Meysam Valiolahi.

notes · the Mr. Sarang series

The speed trap

The only sound in the shop was Mr. Sarang's wooden spoon against the glass jar — he was stirring a stiff vanilla ice cream with great patience, humming a line of Hafez under his breath. That afternoon I learned why "free speed" is the most expensive illusion of our century.

the Mr. Sarang series second part · based on the wild horse

The neon shop across the street

Watching him work, I remembered last week's conversation — wild horses and broken compasses. I asked: "Mr. Sarang, what do the younger crowd say when they see you still making ice cream by hand? Don't they call you behind the times?" I told him that every conversation with him feels like passing a university term.

He adjusted his magnifying glasses, laughed from above them, and said: My boy, technology surely advances the person who knows their constraints. You software engineers are making exactly the mistake of that garish shop across the street.

He pointed at a shop that had opened six months earlier, all neon lights and a bizarre name: AI Ice Cream. Fully automatic ice-cream machines, he said. They buy ready-made base. Their speed is ten times mine. And do you know why I know they'll be bankrupt within a year? Because they believe speed has no cost.

I jumped in: "But Mr. Sarang, my world is different. We have tools that apparently impose no limits at all." And I described the two revolutions shaking our industry: BaaS services — the ready-made base, so you never have to get milk from the farm — and AI — the ice-cream machine that writes hours of code in seconds.

He put the spoon down and turned his hearing aid toward me. Let me show you what's missing, in your own language. In his leather notebook he sketched both production processes, side by side. Then he tapped his finger on random quality control and said: Look at what got removed: tasting, temperature adjustment, talking to the customer. You compress processes with BaaS and AI — and as someone who understands machines under the hood, I'm telling you: this is the biggest cost trap of the century.

The paradox of free speed

When your managers see these tools, they get drunk on them, he went on. They hold meetings titled "AI First" and say: you must ten-times your deployment!

"Exactly," I said. "I heard that very sentence in a meeting last week."

He smiled: Then show me what got built. I opened my notebook and showed him the real chart of an eight-person team that had taken its deployments from 4 to 38 per month with Firebase, Copilot, and Claude Code. He stared at it for a few seconds.

Exactly what the neon shop experienced. In the first month, profit doubled. By month six, electricity and repairs had eaten all of it. The organisation's focus drifts from "does this feature have business value?" to "how do we build it faster?" You become a factory producing ice cream at terrifying speed — without knowing whether anyone actually wants that much ice cream.

Then he mentioned the slogan: You young people have slogans, right? That big social-network company, Facebook, ran for years on "move fast and break things". Same logic — speed over quality. The result was privacy crises, Cambridge Analytica, heavy regulation, and eventually a changed slogan: "move fast with stable infrastructure". Which is what I learned in my first year:

You cannot keep moving by constantly breaking. The day comes when the cost of the breaking exceeds the profit of the speed.

He waved his hand toward the street and reached for military history. 1944 — the British design an operation called Market Garden. Montgomery promises: "the war will be over by Christmas." Speed itself becomes the strategy. The paratroopers advance fast, take the bridges — but the supply lines lag behind. The front forces are encircled without support. They had built "a bridge too far" — exactly like a team that prototypes with AI at full speed but has no production infrastructure underneath. Everything moves beautifully until the market's reality arrives.

The "AI First" trap: when the tool becomes the strategy

"But Mr. Sarang," I protested, "in our company 'AI First' is the official slogan. If we don't use it, competitors pull ahead."

He shook his head, released Hafez for the moment, and quoted Rumi instead — then delivered the verdict: "AI First" is not a strategy. It is a strategy error. No successful company succeeded because it used AI. Companies succeeded by solving a real problem — AI was merely their instrument.

I remembered Michael Porter in Harvard Business Review, years ago: operational effectiveness is not strategy. Companies can be operationally excellent and still perform poorly, because they never allow a distinctive strategy to emerge. (Porter, "What Is Strategy?", HBR, 1996.) That's exactly it: when everyone uses AI, using AI is no longer a strategy — it is table stakes.

Nicholas Carr said the same about IT twenty years ago, he added. And then another war: In World War One, Germany had the Schlieffen Plan. Simple logic: speed equals victory. Take France in six weeks, then turn to Russia. Again — speed treated as strategy itself. The rapid advance stretched the supply lines, exhausted the troops, and stalled at the Battle of the Marne. Four years of trench warfare followed. The same mistake repeated: speed mistaken for the strategy, when it is only ever an instrument of strategy. When speed itself is the strategy, it's a trap.

