Most people see a tangled business as a problem. I see the opportunity. The mess is where the mispriced fixes hide.

Everyone treats complexity like a disease. Simplify, streamline, clean it up. I treat it as the raw material. The messier the business, the more opportunity is sitting in it, unclaimed and usually mispriced.

Here's the logic. In a simple, legible system, the obvious fixes got made years ago. Anyone could see them. In a tangled one, huge wins are hiding in plain sight, not because they're hard, but because no one can hold the whole thing in their head at once to spot them. The complexity is exactly what's keeping the value hidden. And hidden value is cheap to claim.

Read the full essay and follow along with the transcript

I'm a giver by nature, and being a giver has a failure mode nobody warns you about. Coming out of the holidays, here is the boundary I've found that lets me stay generous without being taken advantage of.

The holidays just ended. Rosh Hashana and then Yom Kippur, the Jewish New Year and the Day of Atonement. For anyone who did not grow up with them, they are basically a forced annual audit of your life: a stretch of days where you stop, take stock of who you have been, and decide who you want to be. A week and change of that, plus time with friends, some travel, and a lot of sitting with my own head.

What kept surfacing was not work. It was people, and specifically one question I have wrestled with for years. I am a giver by nature. So how do I keep giving without getting taken advantage of?

Read the full essay and follow along with the transcript

Advisors hand you a deck. I hand you a working system. The difference is who owns the outcome.

The difference between a consultant and an operator isn't seniority, and it isn't how smart anyone is. It's what you hand over at the end. A consultant hands you a recommendation. I hand you the working thing.

A deck is a description of a solution. It is not the solution. The gap between "here's what you should do" and "here's the thing, running" is where most value quietly dies, and it's the client left to cross that gap, alone, after the expensive person has already gone home.

Read the full essay and follow along with the transcript

The order everyone gets wrong. Speed comes from trusting a trained gut, then using data to catch yourself.

Everyone says be data-driven like it ends the argument. I'm not, at least not in that order. I run gut first, data second, system third, and the order is the entire point. Get it out of sequence and you end up slow, or confidently wrong, or both.

Most people fail one of two ways. They never leave their gut and get burned when the pattern doesn't hold. Or they lead with data, no instinct, and drown in dashboards that never quite decide anything. The move isn't choosing one over the other. It's running all three, in the right order.

Read the full essay and follow along with the transcript

Healthcare, packaging, consumer, Web3. Different products, identical failure. Growth breaks in the same place every time.

People assume industries are different worlds, and from the outside they are. Healthcare has nothing in common with packaging. Packaging has nothing in common with a Web3 startup. Different products, different buyers, different rules. Then you get inside the growth problem and it's the same problem every time. Almost boringly so.

I've now built or fixed growth systems across healthcare, packaging, consumer, and Web3. Wildly different companies. Identical failure mode. Channels that don't talk to each other. No system holding it together. Growth that works only when a specific person is paying attention and quietly stalls when they aren't. Change the product and the story doesn't change at all.

Read the full essay and follow along with the transcript

The value isn't the big AI project. It's the backlog of small fixes that were never worth the effort until now.

Every business I walk into has a list of small broken things everyone knows about and nobody fixes. The report that gets rebuilt by hand every Monday. The two systems that don't talk, so someone retypes the same data between them. The follow-up that only happens when the right person remembers. None of it is a crisis. All of it leaks.

The reason it never got fixed isn't laziness. It's math. For years the calculus on a small fix was honest and brutal: this would take an engineer a week, the problem only costs us a few hours a month, so leave it. Multiply that across a company and you get a long tail of problems that were correctly ignored. Each one too small to justify the fix. Together, a slow bleed nobody's accountable for.

Read the full essay and follow along with the transcript

If it falls apart the moment you stop watching, you didn't build a system. You were the system.

The best thing I can hear about work I've done is that nobody remembers I did it. The fix is just how things run now. No hero, no dependency, no name attached. That's the whole goal, and most people build the exact opposite without noticing.

They build things that work while they're watching. Then they call it a system, when really the system was them. Take the person out and the thing sags back to where it started.

Read the full essay and follow along with the transcript

At the frontier you can't ask how it's done, because it hasn't been. So you replace precedent with something else.

Early in my career, the work had answers you could find. Google Ads, marketing platforms, the standard growth motions. Someone had always done it before me. There were docs, case studies, a person one Slack message away who'd solved the exact thing. The skill back then was mostly learning the existing playbook and running it well.

That skill has a shelf life, and lately it runs out fast.

Read the full essay and follow along with the transcript

Every owner can name the thing holding the business back. They're almost always naming the wrong thing.

Every owner I talk to can tell you exactly what's holding the business back. Marketing isn't generating enough. Sales isn't closing. The market softened. There's always an answer, it's always confident, and it's almost always wrong.

Not wrong because the owner is careless. Wrong because the explanation everyone settled on is the one you can see. The real constraint is usually somewhere you can't, and that's precisely why it's still there.

Read the full essay and follow along with the transcript

The LLM is one component. The memory system, capture pipeline, retrieval fusion, and decision architecture around it is where the real leverage lives.

I wrote a couple months ago about memory being infrastructure and building your own tools. Since then I've been deep in the implementation. This post is about the patterns that actually worked.

The short version: the LLM is one component. The architecture around it is where the real leverage lives. And "architecture" isn't a hand-wave. It's a set of specific, solvable engineering problems that most people skip because they're busy optimizing prompts.

Read the full essay and follow along with the transcript