John’s newsletter 3/52

John’s newsletter 3/52

Welcome to week 3 of my attempt to send this newsletter on a weekly basis. So far I’m finding it to be a slog. I keep putting it off until it’s too late.¹ Hopefully practice make perfect!

Momentum

When I write, the results are more successful if I've been writing other things recently. To write a blog post that people will be interested in, I first have to write five other blog posts to get into the swing of things. Then, it usually happens that after I write one that people find interesting, I’ll stop writing, and starting again requires a lot of deliberate effort.

So some advice to myself: If you’re doing something consistently, and it’s going well - try to keep doing it.

This is easier said than done. I personally find it quite hard to prioritise blocks of time to do low-value things like write blog posts. There always seems to be something more important than writing random things for a tiny audience on the internet. For the next month, I’m going to try and more aggressively schedule time do write in my calendar.

Interesting things from around the web

GitHub Copilot exerts downward pressure on code quality

Copilot and other AI tools are meant to lower the barrier of entry to creating software. But maybe lowering the barrier to entry in this way isn’t actually what we want. Copilot has been shown to help people complete tasks rapidly, but a new paper also shows that it increases code churn. The implication is that people are writing lots of code with AI that turns out to not be fit for purpose, and has to be replaced. This could still be a net win for some organisations if replacing the code is also really fast. But if you really care about quality and correctness, it’s probably not good.

When I was at Techstars' London accelerator, I spoke with a CTO who had been hired by a startup with an existing technical product that had been built by a non-technical CEO using Copilot and ChatGPT. It seemed like the majority of their time was being spent trying to understand the mess that had been created. Now, maybe these AI tools allowed this business to get started when it otherwise wouldn’t have. Or maybe they led to the creation of a worthless MVP that was being subject to a sunk cost fallacy and would have been better understood by being deleted.

Practical Betterments

Continuing in the proud tradition of Lifehacker, this site lists a bunch of simple tricks to streamline or improve your life. They range from the smart and effective (lower your boiler temperature) to the insanely marginal (lubricate your keyholes). I love this kind of project, and in fact, I’ve been meaning to publish something similar for a while. Seeing this might be the motivation I needed to actually do it.