The apprentice and the machine

He put the spoon down again, remembering his youth. Thirty-odd years ago I bought my first semi-automatic ice-cream machine. Owning it then was like owning AI today. I had a young apprentice — very bright. He learned the machine in a single day. When it ran, his output was triple everyone else's.

The magic of fast output fools everyone — even smart people. As the shopkeeper, seeing that volume, I thought the boy was a genius. He believed it too: because his speed was high, he no longer needed the fundamentals. Over time he grew confident enough to start advising the older workers.

"So what was the problem? Speed is good, no?"

Here is the problem: that machine never transferred the experience of ice-cream making to the boy. It only amplified his arm. He didn't know that when the milk's fat percentage changes, the freezing temperature must be adjusted. He just pressed "start". He struck the table with his finger. A few months passed. On the surface we were breaking sales records. But under the skin of the shop, disaster was unfolding. The machine was wearing out from misuse. Waste from bad texture had tripled. And nobody noticed that the original taste — our competitive advantage — was quietly disappearing. Until one day customers came back and said: your ice cream doesn't taste like it used to.

Years later, he found out what the big management schools call this: the illusion of mastery under automation. And when a tool makes the work feel effortless, the human brain drops out of deep learning. My apprentice thought that because he controlled the machine, he knew ice cream. The same thing can happen to you: a talented young engineer's output multiplies with AI — and since inexperienced people carry high confidence, they slowly develop an illusion of knowledge. What happened to our ice cream's taste can happen to your product.

Gary Pisano described exactly this tension: tolerating failure requires intolerance for incompetence; readiness to experiment requires rigorous discipline. (Pisano, "The Hard Truth About Innovative Cultures", HBR, 2019.) Mr. Sarang's apprentice was a victim of precisely that imbalance: the shop tolerated failure generously, but had no rigorous standard for competence.

Then he offered a real-world example so I'd know this wasn't just an ice-cream parable. Boeing made exactly this mistake. Rival Airbus announced a new model; Boeing rushed. Instead of real design, they bolted bigger engines onto old airframes, then added an automation system — MCAS — without training pilots, without even telling them it existed. Three hundred and forty-six people died. In your industry nobody dies, but your product does — and it takes your users with it.

And then the sentence he struck the table for: The lesson of that day was this: when you hand a powerful tool to an inexperienced person and leave them unsupervised, the main culprit is not them. The main culprit is you — the one who traded quality and principles for temporary speed. That's where the hidden costs start growing.

The iceberg of hidden costs

"Where do these hidden costs hide?" I asked. "Our Jira dashboard shows deployment speed going up."

He laughed: As an engineer you should know better than me, my boy. Costs are like icebergs. Then he opened the layers one by one, in his own language:

Speed without understanding "why" is just speed on the way to bankruptcy.

Mr. Sarang's three operating principles

He showed me a photo of his old refrigerator. Treat AI and BaaS like a refrigerator, not like an ice-cream machine. A refrigerator doesn't spoil the ingredients. It buys you time to spend on designing the flavour. For that, I have three principles:

  1. Constraints first. Know your constraints before you admire your tools. Technology advances the person who knows their limits — it drowns the one who doesn't.
  2. Speed is an instrument, never the strategy. The moment speed becomes the goal, you are re-running the Schlieffen Plan with a deployment pipeline.
  3. Amplifiers amplify everything. A machine multiplies a good recipe and a bad one alike. Fix the recipe first.

Winter is coming

At the end, he drew one more diagram in the notebook: a seasonal calendar of his shop's year. Then he looked up, pointed coldly at the neon shop across the street, and said:

We don't just know summer here. We plan for winter too. Unfortunately your startup rivals only know summer: endless sprints, fast deploys, celebrations. They think high speed means they've won. But I know winter is coming. When the funding crisis or the recession arrives, those who only learned to press the machine's button go bankrupt with a mountain of spoiled ice cream and unpaid invoices. Then he told the old Saadi story about the dervish in famine-stricken Alexandria — contentment as the wealth that survives winter. And indeed, HBR has made the same point about growth: companies that only recognise growth slowly fade away. (Olson & van Bever, "When Growth Stalls", HBR, 2008.)

I was quiet. His words explained, better than any tech conference, why the glittery startups vanish by year three. He closed the fridge, wiped his hands, and said while locking up:

In software engineering, technology never reduces cost — constraint-based thinking and effectiveness do. New tools only speed up the execution of your decisions. If your decisions are wrong, new tools just speed up your bankruptcy.

References