Three questions stood between thousands of offers and a check for $70,000 that changed everything
Listen to the audio version of this article (generated by artificial intelligence).
Hundreds of business plans, whose founders worked on them day and night, died on a single desk in Boston.
It was the office of George Doriot, a professor at Harvard Business School who has an eye for clean suits and unproven companies.
Decades before anyone coined the phrase “venture capital,” Doriot had already begun applying it, one rejection at a time.
His students remembered one rule above all others, because Doriot kept repeating it: Think of supporting a first-rate founder with a second-rate idea. Never back a second-rate founder with a first-rate idea.
This rule got into place in 1957, when a 31-year-old MIT engineer named Ken Olsen came looking for capital to build computers. This idea horrified the financial men of the era, as a series of expensive corporate failures made “computer” practically a curse word. So Dorio’s team advised Olsen to drop it from the field and drive with the products instead.
I became a company Digital Equipment Company. Doriot’s company, American Research and Development, paid $70,000 for a 70% stake. By the time this stake was finally distributed to shareholders, it was worth more than $350 million. This means a return of up to five thousand times, and the trade proves that venture capital can work.
Dorio saw thousands of proposals. Of these, he financed a small piece. But his fortune came from this filtered approach.
Last week, I showed you how the biggest fortunes of this era are being created – Spark Capital’s early stake in Anthropica Roth IRA was built on pre-IPO stocks Roblox Company (RBLX)Genentech’s IPO has been opened. And I told you that the velvet rope that kept ordinary investors away from these trades is finally starting to move.
I also told you that access is just the starting point.
Today’s issue is not another story about a fortune someone else has already made, but with the actual mechanism I use before a single dollar of my money goes into a private deal.
Three questions.
The same three, asked in the same order every time.
I’ll walk you through each one in enough depth that you can manage them yourself in the next deal that crosses your desk, long after that issue has closed — who’s actually running the company, whether the thing they’ve built solves a problem people can’t walk away from, and whether the clock on the specific opportunity and the clock on the broader market are pointing in the same direction simultaneously.
The number behind every great project story
I want to give you an uncomfortable number before we go any further.
Nearly nine out of ten early-stage bets fail to return investors’ money. Across a typical portfolio, something like one out of ten produces a big winner that makes the entire portfolio tick. And more than half of the positions you take will likely return you less than you invested. These are the normal calculations expected for investing in early-stage companies, even for professionals who do it for a living.
If you walk through a newly opened door and invest your money in the first company you see, you will simply find a faster path to the “nine out of ten.” The door takes you into the room. The framework is what keeps you grounded once you’re inside.
So today I want to introduce you to the actual framework. This is the same three-part framework I use at every company before I recommend one – to members, or to myself.
I call it PPT: People, Product, Timing.
People: The most important candidate
If I had to tip the three, I would honestly call it closer to people, people, people. Or the first-class Dorio base, which was rebuilt for 2026.
You can find the perfect product launched at the perfect moment, and it will still fail if the people running it can’t execute under pressure, adapt when their first plan doesn’t work, or pull the team together during the inevitable phase where something doesn’t go right.
Arthur Ruck followed this framework when he supported the young engineers “The Traitorous Eight” who escaped from William Shockley’s laboratory – a story I shared with you a few days ago. Rock had no product to evaluate and no revenue to design. He bet on the people, and he became the bet Fairchild Semiconductorthe company that seeded almost everything we now call Silicon Valley.
The tricky part is that the “right” founder rarely looks like what you expect on paper. Take Toby Lutke, who co-founded it Shopify company (place). He left traditional education in Germany after 10th grade to enter into hands-on computer programming training, having already taught himself programming by rewriting his own video games around the age of 11. On a CV, “left school at sixteen” can seem like a red flag. In fact, he’s been doing the actual work — writing software every day — longer than most of his eventual competitors.
Lutke and his co-founders set out in 2004 to open an online snowboard store. He was so frustrated by existing e-commerce tools that he built his own store management platform. This side project became Shopify.
And the lesson generalizes long after the dropout: The signal you really want—deep, obsessive competence in the specific problem the company solves—often never shows up on a traditional resume. You have to look beyond the credentials and evaluate the person. Doriot had described Lütke as a first-class founder on the horizon.
