The Distance Between Advice and Execution
- Rich Washburn
- 5 days ago
- 6 min read
Todd called me twenty minutes ago...
He'd just finished something with Codex that he'd been putting off for years. Two old email exports — Dovecot archives from his Bluehost account — that he wanted consolidated into Thunderbird. Original dates, headers, folder structure, duplicates removed. Clean. The kind of task that lives on a list forever because the work isn't hard, it's just miserable. Files, formats, folders, imports, sync issues you don't discover until you think you're done.
Codex inspected the archives, identified the Maildir structure, extracted the compressed messages, converted everything to Thunderbird format, removed four duplicates, imported the result, uploaded folders to his live Titan account, waited for sync, and verified the counts. 793 emails. 13 minutes.
He watched it operate Thunderbird on his computer. Not tell him how to do it. Do it.
He wrote about it on the Eliakim site a few minutes later, and the line that jumped out at me was: "It did not merely tell me how to do the job. It did the job. That distinction matters." That distinction is the whole thing.
After he told me about the email, I said — hold on, let me show you something else. Todd's had a long, vicious battle with expenses. The kind of battle where the receipts pile up, the spreadsheets multiply, and the whole thing gets worse every month because nobody wants to deal with it. I went to GitHub, found an open-source expenses project, and handed the link to Codex. Spin this up locally in a secure sandbox. Set it up for expense tracking. Get it on my phone.
It did. That's the part I'm still processing too — not just that it works, but that the distance between "I found a tool" and "the tool is running on my phone" is now measured in minutes.
A couple weeks before this, I wrote about something I called the light bulb moment. I'd spent an afternoon at my friend Rolf's house, talking through AI in the concrete, "what can this actually do for you" sense. Rolf's been in high-end marine sales for thirty years. He understands markets, deals, logistics, client psychology. He doesn't understand technology the way I do, and that's fine — that's the point.
What clicked for Rolf that afternoon wasn't a feature or a demo. It was the realization that the skill he'd been building for thirty years — structuring intent, communicating it clearly, reviewing output, iterating — was the exact skill that matters now. The execution layer has been automated. The thinking layer hasn't. And the thinking layer is where he was already good. That's the theory. Todd just handed me the lab result.
There's a gap that has existed in every business, every profession, every project since the beginning of organized work. The gap between knowing what needs to happen and making it happen. Between the person who can see the path and the people who have to walk it. We've built entire organizational structures around that gap. Managers exist to bridge it. Project plans exist to sequence it. Meetings exist to coordinate it. Software exists to track it. Whole layers of professional services — consulting, integration, implementation — exist because someone can see what needs to happen but can't make it happen without help.
What Todd watched Codex do twenty minutes ago is the beginning of that gap closing. He didn't get advice about his email. He didn't get a tutorial. He got the work done. The thing that was too tedious, too fragmented, too easy to postpone — the thing that would have consumed a weekend or a hired consultant or a permanent spot on the to-do list — just got done. In 13 minutes. By a machine that understood the instruction and executed it. The gap between "I know what needs to happen" and "it happened" collapsed to almost nothing.
I really do love it when it clicks. That's the thing I'm always chasing when I'm showing someone something new. Some new tool, some new trick, some new way of doing things. The click isn't the "oh that's cool" moment. That's easy. Anyone can be impressed by a demo. The click is when their mental model reorganizes itself in real time — when they go from "this is a neat toy" to "oh, this can actually do the work."
Rolf's click was realizing that AI is not software he has to master. It's intelligence he already knows how to direct. Thirty years of structuring intent, and the bottleneck was never his thinking. It was the execution layer behind it.
Todd's click was more visceral. He handed Codex a real, ugly, years-old problem and watched it cross the boundary from advice into execution. 793 emails. 13 minutes. And then we spun up an expense tracker from a GitHub link in the time it takes to make coffee. Same moment. Different angle. Same click.
Once that click happens, there's no going back. I've watched it enough times now to know the pattern. They immediately start mentally scanning every unfinished project, every abandoned idea, every administrative nightmare, every "I'll deal with that someday" task in their life. The mental inventory changes. The list of things that are "impossible" shrinks. The list of things that are just "tedious" disappears entirely. I keep thinking about what happens when this clicks at scale.
Not just for email migration or expense tracking. For everything. The report that needs writing. The spreadsheet that needs building. The data that needs cleaning. The document that needs drafting. The code that needs writing. The analysis that needs running.
All of these things have lived in the gap between intent and execution. We hire people to live in that gap. We build departments to manage it. We create job titles that are essentially "the person who translates what the boss wants into what the system needs." We've organized the entire white-collar economy around the fact that thinking and doing are separate activities performed by separate people. When the execution layer can take a clear instruction and run it — not eventually, not with supervision, but now — the structure built around the gap starts to look strange.
Here's where I want to be careful, because this is the part where people hear "AI is going to replace everyone" and stop listening. That's not quite what's happening. What's happening is more interesting and more unsettling than that.
The skill that matters — the skill that has always mattered — is the one Rolf has. The ability to look at a situation, understand what actually matters, figure out what needs to happen, and communicate it with enough clarity that someone else can execute. For thirty years, "someone else" meant a person. Now it can mean a machine. The skill transfers. The speed doesn't. Todd didn't need to learn Codex. He needed to hand it a job he already knew how to describe. The thing he'd been putting off wasn't hard to understand — it was just miserable to execute. The gap between "I know what I want" and "I have it" was filled with tedious, fragmented labor that no one wanted to do. That's the gap that's closing.
Todd ended his piece with a line that's better than anything I could write about it: "Just give it one real job. Give it the annoying thing that has been sitting on your list because the work is too tedious, too fragmented, or too easy to postpone. Then judge it by what happens."
That's the whole instruction. Not "learn AI." Not "adopt AI." Not "transform your business with AI." Give it one real job. The one you've been avoiding. The one that's too small to hire someone for and too big to do in five minutes. Then watch what happens.
Rolf got the theory. Todd got the proof. Same moment, different angle. And twenty minutes after Todd's email finished, we were setting up an expense tracker from a GitHub link, because once it clicks, you start looking at everything differently.
The light bulb moment isn't about the machine. It's about the person realizing the thing they've been doing their whole career — structuring intent, directing execution, reviewing results — now has a much faster, much cheaper, much more willing execution partner.
The question isn't whether this scales. It does. The question is how long the people sitting on the old organizational structures — the ones built around the gap between thinking and doing — get to pretend the gap is still there. Because Todd's email migration took 13 minutes. And that gap isn't coming back.
Rich Washburn is a technologist and strategist working at the intersection of AI, infrastructure, and capital. He is Managing Partner and Chief AI Officer at Eliakim Capital.
Comments