Nicholas C. Zakas

I'm excited to share my first new book (or e-book) in over five years! Understanding JavaScript Promises explains not just how promises work but also how to use them in the real world. And the best part? It's free. Grab yours now. bit.ly/promises-ebook

@anxietytech AWS console at least feels like I know where I’m at. Azure feels like going into a dungeon without a lantern.

@slicknet I mean AWS console is basically everything is 1-level deep and it’s also inscrutable

Having spent the past couple days both in the Azure dashboard and Edge extension store, it's clear Microsoft *loves* taxonomies. Every page is a deep swim through multiple levels. It all makes sense at a global level, but the UX is such a slog. Big company UI at its worst.

My experience publishing browser extensions: 1. Firefox - super easy, barely an inconvenience. 2. Chrome - multi-page questionnaire, pay a fee, give us your firstborn. 3. Edge - dig through endless menus to find a screen that doesn't work. 4. Safari - YOU MUST BUY A MAC!

Big shout out to @slicknet for this now impossible to find article about “Code review workshops" that definitely deserve a second look in the AI age web.archive.org/web/2016030509…

@anxietytech Wow, blast from the past.

Wrote an article about code reviews in the AI software development times we're in, mostly in response to another article about code reviews I found really insightful dev.to/xjamundx/some-…

Yup! The models are pretty good at evaluating the code against the spec to find digressions and updating them. When I plan the next feature, I have a subagent look for existing related tech specs and summarize them to give the tech spec writer agent a good foundation to start from. And when stuff breaks, I personally find it helpful to look back at the tech spec to figure out what’s going on.

@slicknet I currently keep them til my final commit and then delete them when sending PR for review. I worry about polluting context if the docs get out-of-date in future. You keep them in the repo? And keep them all up-to-date?

If you have AI writing tech specs for you, be sure to also have AI update the tech spec after implementation to ensure it's up to date with the final approach.

@frontstuff_io There is. This is the same pain a lot of folks feel when moving into a tech lead or staff+ role for the first time — less coding, more oversight. We’re now asking it of everyone whereas previously you’d have more experience before needing to make such a shift.

@slicknet Great read! These predictions seem quite plausible. I believe the shift from shipping code to shipping value is beneficial, though there will be intense growing pains.

Coding agents are commoditizing the skills we built our identity on. I wrote about what that means, why it hurts, and what might be on the other side. sarahdayan.com/blog/whats-lef…

@pamelafox Ah see I keep the spec as an artifact to refer back to. It can also be helpful context for an AI doing code updates. I keep all PRDs, tech specs, and task lists.

@slicknet I find I can spend a few hours ensuring spec and code are up to date, at least when I've done a prompt like "review code and make sure it is 100% consistent with doc" and vice versa. I did it in that case as it was my public doc too, but usually I just delete the spec.

I tried uploading a scanned PDF to Google Drive and the OCR to a doc was awful. I asked Claude Desktop how to get a clean conversion. It said, "give it to me." Five minutes later, I have a damn near perfect doc file. 🤯

Actual steps to turn it off: - Hold the light button for three seconds - Press 7 - Press 2 (would be 1 to turn on) - Press Start/Set 🤷‍♂️

Never thought buying a new oven would involve and extensive internet search for "disable wifi on oven."

One of my goals with Bredbox was to make the dev environment run 100% locally. That came in handy yesterday when I had three hours without internet.

“Artisanal software engineering” — folks, we finally got a name for the thing we’ve been doing for decades.

My internet was out for three hours this morning. During that time, I had to revert to artisanal software engineering (no AI). And you know what? It was kind of fun. I got less done but I was able to get deeper into the code. I also felt calmer without all the inline suggestions popping up constantly.

I find myself using GPT-5.3-Codex a lot more now in VS Code because it's the only one that consistently reads nested AGENTS.md files. This makes the experience so much smoother.

Using AI to write code has made me realize there are two types of code I've traditionally written: 1. Code that I want to know how it works 2. Code that I want to exist but don't care how it works I delegated 100% of the second category to AI now. Maybe 50% of the first.

The Supabase MCP server is really nice. I use Bredbox's production MCP server in read-only mode which enables the agent to find problems more easily. It just debugged an OAuth error in production by pulling the logs. Saved a ton of time.