The Markdown Summary of Your Life

A transcript remembers everything. A summary remembers what matters.

Not long ago, I was in the middle of an AI workshop explaining why context has to be maintained deliberately. Otherwise, conversations drift. As language models process more prompts, new ideas, and changing priorities, the original purpose slowly becomes harder to see. It’s not that the model has forgotten everything. It simply loses sight of what matters most. That’s why I teach people to create markdown summaries. They preserve the objective, the important context, the key decisions, and the direction of the conversation so you can pick up tomorrow without starting over.

During that workshop, someone raised their hand and asked a question I wasn’t expecting.

“Do you do something like that for your own life?”

I smiled and answered without hesitation. I talked about journaling, reflecting, and taking time to think. It sounded thoughtful. It sounded intentional. The workshop continued, people nodded, and we moved on to the next topic.

But my mind didn’t.

On the drive home, that question kept replaying in my head. The more I thought about my answer, the more uncomfortable I became.

I hadn’t lied.

But I hadn’t been completely honest, either.

I had answered with the version of myself I wanted to be, not the one I consistently practiced.

Yes, I journal from time to time. Yes, I reflect. Yes, I write these essays. But if someone asked me to hand them the document that captures the objective of my life, the values I refuse to compromise, the lessons I’ve already paid to learn, the questions I’m still wrestling with, and the direction I’m intentionally moving, I couldn’t do it.

Not because it didn’t exist in my head.

Because it didn’t exist anywhere.

That realization stayed with me for days.

Here I was teaching people how to keep AI conversations from drifting, while quietly allowing the most important conversation in my life to drift a little more each year. Not dramatically. Not enough for anyone else to notice. Just enough that I occasionally found myself asking familiar questions.

When did this become so important?

Is this still who I want to be?

How did I get so far from what I said mattered?

It wasn’t that I’d lost my memory. I’d lost my context.

And that’s when it occurred to me that perhaps I didn’t need another journal.

I needed my own markdown file.

What a markdown file actually does

A markdown file, in the way I teach it, doesn’t preserve the conversation. It preserves the direction of the conversation.

It captures the objective — what we’re trying to accomplish. The context that shaped the conversation. The decisions already made. The constraints that shouldn’t change. The questions still open. And where the conversation should begin next time.

It’s a lean document, deliberately. Not a transcript. A steering file.

The reason it matters isn’t that AI conversations get long. It’s that as they get long, they generate confident, fluent drift. The model doesn’t announce that it’s lost the plot. It performs the plot while doing something adjacent. Because the surface still sounds right, nobody audits it. That’s the failure mode the markdown file protects against — not memory loss, but the invisible substitution of what the conversation was for something that superficially resembles it.

That’s the mechanism. And that’s where the parallel to a human life gets interesting.

The corrupted map

There’s a specific failure mode in prompt engineering that’s worth naming.

If you give a model a system prompt describing behavior it can’t actually produce, you don’t get neutral failure. You don’t get an error message saying I can’t do this. You get confident, fluent drift. The model performs the persona while doing something else, and because the file sounds right, nobody audits it.

That’s what an aspirational personal summary does.

It doesn’t leave you lost. It leaves you certain you’re on course while you’re not.

This is the point most journaling advocates miss. They say write down what matters and you’ll stay grounded. But the first draft anyone writes of a personal summary is almost always a portrait of who they think they should be — not who they actually are. And a corrupted map is worse than no map at all, because it makes drift harder to detect.

Every document has an implied reader, and the moment you imagine one, the should-self starts writing. Most people are unconsciously drafting for a hypothetical biographer. A future colleague. The version of themselves that would be flattered to be read.

Here’s a useful test. Would you be comfortable if a colleague read the line aloud in a meeting? If yes, it’s probably too clean to be true.

The real file isn’t written by the best version of you for the biographer. It’s written by the best version of you for the worst version of you. The depleted, 11pm, mid-conflict, being-offered-something version. That’s the audience. Journaling is expressive. A context file is operational.

Variance analysis

Every organization I’ve watched adopt a new tool says the same thing. We’ve rolled it out. Adoption is strong. The team’s using it. Then someone pulls the telemetry and it turns out the actual usage is eight minutes a week.

Nobody’s lying. Self-report and behavior just diverge, always in the flattering direction.

Your personal file has the same failure mode. And the fix is the same: reconcile the claim against the logs.

Your calendar is a log. Your bank statement is a log. Your camera roll, your sent folder, your browser history. These are the telemetry of your actual life. The summary is the stated adoption.

A value you can’t evidence with an artifact from the last 90 days isn’t a value. It’s a goal.

Those belong in different sections of the file. Values are protected by evidence. Goals are protected by intent. Confusing the two is what allows the aspirational summary to feel like a life inventory when it’s actually a wish list. Once you separate them, the wish list becomes useful — you can see clearly what you want to grow into, without pretending you’ve already arrived there.

This gives you something falsifiable to do instead of a vibe to have.

The log matters more than the summary

In long AI conversations, what saves you isn’t the statement of goals. It’s the record of decisions and the alternatives you rejected and why.

