The Shift.

I recently put out a video sharing my perspective on what's happening in the software engineering industry right now. Nineteen minutes...straight to camera...no slides. Just the signal I've been tracking and what I think it means.

But fifteen minutes can only carry so much. And the more conversations I've had since...with engineers on my team, with friends at other companies, with folks deep in the AI space...the more I realized the deeper cuts needed a home. The evidence...the frameworks...the thinking underneath the thinking.

So this is that.

If you watched the video...this goes further. If you didn't...this stands on its own. Either way...I'm going to lay out what I'm seeing, why I believe what I believe, and what I think it means for our craft.

There is a lot of noise out there right now. A lot of AI-generated slop. A lot of hype cycles masquerading as insight. And no shortage of people with strong opinions who aren't actually building anything.

I was deliberate about not rushing to put something out. The last thing anyone needs is more noise.

But I pay close attention to signal over noise. I watch patterns. And over the last several months...too many things have converged for me to stay quiet about it. Not one signal. Not two. A cascade. Across geopolitics, across industry, across what's happening inside the companies I talk to, and across what I'm experiencing firsthand by building.

So here it is. Straight up.

Now let me tell you why I believe this.

If you only look at this through the lens of "AI is writing code now"...you miss the bigger picture. There are three layers all converging at the same time. And when you stack them...you start seeing something much larger than a technology trend.

The geopolitical layer is real. This isn't a tech conference topic anymore. In February 2026, the India AI Summit brought every major AI CEO into one room...Altman, Amodei, Pichai, Hassabis...with heads of state from over a hundred countries. The language being used publicly, on stage, to world leaders, was "a couple of years" from major capability jumps. Whether you believe the exact timelines or not...what matters is that the people building these models are making infrastructure bets, hiring bets, and product bets based on those timelines being real.

There's an arms race happening right now to own the capability, the technology, and the gateway for frontier AI models. It's playing out like when Google became the de facto entry point for the internet through search...or how Google and Apple became the two dominant forces in the mobile ecosystem through Android and iOS. Same kind of positioning. Except the stakes are much higher, the impact much broader, and the space much more ferocious.

And then layer on what tariffs are doing...shaking entire companies overnight, reshaping trade and supply chains, questioning fundamental business models. That's macro forces at work. AI is sitting right in the middle of that same kind of tectonic shift...except this one touches every industry, every function, every role.

For Canada specifically...the situation is acute. Our productivity growth diverged from the US over a decade ago. We produce world-class talent but we struggle to commercialize it. AI could widen that gap. Or it could be the thing that closes it. That's a choice that companies and leaders and engineers are making right now.

Where economic value sits inside a company...that's moving. Look at what Satya Nadella has been saying about SaaS businesses...essentially that they become CRUD interfaces. Think about that. In an agent-first world, where as a user you don't want to log into a dozen different applications...why would you? Just have your agents do things on your behalf. Compare your insurance quotes. Check your accounts. Book the flight. Schedule the follow-up. One conversation instead of five apps.

We've seen this platform shift pattern before. Websites ate share from desktop applications when the internet took off. Mobile apps ate share from websites. And now...agents are starting to eat share from mobile apps. Each platform shift moved faster than the one before. And this one...this one is moving the fastest.

The competitive moats built on access to capital and large development teams...those moats are being taken down. Where economic value gets created and captured is being fundamentally reset.

And how work actually gets done inside organizations...is about to go through a profound shift. Not just in engineering...everywhere. But engineering is the canary in the coal mine because it's where the models are most capable right now.

When you intersect these three layers...geopolitical shifts underway, economic value and competitive moats moving, and how work gets done changing...the craft of engineering as we knew it for the last few decades is turning into something completely new.

I'm a pattern guy. I look at how things have moved historically. And when I zoom out...every few decades, the fundamental layer of how value gets created changes. And every time it does, the organizations that see it early build the next era. The ones that don't...get absorbed by it.

Remember when companies added a "mobile" tab to their desktop website and called it a mobile strategy? How'd that work out?

The ones who got it right...who understood that mobile wasn't a channel but a paradigm...they built the next generation of dominant companies. The ones who bolted on a responsive design and moved on...they're still playing catch-up. Or they're gone.

Now...here we are again. Agent era. Human goes to augmented.

And I'm watching...almost in real time...the same mistake being made. Companies taking their existing processes, their existing tools, their existing way of working...and bolting AI onto it. A chatbot here...a copilot there...an automation tool somewhere else. And calling it transformation.

But there's a deeper current underneath this one that changed how I think about everything. For decades, companies created value primarily through labor efficiency...hire the right people, organize them well, give them tools to work faster. That model is being rewritten. Value creation is migrating from labor efficiency to cognitive leverage. From "how many smart people can we hire" to "how effectively can we amplify the judgment and expertise of the people we already have."

An organization with 500 engineers equipped with the right AI infrastructure can outpace an organization with 5,000 engineers without it. That's not a productivity tool. That's a different game entirely.

And unlike every previous era shift...AI improves itself. The tool driving the change is also accelerating the change. That's never happened before.

Not from startups in a garage. From companies you know. Companies serving hundreds of millions of users. And the pattern across all of them is identical.

