He Has Never Written a Line of Code. His Company Sells to 35 Federal Agencies.

Levtov is the founder and CEO of Craft, a 12-year-old San Francisco supply chain intelligence company. In an interview published by Crunchbase News on August 27, 2026, he described the company as working with 35 federal agencies and generating double-digit millions of dollars in annual recurring revenue. He has raised $42 million for it.

He has also never written a line of code, and said so plainly: "I've never written a line of code in my life. And I still haven't written a single line of code."

His background is not the usual one. He grew up in a family of musicians who emigrated from the Soviet Union to England, started cello at four, trained at the Royal College of Music, did a Columbia-Juilliard exchange, and took an English literature degree. Then Goldman Sachs, Stanford Business School, an ad-tech startup called Spot Runner that grew from 10 people to about 200 while he was there, a stint as a VC at Venrock, and operating roles at Crunchyroll and Deutsche Telekom.

The reason this matters is not that it is inspirational. It is that Levtov is unusually specific about what the missing skill actually cost him, and most founder profiles are not.

Two things, and both are measured in time.

The first is speed at the very start. "My first coder was literally a $20 an hour Odesk or Upwork person," he said. A technical founder writes the first version on nights and weekends. A non-technical founder has to find a developer, explain the vision, evaluate whether the result matches it, and then find the money to pay for it. That is four steps where the other person has one.

The second is worse. He hired the wrong technical people and lacked the expertise to notice quickly. He described it as an inevitable gap between the non-technical founder's knowledge and the actual innards of the software the business lives on. He believes those hiring mistakes slowed the company's growth.

That second cost is the one first-time founders underestimate. Slow initial builds are annoying. A wrong senior engineering hire that runs for nine months before anyone can name the problem is expensive in a way that does not show up on any dashboard.

Partly, and in a way that was more discouraging than exclusionary.

Raising in London in 2015 and 2016, he read venture fund websites that said, in effect, we back technical founders. He described the reaction as a moment of clarity: "oh I guess that means not me, right?"

But he does not claim he was shut out. He had Stanford and Venrock on his resume, and he raised from Downing Ventures in the UK and later Uncork Capital after moving back to Silicon Valley.

That nuance is worth sitting with. The stated preference cost him confidence and time, not access, because he had other credentials that read as legible risk reduction to an investor.

If you have neither a technical background nor a brand-name resume, the honest read is that you are working against a real filter. The response is not to argue with it. It is to bring evidence that substitutes for it: paying customers, a signed pilot, a waitlist with real names, or a technical cofounder who is actually committed rather than advising.

The breakthrough came from outside the product.

Craft started as a failed attempt to build an enterprise social network. Along the way the team built company profiles by scraping corporate websites, job pages and management pages. Those profiles started ranking in Google, eventually appearing in about 100 million search results a month and drawing roughly 2.25 million organic visitors.

Then someone at Lockheed Martin got in touch and pointed out that this data could track changes across a sprawling supply chain: hiring shifts, executive departures, new product offerings. Lockheed became the first enterprise customer.

In 2020 the U.S. Air Force reached out about monitoring 300,000 companies in the defense industrial base. Craft closed a five-year, $6.5 million deal 94 days later.

"We figured out that our company is actually a supply chain company, and we haven't looked back since then," Levtov said.

His observation about those customers is the part worth underlining. They were not developers asking for better developer tools. They were business users with business problems, and hearing a business problem correctly is a skill that has nothing to do with writing code.

Levtov does not say that, and neither should you.

He grants the logic: a technical founder has "a direct line between the business concept and the code in which it's executed." No translation layer, no lag, no risk of the built thing quietly diverging from the intended thing.

What he argues instead is that the advantage is stage-dependent. Technical founders have the edge earliest, when the job is to build something that works at all. As a company scales, the balance shifts toward hiring, selling, positioning and dealmaking. At different points, he said, it is all about the tech, and at others it is all about the dealmaking, or the positioning, or the marketing.

