Legacy Rehabilitation
A while back, I wrote about why I moved from the East Coast to San Francisco after more than 25 years of professional experience.
At the time, the explanation was partly geographic. I wanted to be closer to the people building the future. I wanted to be where the density of ambition, capital, technical talent, and company formation was highest.
That was true. But I now think there was a deeper reason. I was trying to rehabilitate myself from legacy thinking.
Not legacy in the sense of age. I do not think experience is the problem. In many cases, experience is the only way to understand where the real problems are.
The problem is different.
After decades in and around financial services, enterprise technology, and regulated industries, you absorb certain assumptions. You start to accept slowness as inevitable. You start to accept complexity as natural. You start to believe that moving money, underwriting risk, originating loans, buying property, servicing assets, or offering financial products to customers must always be slow, expensive, and operationally heavy.
Eventually, the system teaches you to confuse waste with risk. That is the part I wanted to unlearn.
The Disappointment of Legacy Technology
I have spent much of my career in financial services and adjacent industries. I have seen wave after wave of technology arrive with the promise of transformation.
Core systems. Workflow systems. Data warehouses. SaaS platforms. APIs. Cloud migration. Digital transformation. Automation programs. Now AI.
Each wave improved something.
But the fundamental customer experience has not changed nearly enough. To this day, the speed of offering many financial products to customers remains too slow. The cost remains too high. The process remains too fragmented. The amount of human coordination required remains absurd.
Some of that cost is legitimate. Risk has a price. Compliance has a price. Capital has a price. Companies need healthy margins to survive.
But a large part of the cost is not the price of risk. It is the price of waste.
It is the price of duplicated work, broken data, manual reconciliation, organizational handoffs, institutional defensiveness, and technology stacks built around the assumption that the current process is permanent.
That waste creates margin for some participants. It creates friction for everyone else.
My view is that AI will eliminate a meaningful amount of that waste very quickly. Not because AI is magic, and not because models are suddenly smarter than every expert in the industry.
AI matters because it changes the cost of coordination, investigation, analysis, and execution. It exposes the inefficiency that legacy systems normalized.
Legacy Rehabilitation
So my move to San Francisco was not just a relocation. It was part of a broader process I now think of as legacy rehabilitation.
The first step was changing my network. That sounds simple. It was not simple for me.
I am an engineer by training and temperament. I am not naturally open or socially fluid. I do not enjoy networking in the conventional sense. For most of my career, I operated inside communities where I often felt misaligned.
Not better than them. Not worse than them.
Just misaligned.
I would look at problems one way, and the surrounding system would look at them another way. I would see structural waste, while others saw process. I would see incentive problems, while others saw vendor selection. I would see the need for new infrastructure, while others wanted another layer on top of the old one.
For a long time, I interpreted that misalignment as a personal oddity. Maybe I just thought differently. Again, not in a good way or a bad way. Just differently.
But AI gave me a new way to test that assumption.
Using AI to Find Intellectual Neighbors
One unexpected use of AI has been personal. I started using it to map intellectual alignment. Instead of using social media only as a place to broadcast or consume, I began treating it as a data source. I looked at the way people communicated publicly: their posts, essays, arguments, reactions, and patterns of thought.
Then I used AI to help me understand them. Not in a superficial way. Not “what topics does this person talk about?” But:
How do they reason?
What problems do they return to?
What tradeoffs do they notice?
What do they consider obvious that others miss?
Are they focused on systems or symptoms?
Do they prefer status or substance?
Do they understand technology as tooling, or as a change in operating model?
Are they trying to preserve the old world, or build the next one?
Over time, this became a kind of personal RAG system. I could compare my own writing and thinking against the writing and thinking of others. I could identify who seemed aligned, who seemed adjacent, and who seemed directionally different.
This changed something for me. I began to realize there are many people who think in a way that feels familiar to me. They may work in different fields. They may use different language. They may come from venture, engineering, crypto, AI, financial infrastructure, real estate, or enterprise software.
But the thought pattern is similar.
They are interested in hard problems. They distrust surface-level transformation. They care about architecture, incentives, systems, and execution. They understand that technology does not create value just by existing; it creates value when it changes how work actually happens.
For someone who spent years feeling intellectually misaligned, this was useful.
AI became a way to find friends. Not necessarily friends in the traditional social sense. More like intellectual neighbors. People whose thinking provides evidence that you are not alone in how you see the world.
Why This Matters for Building
This is not just a personal story. It matters because the network you are in shapes the company you can build.
Legacy environments have gravity. They pull you toward legacy conclusions. They reward incrementalism. They overvalue consensus. They punish uncomfortable clarity. They teach you to describe problems in language that makes the existing system feel safer.
If you spend enough time there, even your ambition gets normalized downward.
That is dangerous right now because AI is not just another enterprise tool. It is not simply a productivity feature. It is a new execution model.
When one person can direct team-scale output, the bottleneck shifts from labor to judgment. When agents can perform investigation, reconciliation, generation, and execution, the shape of the organization changes. When software can operate as work rather than just support work, the economics of many legacy industries change.
But seeing that clearly requires being around people who are not emotionally committed to preserving the current structure.
That is why network matters. The right network does not just provide introductions. It changes what feels possible.
From Social Graph to Thought Graph
I think this is one of the more interesting personal applications of AI.
Historically, your network was constrained by geography, job history, education, and social access. You met people through companies, conferences, investors, friends, and chance. That still matters.
But AI allows something different: a thought graph.
A thought graph is not based on who knows whom. It is based on how people think. It can be built from writing, public conversations, technical work, investment theses, open-source contributions, and repeated patterns of reasoning.
For me, that became a way to answer a practical question: Who is actually aligned with the way I think about the world?
Not who has the right title. Not who has the most followers. Not who is fashionable in the current cycle. Not who performs intelligence well in public.
But who is reasoning from similar first principles?
That is a very different way to build a network. And for founders, it may become increasingly important.
The Rehabilitation Is Ongoing
I do not think legacy rehabilitation is something you complete. It is ongoing. Every time I look at a financial process that takes weeks and ask why it cannot take minutes, that is part of it. Every time I see a company treat AI as a bolt-on tool rather than a redesign of the operating model, that is part of it. Every time I find myself accepting some industry constraint as permanent, I try to ask whether it is actually a constraint or just institutional memory.
That is the real work.
To build something new, you have to distinguish between:
risk and waste
complexity and confusion
regulation and institutional habit
margin earned through value and margin extracted through friction
transformation and retooling
network alignment and social proximity
This is why I moved. Not just to be in San Francisco. To place myself in a network that makes legacy assumptions easier to see and harder to keep.
Closing Thought
AI is often discussed as a way to make people more productive. That framing is too narrow.
AI also helps people find alignment. It helps expose the structure of thought. It helps individuals rebuild their networks around ideas rather than proximity. For me, that has been part of the journey.
I am trying to build in industries where the legacy system has confused waste with risk for too long. To do that, I needed to change not only what I was building, but who I was thinking with.
That is what I mean by legacy rehabilitation. It is the process of unlearning the assumptions of systems that no longer deserve to define the future.
And, surprisingly, one of the most useful tools for that has been AI.
