👋 Hey friends,

This is the final newsletter of the year! Here’s what I’ve got in store for you this week:

Let’s get into it 👇

I started making and posting content this year in May and it’s been an incredible journey. I’ve had so much fun making videos, reading up on obscure data structures and algorithms, planning on call scenarios (or reciting them from memory), and just reading your comments. As a fun way to celebrate the year, I jotted down some quick numbers to share:

Honestly, I’m just grateful. If you’ve ever watched a video, shared a post, replied to an email, or told a friend about the newsletter… thank you.

Even safe migrations can go sideways if they're poorly timed, rushed, or not coordinated. Schema changes are as much a team process as they are a technical task - and the way you structure your workflow can dramatically reduce risk.

Here are some practical habits that make migrations safer, easier to deploy, and less likely to surprise anyone at 2AM:

Naming your migration AddStatusToUsers is fine, but leaving comments inside the file explaining why you're doing something is even better - especially if it's a workaround or involves delayed backfilling.

Example:

Your future teammates (or future you) will thank you.

If your application code expects a column or table that hasn't been deployed yet, it can break production if the app deploys before the migration runs - or if the migration fails halfway through.

Safer pattern:

This requires separating schema and code deploys when they're tightly coupled - or using runtime conditionals (e.g., feature flags, presence checks).

In case you missed it, Anthropic published a new open standard: Agent Skills. It’s an interesting idea, but also an incredibly simple one: it’s just a folder that contains SKILLS.md (instructions + metadata), plus optionally scripts, references, and assets/templates. The power behind it is to “write once, use everywhere”, and in doing so your team’s procedural knowledge becomes a reusable package that agents can load when relevant. Instead of building a different “agent” for every niche job, you keep a general agent and give it a library of skills (like internal playbooks). That makes agent behavior more consistent, repeatable, and shareable across tools/teams. I haven’t had the ability to try this out yet, but will do so soon. In case you’re using something besides Claude, Codex also has been building off this standard and Copilot apparently supports it as well.

That’s a wrap for this week as well as this year. If something here made your day smoother, feel free to reply and tell me about it. And if you think a friend or teammate would enjoy this too, I’d be grateful if you shared it with them.

Until next year,
Steven