Why the most important work comes from people who care about things nobody asked them to care about.

Been receiving lots of questions about this from folks who expect AI engineers to result in a wave of mass unemployment. I think the crux of the question is whether there can be such a thing as "too much code" in the world. As farming productivity skyrocketed in the last few hundred years, we really didn't need all that food, and so the % of the American workforce working in agriculture went from 70% in 1840 to 2% today. But other goods showed a very different trend, and their consumption increased as productivity increased and prices decreased (the famous Jevons paradox). en.wikipedia.org/wiki/Jevons… This was the case of aluminum and solar panels — Jevons paradox tends to be the rule in commodities, and trends shown by food the exception: We consume more than 10x more aluminum today than we did in 1950. The trend for solar panels is even more spectacular — we consume more than 1,000x as many modules today(!!!) as we did in 2002. And the trend is accelerating! My contention is that code will follow the same path, and I think every single big tech company in existence today is one more point supporting this thesis. Consider the Uber app — Uber spends about $3B per year on R&D right now, per their public filings. You could say the Uber app is an object that's cost roughly $10-20B to produce. For reference, the Golden Gate Bridge cost $600M (yes, adjusted for inflation). The Uber app is 20-40 Golden Gate Bridges. We don't realize it because these things are so immaterial, and they seem so small stuck behind the glass of our phones — but software has been very, very expensive. I bet that if you were to do the same math for Google, Microsoft Windows, or iOS, you'd land on something closer to $100B. Each of these companies has voted with their wallet — they need a lot more software, and are willing to pay *a lot* of money to get it. That makes sense when you consider what software is — it's just rules and logic. There is no hard coded number anywhere specifying the maximum amount of logic that a civilization can use — my contention is that the more, the better. Now the question left is — sure, we'll need an infinite amount of software. But won't AI write it all? And here, I appeal to good old comparative advantage. If you have an asset producing billions of dollars an hour (AI engineers will), it will rationally make sense to pay someone 6 figures who can improve this system by 1e5 / 1e9 = 0.0001%. I have a hard time thinking human engineers won't be able to cross that hurdle — they'll just operate on a different layer of abstraction, just like we don't code in machine code anymore, but in high-level languages. And even those still require technical skills — Interestingly, this still applies even in the extreme scenario of "no code," aiming to let even your grandmother build applications. But I think we've all learnt by now that even the no-codest of the no-code products require someone with a programming mindset to build great products.

I’ve secretly 🤫 been using a new product every week that is now live for everyone. Today we are launching Reforge Artifacts and it’s completely free 🆓. Take a look 👇 Artifacts let you access the real work from those who have done it before, so you’ll never have to start from scratch. Some examples: 👆 Product Review Systems from Casey Winters, Tom Willerer, and others 🧠 Product Strategy and Roadmaps from Adam Fishman and Sachin Rekhi 🪜 Career Leveling Guides from Julie Zhou, Barron Ernst, and Kevan Lee 🧪 Growth Product Experiments from Lauryn Isford, Matt Woods, and Ben Williams 📈 Quantitative analyses from John Egan, Dan Wolchonok, and Yousuf Bhaijee 🧐 User research projects from Amber Rucker, Mike Fiorillo, Shelly Eisen-Livneh And sooooo many more. You can sign up here → reforge.com Preview some artifacts here → artifacts.reforge.com/artifa… These are NOT blank templates. Artifacts are the real work that contain the substance, nuance, insights, and realness. Here are a few things you can do with Artifacts: 🔎 Find artifacts relevant to what you are working on 💡 Access notes from the creator about the story, lessons, and insights ⭐️ Save artifacts and share them with colleagues 🔁 Remix artifacts to create your own version Artifacts is such a simple but powerful idea I can’t believe we didn’t think of it before. I’ve wasted so much time recreating something thousands of others have already done. Artifacts can help accelerate a lot of work (like open-source code) by enabling building on each other. Would love to know what you think.

Say hello to your new virtual assistant: Coda AI. ✨ Summarize meeting notes & transcripts in a snap. ✨ Quickly prep for customer calls. ✨ Whatever you dream up – it’s stackable with Coda’s other building blocks like tables, controls, text, & formulas. bit.ly/coda-ai

The @Coda_hq product team has always stood out as one of the most thoughtful, deliberate, first-principled product culture out there. In part 2 of my series on how the best product teams build product, I interviewed @lshackleton on their approach to product. A few highlights:

OpenAI's new ChatGPT writes a Seinfeld scene in which Jerry needs to learn the bubble sort algorithm:

Big news: we’re pleased to introduce the all-new Coda 3.0 — the doc that brings it all together. With a brand new editor, an open Packs platform, and hundreds of your most requested updates, Coda 3.0 is ready for your team. Read more about Coda 3.0: bit.ly/3dot0-t