If you've been around the founder world for more than five minutes, you've heard about Elon Musk's "algorithm." It's the five-step framework Walter Isaacson described in his 2023 biography.
Jon McNeill's new book The Algorithm is the first time someone who actually ran the playbook from the inside put it on paper. McNeill was president of Tesla during the 30-month period when revenue went from $2B to $20B. He came in steeped in Toyota lean principles and left convinced that Musk's approach was the opposite of lean, and in many ways more powerful.
He's since applied the same framework as COO of Lyft, as a board member at General Motors, Lululemon, and Asurion, and as cofounder of DVx Ventures, where he's launched twelve companies. The book pulls case studies from all of them, which is part of what makes it more useful than the Isaacson chapter most people are quoting.
Here's what's in the book and what I think it means for the founders, operators, and investors I work with.
The five principles
Question every requirement.
Every rule was made by a person, not a department. Get the name. Then challenge it. Requirements from smart people are the most dangerous because nobody pushes back.
The most famous example: Tesla negotiated the first 100% foreign-owned auto plant in China by refusing to accept the joint-venture rule everyone else treated as law. GM, Microsoft, Airbus, and Volkswagen had all accepted it. Musk sent McNeill to Beijing in 2015 with instructions not to.
A smaller and arguably more relatable example: Tesla's online car-buying flow used to take 64 clicks. The team got it down to 12. Part of the savings came from streamlining models and options. The much bigger savings came from cutting more than 40 pages of legal disclosures customers had to scroll through to finance a car. When the team actually researched it, none of that boilerplate was required by law or regulation. Corporate attorneys had added it over decades because nobody ever asked who put it there.
That's the principle in miniature. The rule was real. The reason was not.
Delete every possible step.
Cut more than feels comfortable. If you don't add back at least 10% of what you deleted, you didn't cut enough.
The body shop story is the famous one. Tesla's manufacturing chief looked at the century-old assumption that a car factory needs a body shop with hundreds of robots stamping and welding parts together. He asked why. The answer was tradition. Tesla replaced the entire shop with single giant castings produced by a few enormous machines. Cheaper, faster, more consistent quality, and a five-to-seven-year lead on every other automaker. There's now a global shortage of the casting machinery competitors are scrambling to buy.
The deeper point isn't the casting. It's that the body shop existed because every car company had one. Questioning the requirement and deleting the step usually go together. You can't delete what you haven't first dared to question.
Simplify and optimize.
Only after questioning and deleting. The order matters. Optimizing a process you should have killed is one of the most expensive mistakes in operations. Most lean methodologies start here. The Algorithm argues that starting here is exactly the trap, because you end up making bad processes more efficient instead of asking whether they should exist at all.
Accelerate cycle time.
Compress every loop: design, decision, iteration, feedback. Faster cycles compound learning faster than competitors can react.
The Lululemon example sticks with me. When the company was producing apparel for the Canadian Olympic team, the design and production cycle had to compress dramatically. Rather than email back and forth with overseas factories about color matching, designers and product managers flew to Asia together so they could verify shades of red on-site in real time. The cycle that used to take weeks took days. Same people, same factories, faster loop.
This is the principle most founders miss because it doesn't feel like a principle. It feels like a logistics decision. It's actually a strategic one.
Automate last.
Tesla's near-bankruptcy moment came from automating the Model 3 line before the process was simplified. It was close to a billion-dollar mistake that forced them to build cars by hand in a tent in the parking lot. Humans optimize the process first. Robots come last, or you bake complexity into expensive systems you can't unwind.
The discipline this principle enforces is making yourself earn step five by completing the first four. Automation feels like progress. That's why everyone skips ahead.
What I want founders to actually take from this
The framework is easy to understand and hard to argue with. Here's where I think founders most need to wrestle with it.
The order really matters
Most founders I work with already do some version of "simplify" and "automate." What they don't do is run the steps in order.
They optimize before they delete. They automate before they simplify. They build a workflow in n8n or Zapier to handle a process that shouldn't exist. They hire a contractor to clean up a service offering that should have been killed last quarter.
This is the trap Toyota lean training builds into people. Lean improves what's there. The Algorithm asks first whether what's there should be there at all. Different starting question, completely different outcome.
Question every requirement is a permanent job
By month 18 of any company, you'll have requirements that started as good ideas, became habits, and ossified into rules nobody remembers making. Standup at 9am. Demo every other Friday. The CEO approves all hires. Sales calls follow this five-stage script.
The discipline is to make those rules name-bearing again. Who decided this? Are they still here? Does the reasoning still hold?
The Tesla loan disclosures are the cautionary tale. Forty-plus pages of legal text that customers had to click through to buy a car. None of it required. All of it added by well-meaning attorneys over decades. Every company has its own version of those forty pages. Most of you reading this can name three of yours right now.
Many founders avoid this work because the people who made those rules are sitting two desks away. The alternative is running a company shaped by past decisions you'd no longer make.
Automate last is the principle most founders are getting wrong right now
Every founder I talk to wants to bolt agents and RAG and workflow automation onto their operations. Many of them are automating processes that are broken.
The right sequence is: question whether the process should exist, delete what you can, get a human to run a simplified version until it works, then automate. AI doesn't change this sequence. It makes the consequences of skipping it worse, because automated bad processes scale faster than manual ones.
Tesla's Model 3 tent in the parking lot is the version of this story founders need to remember. They had the most automated car factory in the world. It didn't work. They had to rip out automation and build cars by hand until the process was simple enough to be worth automating again. If a company with Tesla's resources almost died from getting this order wrong, your AI-first ops stack should probably get a second look.