing... Let's go. We have a country to build.

So every Canadian maker who needs boards is going to buy from Will, right?

Canadian entrepreneurs finally saying the quiet part out loud AND being heard. Canada's business environment is too conservative and takes zero risks.

Everyone of a certain age now feel old.

My X feed is now hardware manufacturing and the most optimistic AI people around. Loving every second. Keep on building, you beautiful people.

I live near a data center. I wish there were more of them. We've been surrounded by data centers for a while. They're not noisy outside. This phase feels like the Final Boss of NIMBY.

To do list item checked off: make a SimTower clone.

What do we need to do to scale this up? Dude is manufacturing hardware boards in canada. We need MORE OF THIS. (Also that lab is giving me serious envy and I love it.)

You can just build hardware and use it to run shows.

Linus Torvalds' takes continue to be gold:

97% of self help books could be replaced by the following cover:

Gateway is coming along and it is snazzy.

Making hardware appliances is fun

Mini PC dry fit is done. Next up are the IO shields and heat set inserts!

I don't know what this late 90's art style is, but it needs to make a comeback.

I need recommendations on a good starter FPGA. Preferably something that can do SDI or HDMI input and/or output.

Tonight I was going through some code Codex had added to one of our APIs. And it looked... grafted on. It worked. But it was convoluted, awkward to read, and didn’t match the aesthetics of the existing codebase at all. Eventually I reverted the entire change and told Codex, essentially: Look at the rest of this code. Match its style. Match its simplicity. Solve the problem like this codebase would solve it. The result was dramatically better. I’m not claiming to be some sort of coding genius. But after working with Codex for a while, I’m getting much better at focusing it on how I want software built. AI coding tools remind me a lot of a junior hire who is incredibly excited to do literally anything. In the AV world, it’s the tech working their first event where even getting to tape down cables is cool. That enthusiasm is fantastic. But you still have to focus it. You figure out what they’re great at. You give them examples of good work. You review what they produce. And over time, you nudge them away from the things they’re merely okay at. Yes, vibe coding is fun. But I think the next step is developing some taste around AI-generated software. We need to inspect the code. Not just: Does it work? But: Does it belong here? Is it simple? Can another human understand it? Does it make the next change easier instead of harder? Because we’re now writing code for two audiences: Humans who need to understand and maintain it. And AI agents that are going to come back and work on it later. It'll be our taste that guides us to good software, with the help of AI.