Everything I just described...the evidence, the convergence...that's what's already happened. What should really get your attention is the speed at which it's accelerating.

While the India AI Summit was happening...Anthropic launched Claude Code Security. AI that scans entire codebases for vulnerabilities...not pattern matching like traditional tools...actually reasoning through code like a human security researcher, tracing data flows, understanding how components interact. Found over 500 vulnerabilities in open-source projects that had gone undetected for decades. Companies like Veracode, Snyk, and Checkmarx got put on notice. Markets reacted in real time...cybersecurity ETFs dropped to their lowest level since November 2023...billions in market cap moved from one product announcement in one day.

Think about the trajectory underneath this. Models started as text in, text out. Then came tool use...giving them hands to call APIs and interact with systems. Then computer use...operating a screen, clicking buttons, navigating interfaces. Then coding started getting solved. And once coding is solved...adjacent spaces become quick follows. Security was literally this week. Testing, infrastructure, ops, design...all on the table.

Every era shift I mentioned...each one reshaped how industries work. And this one is compressing the timeline in a way we've never seen before.

There are two approaches to what's happening right now.

Most organizations...and I mean the vast majority...are running Approach 1. And calling it transformation.

Your org chart didn't change. Your handoff points didn't change. Your decision latency didn't change. Just the typing got faster.

And the biggest risk isn't choosing the wrong approach. It's confusing Approach 1 for Approach 2. It's thinking you're transforming when you're actually just optimizing the old thing. Because while you're doing that...the Spotifys and OpenAIs of the world are building an entirely different machine.

I know what a lot of engineers are feeling right now. I've felt it myself. The question in your head: "What happens to me?"

I'm not going to give you some corporate answer like "don't worry, humans will always be needed." That's vague. And vague doesn't help when you're staring at a paradigm shift.

So let me be specific.

For 40 years, the primary activity of a software engineer was writing code. The core question was "how do I implement this?" That's inverting. The primary activity of an engineer in this new model is designing environments. The core question isn't "how do I implement this?"...it's "what capability is missing?"

You're not writing the code. You're designing the system that agents write code within. Defining the constraints...setting the guardrails...building the context...curating the knowledge. Evaluating the output. Making the judgment calls that agents can't make.

The career identity thing. I think the industry isn't talking about this enough. If you've spent 10 years becoming an exceptional code writer...and someone tells you "the craft moved"...that's disorienting. It doesn't matter that the new thing might be "better" or "more valuable." It's still a loss of something you built your identity around. And I think we have to acknowledge that. Not dismiss it. Not hand-wave it with "upskill and you'll be fine." Actually sit with it and say...yeah, this is a real transition. It's going to be uncomfortable.

But the engineers who figure out how to operate at this new level...they will be the most valuable engineers in the market. Not as a consolation prize. As a genuine upgrade.

ThoughtWorks ran an engineering retreat recently with senior engineers and architects deep in agent-assisted development. They named the actual skills.

These are hard skills. Genuinely hard. And they need to be explicit competencies...not assumed. Not something you "pick up." Something you deliberately develop.

It's not just the engineer's role that inverts...it's the entire delivery lifecycle.

Traditional SDLC: Requirements → Design → Build → Test → Deploy → Maintain. Linear. Sequential. Lots of handoffs.

The AI-native version is fundamentally different.

And here's what most people miss...the bottleneck in the old model was never the coding. It never was. The bottleneck was the decision latency, the handoffs between siloed functions, the meetings about meetings about meetings. Coding was maybe 10-15% of the time. Everything else was organizational friction.

So when people say "AI just makes the typing faster"...they're missing the entire point. An AI-native process doesn't just speed up the typing. It eliminates the handoffs. It collapses the coordination overhead. It moves decision-making to where the information is.

On governance...for anyone in a regulated industry thinking "that sounds great but we have compliance requirements"...current model is governance as ceremony. Point-in-time reviews...manual approvals...sign-off gates. New model is governance as code. Continuous verification...policy as code...every agent action logged, auditable, traceable. This is better governance. Not less governance. More auditable...more consistent...less dependent on a human being awake and attentive at exactly the right moment.

There's a maturity curve here...from Level 0 (using Copilot to help write code faster) all the way to Level 4 (agents do everything, humans define acceptance criteria). Most organizations are at Level 0 or 1. The evidence companies are at Level 3 and 4. And the gap between those levels isn't just speed...it's a fundamentally different way of operating.

I've written the full breakdown of the AI-Native SDLC...each maturity level, what changes at each stage, how governance works, and how to think about the transition...in a dedicated deep-dive: The AI-Native SDLC.

Let me paint a picture of what most large organizations look like right now with AI.

Team A built a chatbot. Team B integrated Copilot. Team C has automation scripts using GPT...Team D is running an experiment with agents...Team E signed a vendor contract for an AI analytics tool...Team F heard about what Team D is doing and is building their own version.

Everybody's doing something. Nobody's connected to anybody else. Each team chose their own model, their own framework, their own approach. No shared learning...no shared infrastructure...no shared governance...no visibility across the organization.

I call this Island Syndrome. And it's everywhere.

