“But I sent the payload through the dry-run endpoint, and it still committed the charge. How is that possible?”
Marcus didn’t wait for the Slack notification sound. He sat staring at the channel, watching the little pencil icon dance next to Dave’s name. It was , that specific hour when the office air conditioning usually shifts into a lower, more mournful hum. Marcus had been running platform engineering for , and he was currently witnessing a minor physics miracle: a “read-only” flag that had somehow triggered a five-figure refund.
The Slack thread already had eleven people in it. There were product managers from the Tokyo office, a distraught account executive, and three junior devs who were mostly just there to witness the car crash. None of them had an answer.
“Check the 2019 override flag in the metadata header. If the customer ID ends in a prime number and the currency is JPY, the dry-run wrapper ignores the boolean. It’s an old legacy fix for the SoftBank integration. Just manually null the ‘execute’ key and retry.”
– Dave, via Slack
Marcus typed a quick “Thanks, Dave. Life saver.”
He didn’t ask how Dave knew. He didn’t ask why a legacy fix from was still overriding the core API logic of their primary revenue engine. He didn’t ask because he was just relieved the bleeding had stopped. This was the forty-first time this quarter Marcus had reached out to Dave for a “quick pointer” on the billing module. He had never once asked why he still had to.
The Transaction Tax of Legibility
When we look at a scenario like this, we tend to categorize it as a failure of documentation. We treat it as a technical oversight, a bit of “tribal knowledge” that simply needs to be transcribed into a Confluence page. But this is a clinical misunderstanding of why information aggregates in a single human mind.
In most brownfield systems-those sprawling, revenue-generating monoliths that have survived three management changes and four rebranding exercises-the concentration of knowledge isn’t an accident. To the organization, Dave is a “single point of failure.” To Dave, being a single point of failure is the only thing that makes his job feel secure in a world of shifting roadmaps and quarterly layoffs.
We can analyze this through the lens of what I call the “Transaction Tax of Legibility.” In a technical sense, the billing module is likely a mess of “Eventual Consistency”-a term we use to describe systems where data doesn’t update everywhere at once, but promises to eventually. Glossed for the layperson, it’s like a game of telephone where the last person in the chain eventually gets the right message, provided no one hangs up.
But there is a human version of eventual consistency, too. It’s the idea that if Dave explains how the system works, his value to the company will eventually stabilize at a lower, more replaceable level. He is incentivized to stay in the “Eventual” phase forever.
The “Translation Tax” represents time spent waiting for experts to explain undocumented logic.
Rewarding the Hero, Punishing the Processor
The problem isn’t that Dave is a hoarder. It’s that the company’s incentive structure is built on a fundamental paradox. Organizations claim to value “scale” and “process,” yet they consistently reward the “heroics” of the individual who can fix the unfixable at .
If Dave spends documenting the SoftBank override and the prime-number currency logic, he isn’t rewarded with a promotion; he is rewarded with a smaller sphere of influence. He has converted a “defensible position” into a “shared commodity.”
I recently tried to explain the mechanics of cryptocurrency to a neighbor, and I found myself hitting the same wall. The more I simplified the cryptographic hashing and the decentralized ledger, the more he looked at me like I was a magician revealing that the rabbit was actually just hidden in a false-bottomed hat. People don’t actually want to know how the trick works; they want to know that someone is there to perform it when the audience is watching.
The Dangerous Secret of the Tank
William P.K. notes that aquarium “plumbing guys” never write anything down. They like the fact that the curator has to call them at dinner.
William P.K., an acquaintance of mine who works as an aquarium maintenance diver, once told me that the most dangerous part of his job isn’t the sharks or the cold water-it’s the “hidden plumbing.” Every large-scale tank has a custom-built life support system that only one or two guys truly understand.
If those guys leave, the pH levels drift, the nitrate spikes, and suddenly you have $40,000 worth of dead coral. William noted that the “plumbing guys” never write anything down. They like the fact that the curator has to call them at dinner if a pump starts rattling. It’s a quiet power.
When you look at the data, the reality of this “knowledge tax” is staggering. In a study of mid-market engineering teams, it was found that the “Translation Tax”-the time spent waiting for a specific person to explain a specific line of code-can account for as much as 31% of the total engineering budget. That isn’t 31% of the work; it’s 31% of the *capacity* simply vanishing into the air while people wait for a Slack reply.
Why External Delivery Fails
The traditional way to solve this is to hire a technical writer or tell Dave to “take a Friday for docs.” It never works. It doesn’t work because documentation is static, while the code is a living, breathing creature that changes every time a new customer signs a contract with a weird requirement.
This is where the model of the “external delivery partner” usually fails, too. Most agencies want to take the mess, go into a , and come back with a “new” system that is just as undocumented as the old one. They sell you the replacement of the problem rather than the resolution of the risk.
To actually break the cycle, you have to change the environment in which the knowledge exists. You have to move from a “hero-based” delivery model to a “legible” delivery model. This is the core of what Limestone Digital does when they embed an AI Velocity Pod into a team.
“Every question Dave answers at 11 PM is a failure of leadership, not a success of engineering.”
They don’t just ask Dave to write things down; they use AI architects and senior engineers to “observe” the system as it runs, mapping the actual delivery throughput against a baseline.
Making Knowledge a Byproduct
If you want to know how the billing module works, you don’t ask Dave to write a book. You put a pair of engineers inside his daily stand-ups, let them commit to the same repository, and use AI-assisted workflows to document the “why” as the “what” is happening. You make the knowledge transfer a byproduct of the work rather than an extracurricular chore that Dave feels threatened by.
Because here is the uncomfortable truth: every question Dave answers at is a failure of leadership, not a success of engineering.
When Marcus got that answer about the SoftBank flag, he felt a surge of relief. But what he should have felt was a cold chill. He had just confirmed that his most critical revenue pipeline is dependent on the memory of one man who is currently answering DMs while probably trying to eat dinner.
The organization is paying for this. They are paying for Dave’s security with their own scalability. They are running a multi-million dollar business on a “miracle-based” architecture.
Treating Knowledge Like a Leak
If we want to fix this, we have to stop treating “institutional knowledge” like a precious resource and start treating it like a toxic leak. A leak needs to be contained, mapped, and integrated back into the system so it can be managed by anyone. We have to make it more profitable for Dave to be the guy who built the map than the guy who knows the secret path through the woods.
Until the incentives change-until we start measuring “legibility” with the same fervor we measure “uptime”-the billing module will remain a mystery. And Dave will keep answering at , not because he has to, but because the company has silently told him that his silence is the only thing keeping him safe.
In the end, it’s not about the code. It’s about the fear of becoming ordinary. In every legacy system, there is a Dave who believes that if the system becomes clear, he becomes invisible. Our job isn’t to take his magic away; it’s to give him a bigger stage where he doesn’t have to keep the rabbit in his hat just to get a seat at the table.