Every Conversation Starts Over
Why memory may matter more than intelligence in the way we work with AI.
The smartest employee you've ever worked with
Imagine hiring the sharpest person you've ever worked with. They read everything you hand them. They ask great questions. They make connections you didn't see, and they solve problems in minutes that normally take days.
Then every day at 5:00 p.m., they forget everything. Monday morning they're brilliant again — but they don't remember your biggest customer, the strategy meeting from last Thursday, or the product idea you spent three hours working through together. They don't even remember you. Every morning starts with introductions.
Ridiculous, right? That's how most of us use AI.
The bottleneck wasn't the technology
When I first started leaning on AI every day, I assumed the bottleneck would be the technology. It wasn't. The bottleneck was me — or more precisely, how I managed information.
I'd have an incredible conversation. Work through a customer problem, outline a go-to-market plan, brainstorm a product, research a market. I'd leave thinking, that was useful. Then I'd come back three days later and start over. I wasn't continuing the conversation. I was rebuilding it. Again, and again, and again.
A tax paid in context, not tokens
It finally hit me that I wasn't paying a tax in tokens. I was paying a tax in context. And context is expensive — not because it costs money, but because it costs time.
Every conversation started the same way. Here's the company. Here's the customer. Here's what we've already tried. Here's what we're building. Here's where we got stuck. By the time you've rebuilt all of that, you've spent the first twenty minutes catching up instead of moving forward. We've done it so many times we barely notice anymore.
Businesses already have this exact problem
The funny thing is, companies have lived with this for decades. Someone leaves. Six months later, another person finds a document and asks, "Why did we decide to do it this way?" There was usually a good reason. The reason just disappeared.
It lived in a meeting, a Slack thread, a hallway conversation, someone's notebook, someone's head. The decision survived. The thinking behind it didn't. That's the same failure, one level up — and it's why intelligence is one of the six places growth gets constrained.
I was doing it to myself
Then I noticed I was creating the exact same problem for myself, just faster. My ideas lived everywhere. Some in AI chats. Some in OneNote. Some handwritten. Some buried in email. Some existed only because I happened to remember they existed. That's not a knowledge system. That's a scavenger hunt.
The more I relied on AI, the more obvious it got. AI wasn't the one forgetting. I was — or more accurately, I wasn't giving it a way to remember. I'm not doing it to organize things for their own sake. Honestly, I'm not that person (just ask my wife). I'm doing it to make Future Josh's life easier. He's the one who has to remember why Present Josh thought something mattered, and historically I've made his job harder than it needed to be. Josh's biggest problem is usually Josh.
Intelligence isn't the hard part. Memory is.
We tend to assume the hard part is asking a better question. I've come to believe the harder part is giving the system enough history to produce a better answer. Those are completely different problems. One is intelligence. The other is memory.
That distinction changed how I work. Instead of treating every conversation as a one-off event, I started treating it as another chapter in an ongoing story. Meeting notes, customer conversations, stray ideas, research — everything eventually finds its way back to one place it can accumulate.
There's a second reason this matters more now than it used to. Twenty years ago, most of us weren't generating dozens of pages of thinking in a week. Now we are. That's a gift — until you can't find any of it. The problem isn't creating knowledge anymore. It's keeping it connected. So I stopped thinking about notes, documents, chats, and folders, and started thinking about memory: not human memory, but organizational and personal memory — the kind that survives longer than the conversation that created it.
Where the leverage actually is
I don't think the future belongs to whoever has access to the smartest AI. We're getting very close to a world where that level of capability is available to everyone. The advantage will belong to the people — and the companies — who stop starting over every morning.
That's the whole idea behind The Intelligence Estate: one owned place where the important parts of the work accumulate, connected to what they relate to and to why decisions were made. The habit that keeps it alive is a weekly knowledge sweep — a standing block to pull the week's scattered thinking back into that one place before it's lost. I'll write that one up next.
The smartest employee you'll ever work with is already here. Now we just have to give it a memory worth building on.
If this idea resonates, read The Intelligence Estate next.