The problem isn't that individual teams are doing bad work. Some of it is impressive. The problem is none of it compounds. Team A's learnings don't benefit Team B. The chatbot Team A built...Team C is building from scratch because they don't even know it already exists. Everyone rebuilds the plumbing. Authentication, logging, rate limiting, model routing, context management...the same plumbing rebuilt 15 times across 15 teams.

It would be easy to spend this entire publication pointing at Spotify and OpenAI and "the future of everything." That's what most leaders do. They point at the shiny thing out there and conveniently skip over what's happening inside their own house.

I'm not going to do that.

I lead 650-plus engineers at a 150-year-old financial institution. When I joined, I didn't bring in consultants for a big assessment. I went to the engineers. The people doing the work every day. And I asked them honestly...what's slowing you down?

That last one...think about it. 120 days. And the vast majority of that time was waiting for approvals, environments, other teams, release windows, meetings to get scheduled so decisions could get made. That's not a technology problem. That's an everything problem.

I'm not sharing this to point fingers. The people are talented...the intentions are good...the bottleneck isn't talent. It's everything around the talent. It's the process that accumulated over decades. The organizational structure that creates handoffs. The incentive models that reward risk avoidance over velocity. The legacy of every "transformation" that came before...that added frameworks and governance and ceremonies without ever taking anything away.

And look...if you're at a different company reading this and thinking "well that's not us"...with respect...it probably is. Maybe the specific numbers are different. But the pattern is near-universal across large enterprises.

That's not a productivity improvement...that's a different organization. Hundreds of engineers worth of output added without hiring a single person. Just by removing friction.

And those are 9,750 hours per week of engineers who could be learning to work in the new paradigm. Building AI-native systems. Developing the skills to operate at the next level. But right now...they can't. Because they're stuck in the 33 steps, the meetings, the firefighting, the waiting.

I go deeper on the systemic analysis...what's actually causing these numbers and what the fix looks like...in the SDLC Redesign piece.

Everything I've laid out comes down to two tracks running in parallel.

Most organizations try to pick one. Wrong question.

Track 2 without Track 1...is building on a broken foundation. You can't do agent-primary development when your engineers can't even deploy without 33 steps. The agents will just be faster at hitting the same bottlenecks.

Track 1 without Track 2...is perfecting a process that's about to be obsolete. You'll have the most efficient traditional SDLC in the industry...just in time for it to not matter.

You can't go fast on a broken road. But you also can't pave a road to nowhere.

And the operating system layer I described...that's what connects them. Track 1 improvements feed into the orchestration layer. Track 2 capabilities build on top of it. Everything compounds.

One more thing I want to name. Because I think it reveals something about the depth of this shift.

Words like "agents" and "agentic" weren't even commonplace a year ago. Now they're everywhere. The lexicon itself is evolving because the old words don't describe what's happening anymore. When you have to invent new vocabulary to talk about what you're doing...that tells you the change isn't incremental. It's categorical.

And nobody has this fully figured out. If you follow the folks building at these companies...what you see is everyone figuring it out in real time. No one knows what this new thing looks like when it matures. The playbook is being written as we go.

That's exactly why the people who will be best prepared...it's not the ones with the most years of experience or the deepest knowledge of a specific framework. It's the curious ones. The ones with a growth mindset and a bias to learn. The ones who are adaptable. The ones who move.

Here's something I've been thinking about a lot. We're used to organizing our workflows, our systems, our documentation in a way that makes it easier for other humans to pick up the work...to debug, to understand how a system works. That's how we've always done it. But in a world where AI writes the majority of the code...that model breaks. You need codebases, documentation, processes that make it easier for AI to work. Not humans. And that's a vastly different approach than most current development setups.

I want to close with why I actually care about this.

I've been in this industry for over 20 years. I've been through a few of these shifts. I was building digital experiences when mobile was the "new thing." I've watched companies see it early and build the next era. And I've watched companies hesitate...wait for the playbook...wait for certainty...and never recover.

The shift we're in right now is bigger. And it's moving faster. By a lot.

Every time there's been a major shift like this...the craft goes up. We start solving higher-order problems. When we got compilers, we didn't lose programmers...we stopped writing machine code and started building applications. Cloud didn't kill ops engineers...it moved them from racking servers to designing systems. The abstraction goes up. The impact goes up. The value goes up.

And this doesn't mean less software gets built...the opposite. Expect exponential growth in the amount of code being written, products created, speed at which things ship. We're about to go into overdrive. More software will be built in the next few years than in the previous few decades combined.

For an organization like Sun Life...a 150-year-old financial institution...this is actually a massive opportunity. The playing field is leveling. The advantages pure tech companies had...speed, developer density, engineering culture...those are compressing. If we move now, if we lean into this, if we fundamentally rethink how we build and ship...we can compete in ways that weren't possible even a year ago.

To keep it real...I don't have it all figured out. Nobody does. Anyone who says they have the complete playbook for what engineering looks like in three years is lying to you. But I know this much...the worst thing you can do right now is nothing.

So stay deeply curious. Dig in...start experimenting...keep sharing what you learn. And I'll keep sharing what I'm seeing.