Hi, I'm Sal.

For most of my career, I got really good at cooking. I learned the knives, the pans, the timing. I got comfortable in my kitchen.

Then AI came along. But AI didn't change my kitchen. What it did was give me ten more hands to prepare the same meal I had made dozens of times before.

That changed everything.

Not because I don't know how to cook. I do. But when you suddenly have twelve hands, the skill that matters most isn't the cooking anymore. It's the coordination. What should all those hands be doing, when, and why? How do they work together without knocking things over?

That's not a tool problem. That's an architecture problem.

Three things changed when AI arrived: Tools used to be passive. Now they participate. They don't change the space, but they change how we operate in it. And they increase creation at a rate that makes curation the new competitive advantage.

Those shifts are what most organizations are getting wrong right now. They're making the same meal, in the same kitchen, with ten more hands, and no change to the workflows, decision chains, or coordination patterns built for a world with two hands and passive tools. The result is speed without structure. Output without discernment. Scale without quality.

AI made creation easy. It made curation essential.

I design the operating systems that make curation possible at scale.

Roger Martin at the Rotman School built the framework. Integrative thinking: the discipline of holding two opposing models in tension long enough to produce a resolution that's superior to either one. Most organizations pick a side. Speed or governance. Automation or judgment. Scale or accountability. Integrative thinking refuses the choice. It designs for both.

What I do takes that further. In Hegel's language, the word is sublation. Not synthesis as compromise. Transformation. The output isn't a blend of two inputs. It's something structurally new that couldn't have existed without the tension between them.

That instinct shows up in everything I build. Four patterns repeat:

I think in constraints before capabilities. Every one of my 8 production workspaces starts with governance. Behavioral boundaries. Decision gates where AI stops and a human chooses. DO NOT lists. Constitutional documents that load at the start of every session. The constraint architecture always comes first. 19 reasoning patterns, the highest count in the heatmap, and most of them are judgment layers that determine what ships and what doesn't.

I design for the tension, not past it. The operating model I derived from my own work holds five roles in productive tension: governance constrains execution, routing eliminates ambiguity, synthesis transforms raw input into rules (not summaries), execution stays inside defined corridors, and augmentation puts human judgment precisely where irreversible decisions happen. Those roles don't resolve the tension between AI speed and human authority. They make both sides stronger because they exist together.

I build institutional memory by instinct. 18 memory patterns across 8 workspaces. The third-highest cognitive ability in the portfolio. 99-file knowledge bases, persistent context specifications, intelligence briefs with declared authority hierarchies, single-source-of-truth documents loaded as Required Reads before any AI action. This wasn't a deliberate strategy. It was an instinct that repeated until the data made it visible.

I operate at compressed timelines without compressing quality. A 12-pattern GTM diagnostic tool (framework, scorecards, article, outreach note) deployed as a single atomic commit. A 636-line product marketing site built in one AI session. A content business architecture with 5 engines, 12 API keys, and a 10-stage production pipeline designed as a single orchestrated system. The speed comes from fluency with the infrastructure, not from shortcuts.

I build systems that monitor themselves. 14 metacognition patterns. Self-assessment before publish, confidence scoring, pipeline health monitoring, evaluator disagreement detection. Systems that know what they don't know. Frankly, this is the capability most organizations skip entirely. And it's the one that determines whether AI scales safely or just scales.

I work across domains without losing architectural coherence. 8 completely independent workspaces: startup operations, company culture, content business, portfolio engineering, product marketing, skills assessment, go-to-market diagnostics, career management. Same five roles. Same governance-first instinct. Same memory architecture. The signature transferred without modification to the underlying design principles.

I know where I'm thin. Learning loops and social cognition are my two weakest cognitive layers. My systems are excellent at Day 1 judgment. They're less equipped for Day 90 improvement. The feedback mechanisms that would make the systems genuinely adaptive over time are designed but not yet instrumented. I say that because the point of mapping your work to a taxonomy isn't just to see what you've built. It's to see what you haven't.

I'm not an AI engineer. I don't write the models.

I'm not a traditional operator. I don't just use the tools.

I'm the person who designs the operating layer between AI capabilities and organizational outcomes. Governance, routing, memory, coordination, judgment architecture. That layer barely exists in most organizations. The people who can design it are rare. And the gap between organizations that have it and organizations that don't is widening every quarter.

The rest of this page is the evidence. The architecture I derived from 130 operational patterns across 8 independent workspaces. How those patterns map to a cognitive taxonomy designed for evaluating AI systems. And the working systems you can experience yourself.

Those names came from the work itself. No borrowed framework. No academic reference. Just the patterns that kept repeating across systems built to solve completely different problems.

In March 2026, Google DeepMind published a cognitive taxonomy for measuring progress toward AGI. Ten cognitive abilities derived from decades of psychology, neuroscience, and cognitive science, built to evaluate whether AI systems had achieved general capability.

When I mapped my five roles against their framework, the structures aligned across all ten dimensions.

I hadn't seen the paper. My framework used none of the same language. The convergence wasn't designed. It was discovered.

What that means practically: the cognitive abilities that determine whether an AI system is genuinely capable are the same ones that determine whether the operational layer a human builds around AI will hold. Reasoning, memory, planning, metacognition. These aren't just what makes AI smart. They're what makes AI systems reliable. A workflow without memory repeats the same mistakes every session. A system without metacognition can't catch its own errors. A pipeline without planning collapses under load.

Most organizations are building fast. Few are building sound. The difference is architecture. Specifically: governance before execution, memory at every layer, routing that eliminates ambiguity, synthesis that produces rules not summaries, and human judgment placed precisely where the stakes are highest.

That's what 130 patterns across 8 workspaces looks like when the same design mind applies the same principles to completely different problems.

Thinking out loud about integrative design, AI infrastructure, and how organizations change when they stop treating AI as a feature.

Longer pieces drawing on the case study and framework work.