His conclusion is unglamorous and probably right: "It really just takes both. It takes both sides. I think if you can have a technical founder and a non-technical founder, you're probably in the ideal spot."

AI complicates the debate, since software can increasingly be produced without traditional coding expertise. Levtov stops short of claiming technical founders no longer matter, and the gap he described is not about typing code. It is about being able to evaluate the person who does.

Buy evaluation capability before you buy building capability. Before your first engineering hire, find one experienced engineer who will do paid technical interviews and code reviews for you a few hours a month. This is the cheapest insurance available and almost nobody does it.

Sell before you build. Levtov's largest deals came from customers who found the data useful, not from a roadmap. Letters of intent, paid pilots and design partners are all available to someone who cannot code and are more persuasive to investors than a prototype.

Write specifications like a product manager, not like a client. The gap Levtov describes narrows when the requirements are precise, testable, and written down. Vague briefs are where non-technical founders lose the most money.

Structure the first hire to be reversible. Contract to hire, a scoped paid project, or a three-month trial. You will get one of these wrong. Make the wrong one cheap.

And keep the business case somewhere other than your head. If you are mapping customers, hypotheses and hiring plans for the first time, a planning tool like Foundra, a shared doc, or a spreadsheet all work. The tool matters far less than whether your engineer can read what you actually believe about the customer.

Hiring an agency and calling it a product team. Agencies ship deliverables. Products need someone who owns the outcome after launch.

Giving away large equity to the first technical person out of gratitude rather than assessment. Vest everything, cliff everything, and treat the first six months as an evaluation for both sides.

Waiting for a technical cofounder before starting. The strongest way to attract one is to arrive with customers, revenue or a signed pilot. Levtov's data was ranking in Google and drawing millions of visitors before anyone knew what the company was.

And apologizing for the gap in investor meetings. Investors can smell defensiveness. State the plan for covering the technical risk and move to the part where you are strong.

Levtov built a company with 35 federal agency customers and double-digit millions in ARR without writing code, raising $42 million along the way.

The real cost of being non-technical is not slow building. It is being unable to quickly recognize a bad technical hire, which he says slowed the company's growth.

Investor preference for technical founders is real but not absolute, and other credentials or customer evidence can substitute for it.

Craft's pivot came from a customer at Lockheed Martin noticing a use for existing data, which led to a five-year, $6.5 million Air Force deal closed in 94 days.

The advantage is stage-dependent. Technical skill matters most at the beginning, and hiring, selling and positioning matter more as a company scales.

Can I start a software company without knowing how to code?

Clearly yes, and the harder question is whether you can evaluate the people who do. Solve that before your first engineering hire rather than after.

Should I learn to code before starting?

Enough to read code, understand what your team is deciding, and ask useful questions is worth a few months. Aiming to build the product yourself usually is not, unless you enjoy it.

Do I need a technical cofounder to raise money?

Not strictly, and it removes friction. If you cannot find one, bring traction that answers the same underlying worry: can this team actually ship.

How much equity should a first technical hire get?

It depends on stage, salary and whether they are a cofounder or an early employee. The important part is standard vesting with a cliff, so that a mistake is recoverable.

Are AI coding tools closing this gap?

They lower the cost of a first version considerably. They do not yet replace the judgment needed to run a system in production or to tell whether an engineer is doing good work.

What is the fastest way to prove a non-technical founder can sell?

Get one paid pilot from a customer who is not a friend. It answers more investor questions than any deck slide.

1. Crunchbase News: Craft's Ilya Levtov Never Learned To Code. He Built A Software Company Anyway.
2. Crunchbase News: How An Educator Built An Edtech AI Startup With $63M From VCs
3. Crunchbase News: How A Teenage Carpenter Became The Founder Of AI Construction Startup Trunk Tools
4. Crunchbase News: From Humanities To AI, How Ali Hussain Built Fintech Tabs Into A $400M Startup
5. Crunchbase: Craft company profile
6. Crunchbase News: The Week's 10 Biggest Funding Rounds, August 28, 2026