On Joining OpenAI

And the next chapter of Artificial Ignorance.

Next month marks three years of publishing Artificial Ignorance.

I didn’t start this with a master plan. I started it because I wanted to understand what was happening in AI - and I knew I’d learn faster if I forced myself to write clearly (and regularly) about it. Over time, it turned into a small but meaningful rhythm: read, think, tinker, write, repeat.

Now, three years later, that path has taken me somewhere fairly unexpected: to OpenAI.

I joined OpenAI

I’m on the Developer Experience team, which means I’ll help developers learn and build with OpenAI’s technology through docs, guides, tutorials, examples, and more.

It’s no small task: it feels like almost every day the company’s surface area is getting bigger, and in the last year alone, the developer relationship with OpenAI has meaningfully shifted. It used to be that (as a dev) you worked with OpenAI’s API, or used GPT-X in Cursor; now you have Codex and ChatGPT Apps, meaning you can work with/on OpenAI and never touch core API docs.

Still, I’m incredibly excited to be joining. For the past three years, thinking about this technology has kept me up at night, and I’ve thoroughly enjoyed building with it. I’ve been the kind of person who reads dozens of articles about AI before bed because I genuinely want to download as much context into my brain as possible. Now I get to channel that same curiosity into work that helps other people build.

That said, joining OpenAI means Artificial Ignorance needs some clearer guardrails. So a few things will be evolving around here.

What’s changing with Artificial Ignorance

News roundups are going away

The biggest change: I’m discontinuing the news roundups.

I still love following AI news. I still read it constantly. But covering the news publicly, in the same way I’ve been doing, creates too many edge cases and too many ways for the incentives to get weird. Even when you’re careful, “commentary” can slide into “amplification.” And for whatever reason, everything OpenAI does receives an inordinate amount of attention - meaning the margin for misunderstanding gets a lot smaller.

There’s also a simpler truth: news roundups are increasingly commoditized. Plenty of people (and AI tools) do a great job summarizing the week. I don’t think that’s where Artificial Ignorance is most differentiated, and it’s not where I want to spend my limited writing time going forward.

Deep dives are staying (with a more technical focus)

The deep dives are staying. If anything, they’ll become the heart of the writing here.

Most readers already seem to find the deep dives more valuable than the roundups, and they’re also the type of writing I’m most proud of: slower, more thoughtful, and more durable than whatever happened this week.

Over time, I also want to put more emphasis on engineering-specific content - practical writing for people building with AI. Not “ML nuts and bolts,” and not industry hot takes, but the craft layer in between. Sometimes, that might involve primers on topics like RAG and guardrails; other times, it might lean more towards outlooks on how AI is reshaping coding as a whole.

For better or worse, there’s far more happening in AI than any single person can cover. I don’t expect to run out of things to write about, even with a somewhat narrower focus.

Chat: a new experiment

One of the things I loved most about the roundups was a place to formulate small ideas about how the space is advancing. I want to experiment with doing that in a smaller, more intimate setting: Substack Chat.

My initial plan is to use Chat as a place for (paid) subscribers to have more discussion and interaction - especially for things that don’t merit a full post, but are still worth talking about. Two concrete things I’d like to try:

• Office hours: a recurring thread where you can ask questions about building with AI. Architecture, evaluation, product integration, agent design, and all the practical “how do I actually do this?” questions.

• Topic discussions: a single thread to discuss a trend or story in AI. I’m stepping away from public news roundups, but I still want a space for thoughtful conversation about what’s happening.

This is a real experiment, and I’m going to treat it like one: try a few formats, see what works, iterate, and keep what’s valuable. It might work great; it might flop!

TL;DR

If you prefer the scannable version, here it is:

• Less: news roundups

• More: deep dives, especially engineering-specific content for builders

• New: Chat experiments (office hours + topic threads)

• Same: the intent of the newsletter - pragmatic, curiosity-driven writing about how this technology actually works in practice

And just to be abundantly, explicitly clear: you shouldn’t expect inside information here. I won’t be sharing anything confidential or non-public, I won’t amplify leaks or rumors, and I’m unlikely to post anything that reads like a corporate endorsement. The goal stays the same: write what’s useful, grounded, and true.

---

Thank you for reading Artificial Ignorance - whether you joined three years ago or last week. I don’t take it for granted that you’ve chosen to spend your attention here.

I started writing this newsletter to teach myself first, and then to teach others. Perhaps selfishly, I wanted to learn how all of this worked; building an audience wasn’t the primary goal (and still isn’t). I’ve been driven mainly by my own intellectual curiosity, because this is such an incredible technology happening in real time.

I don’t plan on changing any of that anytime soon. I’ll keep writing what I would’ve wanted to read.

And if you’re a builder, I’d love to see you in Chat.