The Everything Engineer is a substack about how to thrive in an AI-enabled economy. Subscribe for on-the-ground reports, tips, and philosophical musing on the future of startups & business.
The Decline of Software Engineering...
Software engineering is in decline.
Software development job postings have decreased by 35% since February 2020, even as overall job postings across all industries have increased by 10%. Developer employment keeps dropping: -4.6% in January 2022, -3.5% in May 2022, and -3.4% in January 2023.
To be fair, much of these stats represent a correction of the pandemic-fueled hiring binges. But they also represent expectations on how AI is going to transform the discipline. GitHub Copilot enables developers to complete coding tasks up to 55.8% faster. The internet abounds with stories on "vibe coding" software from scratch, though not so many stories of those folks maintaining or scaling that software 😉.
I can attest personally to the power of AI-assisted coding: in building innori, my current obsession, I'm a better, faster, more capable coder than I've ever been, despite having written minimal production code over the 5 preceding years that I worked "in management".
So, software engineers are screwed. Right?
...is the Rise of Everything Engineering
Yes: Software engineering jobs are slowly going extinct. Thank god.
Today’s engineering roles are, frankly, prisons of specialization — turning talented builders and problem solvers into "solutions providers". The trajectory of engineers' careers is increasingly defined not by what they build, by their perceived indispensability to what needs to be maintained—because they hold secret maps to ancient systems and obsolete architectures everybody else is too afraid to touch.
Take, for example, pretty much any "principal engineers" you may have known and ask yourself: did they reach the peak of the engineering job title pyramid by virtue of their expected future contributions? Or was it because they were the only person left who understood some legacy system nobody wanted to rebuild?
Well, now a half-decent LLM can tell you how that code base works! That feels a lot like liberation to me.
Engineering was never about coding. Engineering is about building things. What is exciting about AI-assisted coding is not that it replaces engineering jobs, but that it completes the evolution of turning all jobs into engineering jobs. Engineers now have agency to build cross-functionally at scale.
Ever wanted to automate an operations process, from root to fruit? Wanted to raise a round without doing a year building connections in VC? Ever wanted to build a truly excited product? It’s the time to shoot your shot.
Everything is Systems
Here’s a personal example. I co-founded and built the tech at RubiconMD. Every time we release an update to our iOS app, I had a little fun writing the release notes. One went:
Somebody told us our app "looked pretty good... for healthcare" so we cried all night and then rebuilt it with this brand new interface.
Our marketing team was not happy: The message wasn't in the brand "voice"! It was missing punctuation! There should be protocols for writing these that should be followed!
The head of marketing insisted she "own" the writing of release notes henceforth. It wasn’t high value work, so I didn't fuss. Except: marketing couldn't write the release notes without knowing what was in the release, so my job now became:
write what is in the release
email it to the junior marketing team member
review her draft to see if it stated correctly what’s in the release
prod her to prod the head of marketing to review and approve
copy and paste the approved copy into iTunes Connect
Eventually, I just started releasing each update as "bug fixes".
The point is that this was ridiculous, yes, but the marketing team wasn't wrong. There's value in having quality standards in even the smallest communications from your brand. With Rubicon, however, the friction of communication for getting this got so high I just stopped doing it.
With innori, I’ve automated it all. I christened an LLM that runs on my laptop as head of marketing, and automated the whole app release note workflow as part of a larger “knowledge base” project. More on that in a future post.
I just generate, release and forget. Whole thing took 30 minutes to build.
Welcome to the Discretionary Age
This is just one example from the beginning of what I've started calling the Discretionary Age. Workers’ information or knowledge now doesn’t matter as much as their gross agency: or their ability to realize larger and more impactful changes across the complex web of systems and processes that make up their company or industry.
It’s a philosophy I've been applying to every step as I build innori, my latest startup. In the Everything Engineer, you'll get a co-pilot’s view of the journey, with the ability to explore with me some of the bigger ideas that have emerged along the way, such as:
Measuring (and maximizing) "gross agency" in your team
The impending demise of SaaS
Defining & automating an inverted hierarchy of knowledge
Embracing non-reproducible workflows
...and more!
This is a substack for discussing the craft of engineering everything, written on the dawn of a time when possibly everything you need to do to run a company can--and will--be engineered.
Sharpen your pencils, it's time to build.