In practice, this means checking the boring stuff first: what those people have actually built before, where their claims can be independently verified, and whether the people they’ve worked with — other investors, other founders, former colleagues — describe them the way the founders describe themselves. Most candidates fail this process.
Product: Does it solve a problem that people can’t live without?
Once people review the product, the next question is simpler than most investors ask: Does this product meet demand that is so pressing that customers will tolerate an incomplete early version just to get access to it?
Google It is textbook case. By the late 1990s, the Internet had exploded and provided more information than anyone could organize, and search became the only reasonable way to navigate what had just been created. This urgency is why Andy Bechtolsheim wrote his famous check for $100,000 on the spot — which was made out to a company that had not yet been founded. Early investors who clearly saw demand were rewarded on a scale that is still cited in venture circles decades later.
I bring this up now because I think we’re in the early stages of an equally big story for which there’s unmet demand: physical AI — robots that can do real work in the real world.
But the real choke point for physical AI is the training data. A language model can learn from all of the text on the Internet, because that text already exists. The robot has no equivalent shortcut – each skill must be taught through physical demonstration, recorded frame by frame, often with a human operator physically guiding the machine through the task in real time. Industry data on this has moved quickly: the cost of collecting one hour of labeled, usable training data has fallen sharply over the past couple of years thanks to better tools, but it remains an expensive and physically constrained process — and a robot that needs to deal with endless real-world diversity reliably needs vast amounts of that data, well beyond a few hundred hours of demonstrations.
Whichever companies figure out how to produce this training data cheaply, and at scale, are sitting on what I believe will become the defining choke point of the physical AI era — the equivalent of what chip design and cloud infrastructure have become in the last decade of AI. This is the “insatiable demand” test, applied to where I think the next major AI build is actually headed.
Timing: two hours, not one
The last part of the framework covers two separate things that people often collapse into one thing.
The first hour is your entry into a specific company. It is early enough to buy before the bulk of the value is captured by the previous rounds. But it was too late that the company proved the success of its basic idea. This balance is exactly why questions about people and products come first; Which allows you to judge whether an early-stage company is already exposed to enough risk to be worth the timing bet. Remember Genentech: The investors who rushed to Wall Street’s door in 1980 were buying pure potential. Those who understood what Swanson and Boyer had already proven in the laboratory were making a judgment call.
The second hour is the broader environment in which you are investing. On this front, I believe current conditions are as favorable as they have been in a long time. I explained the math a few days ago: The largest tech companies are pumping nearly $700 billion into capital projects this year. About $2 billion every day… and they’re competing with each other for the same capacity.
Companies with this much money and this little time tend to buy what they cannot build fast enough. This competitive pressure is compressing the timeline between “promising private startups” and “life-changing exits for early investors,” whether the exit comes through an IPO or an acquisition by a giant corporation racing to secure a capability it does not have enough time to develop from scratch. Increasingly, acquisitions are the finish line that early investors aim for.
What this tire actually buys you
I want to conclude as Doriot would conclude: Nine out of ten of these bets won’t pan out, and no framework — mine or anyone else’s — changes that basic calculation. Even Bessemer Venture Partners, one of the oldest firms in the space, maintains a generic “anti-portfolio” of the winners it has passed — Google, Apple, Airbnb, FedEx. Anyone who promises you a system that turns investment investing into a sure thing is lying to you.
What people, product and timing actually do is change your odds. It won’t turn a bad bet into a good bet. But if applied consistently, it filters out startups that would never make it past the “nine” — founders who don’t hold up under scrutiny, products that solve problems no one can solve urgently, and companies that are too late or too early to pay attention to. What remains is a much smaller list, and one that is worth betting capital against.
Doriot read thousands of plans to find a single digital device. However, you won’t need to read thousands. on July 30 at 1 p.m. ETI will learn about this exact framework directly, in real time, and apply it to it One specific AI company I recommend for free. You’ll see exactly how your People, Product, and Timing questions play out against the real deal, with a real chance to work on them that same day… plus two bonus recommendations reserved exclusively for members who join that day.
