Hello World

I still remember the day the internet arrived at my desk.

I was working my first industry job as a designer at a niche financial marketing company. At the time, my days were spent designing letterheads for small banks and doing imposition work for their white papers. But as the web became viable for business, our clients began asking us to design their websites. We didn’t have that capability within the team.

One afternoon, the owner of the company walked into the design bullpen and asked a simple question: “Who wants to learn how to make websites?” My hand shot up, and he handed me a $100 gift card to Barnes & Noble. I bought every book they had on HTML and Perl.

That moment profoundly shaped my career. It taught me that when the medium changes, you don’t panic, you adapt and embrace the transition. That pivot eventually led me to a career first as a full-stack Software Engineer, and later to synthesizing my design roots and coding skills as a UX Engineer at Google. Most importantly, it reinforced a core philosophy I learned at Cal Poly SLO: Learn by doing.

I share this story because today, in 2026, we are facing another “Barnes & Noble gift card” moment. If you’ve been paying attention to the intersection of AI, software engineering, and UX design, you can feel the ground shifting. AI is no longer just an autocomplete tool; it is becoming a general substitute for cognitive work.

The Collapse of the Traditional Lifecycle

The traditional, linear software development lifecycle (Research → Design → Code → Test) has completely collapsed. We are moving away from a code-centric world toward a spec-centric design and development process.

For decades, UX professionals and engineers have taken deep pride in our craft. We found value in manually pushing pixels, refining typography, and writing boilerplate syntax. But the economic and technical reality of 2026 is blunt: it no longer makes sense to do that work by hand. The prompt is now the build. With the rise of advanced agents and multimodal models, the friction of production has dropped to near zero, allowing us to generate functional, high-fidelity features, if not whole products, in minutes.

It is easy to feel anxiety right now. Many practitioners fear that the rise of agentic AI means the death of software and UX craftsmanship. I am starting this blog to offer a perspective rooted in a different reality: we are not witnessing the death of our craft, but its remarkable evolution. We are transitioning from a craft-based production model to a supervisory orchestration model. The new skill is not writing the code, it is context engineering.

Entering the Middle Loop

If AI handles the technical execution of drawing a screen or scaffolding a database, where does the engineering rigor go? It moves upstream.

We are entering the Middle Loop, a supervisory orchestration layer that sits between raw code generation and outer-loop deployment. In this space, human intent is becoming the new source of truth. AI models can predict the next pixel or the next line of syntax with astonishing accuracy, but they cannot predict human purpose. They do not possess intrinsic empathy, taste, or strategic judgment.

To thrive in the Middle Loop, UX and engineering leaders have the opportunity to evolve into Architects of Intent. We can move beyond treating design as a static visual export, and elevate it to programmable infrastructure.

Designing the Agentic Canvas

Over the next six or so weeks, I will be publishing a foundational series exploring how we can adapt our practices, workflows, and mindsets to harness agentic velocity and build software reliably at machine speed.

We will explore this transition through six core pillars:

Part 1: The Evolution of Craft
We will explore how the evolution from “Operator” to Intent Architect allows us to delegate high-level goals while maintaining architectural coherence across a fleet of digital workers.

Part 2: Designing the Catalog
AI models don’t have eyes, instead they see grids of floating-point values. We will explore how to solve AI “design blindness” by shifting from static Figma exports to programmable JSON component catalogs and structured DESIGN.md files.

Part 3: The Kaizen of Context
When generation is instant, verification becomes the bottleneck. We will look at why prototyping in production invites us to adopt strict Red/Green Test-Driven Design (TDD) as our ultimate form of prompt engineering and automated validation.

Part 4: The Approval Interface
While perhaps difficult to imagine for those that have yet to experience it firsthand, unmanaged AI fleets create compounding cognitive debt. We will discuss how to design human-in-the-loop orchestration interfaces and UX pull requests to safely review and approve agent actions.

Part 5: Onboarding the Machine
In the agentic era, stale documentation is a major liability. We will explore the daily reality of maintaining the “agent subconscious” and preventing context drift by rigorously curating AGENTS.md files (e.g. CLAUDE.md, GEMINI.md, etc).

Part 6: The Mythical Agent Month
Finally, we will look at the challenges of scaling AI fleets. Without human judgment, agents can easily generate technical debt and “slop forks” at unprecedented speeds, proving that an Intent Architect is more vital than ever.

Code as Craft, Reimagined

The silos separating Product Management, Engineering, and UX are blurring. But your expertise is not obsolete, it has simply shifted up the abstraction stack. The future belongs to those who adopt the learner’s mindset, who can articulate exactly what they mean in machine-readable formats, and who aren’t afraid to get their hands dirty engineering human-AI collaboration.

Hello world. Let’s learn by doing.