The summary tells you where you meant to go. The log tells you every fork you took and why you took it. When the conversation drifts, the log is what lets you find the specific decision that started the drift — not the vague sense that things went wrong somewhere back there.

The same is true personally.

My life is shaped far more by what I’ve declined than what I’ve declared. The jobs I didn’t take. The friendships I let go quiet. The projects I said no to. The invitations I passed on. Almost nobody records the no’s, which is why we relitigate the same ones for years. We remember declaring things. We forget declining them. And so we make the same declines over and over, treating each one as new evidence about ourselves when it’s actually the same decision we’ve made a dozen times.

A useful markdown file has a decisions section that includes the rejections. Not just I chose stewardship over ownership, but I chose stewardship over ownership even when the ownership path was open, and here’s what closed that door for me.

The log is where identity actually lives. The summary just labels the file.

Maintenance is pruning, not accretion

Context files fail by getting longer.

The instinct, once you have a file, is to keep adding to it. Every new insight, every new commitment, every new question. Over time the file becomes a compost heap of everything you’ve ever thought was important. And when a file is that long, the reader — including future you — stops navigating by principle and starts navigating by inertia.

The skill is deletion. And deletion is the part people won’t do.

A line you wrote at 32 about who you were becoming can steer you for a decade past its expiration. Not because you still believe it. Because you never went back and removed it. It sat there in the file, unchallenged, quietly propagating into every downstream decision.

One stale belief in a context file becomes structural over time. It shapes what you notice as relevant, what you file away as important, what you rehearse to yourself as your story. And because it was written down once, it has the authority of something you decided — even though the you who decided it doesn’t exist anymore.

Maintenance means going back and removing what’s no longer true. Not editing it. Removing it. And that’s harder than adding, because removal requires you to admit that something you once believed is no longer operational. Most people would rather append than delete. That’s how context files rot.

The ritual that makes it self-correcting

Here’s the practice that pulls the whole thing together.

Once a quarter, rewrite the file from scratch. Without looking at the old one.

Then diff it against the previous version.

What you can’t reproduce from memory isn’t operating in your life. It’s stored. The diff is the actual instrument — it shows you which values quietly changed, which commitments went dormant, which questions dissolved without you noticing.

And here’s the hard rule that goes with it: never overwrite the old file. Only append the new one, with a dated reason for what changed.

Because the question isn’t whether you revised. The question is whether you revised right after the value got expensive.

Growth and rationalization look identical in the moment. You can’t tell them apart in real time. The person who chose comfort over integrity last Tuesday and the person who genuinely outgrew a rigid belief are indistinguishable — until you look at the revision history.

Growth is stable across time. You can point to it a year later and it still looks like growth. Rationalization requires ongoing maintenance. It has to be re-narrated every time the underlying situation resurfaces. The dated append log makes this visible. If a value quietly softened right after honoring it became inconvenient, that’s information. If it softened over years, through independent challenges and eventually a considered revision, that’s different information.

You don’t get to decide which one it was in the moment. You get to preserve the record and let the pattern show you later.

The workshop question, revisited

A few months after that workshop, I built my first real markdown file.

The first draft was almost entirely the should-self. Cleaner than my life. Values I aspired to more than practiced. Language I would have been proud to have quoted back to me.

Then I did the variance check. I opened my calendar and my bank statement and my sent folder, and I asked what my telemetry actually showed.

The gap was uncomfortable. Not because I was living badly. Because I was living slightly differently than I’d been telling myself.

I revised the file. Not to lower the standard. To separate the values from the goals, and to move the aspirations to their own section where they belonged — as things I was working toward, not things I’d claimed as already mine.

The blind rewrite three months later showed me something else. Two of the commitments I’d been most confident about at the start didn’t survive the recall. I couldn’t reproduce them from memory because they weren’t operating. They’d been in the file, but not in the life.

That’s the whole point of the practice.

The document isn’t the person. The document is the instrument that shows you where the person and the person’s stated direction have diverged. And divergence, caught early, is correctable. Divergence, allowed to run long enough, becomes the new baseline — and by the time you notice, you’re not drifting anymore. You’ve arrived somewhere.

Most of us update everything else. Calendars, budgets, project plans, quarterly reviews. We update the operational documents of every other project we lead. The one project we rarely update the operational document for is our own life.

The question I got in that workshop wasn’t rhetorical. It was diagnostic. And the honest answer, for most people, is the same one I gave — a smooth deflection that sounds thoughtful and isn’t quite true.

The markdown file won’t fix that on its own. But the discipline behind it — variance analysis, decision logging, quarterly blind rewrites, append-only revision — is the closest thing I’ve found to a working audit of a life in motion.

Not because it remembers the past.

Because it makes drift visible in time to do something about it.

And perhaps that’s the lesson AI unexpectedly taught me. The most important conversation you’ll ever help stay on course is the one you’re having with yourself. And the only way to keep it on course is to build the same instruments you’d build for any other conversation you refused to let drift.

The version of you writing the file wants to be flattered.

The version of you reading it, six months from now, needs to be told the truth.

Write for the second one.