Ideating

Of all the definitions of a startup, perhaps the best is one from Paul Graham: a startup is about growth. A startup is a business built to grow extremely rapidly. Rapid growth usually requires some sort of new technology that invalidates the assumptions of incumbents, whether incumbent politicians, incumbent businesses, or incumbent ideas.

The gap between stated preference (what is praised) and expressed preference (what is bought) is an inexhaustible source of startup ideas.

You can condemn hypocrisy. Or you can arbitrage inconsistency.

A framework I use often is the evolution from the physical version to the intermediate form, and then to the internet-native version. If you’re into electrical engineering, you can think of this as the evolution from analog to analog/digital, and then to native digital.

We transitioned from paper to a scanner that scans paper into a digital version, and then to a native digital text file that begins life on the computer. We transitioned from face-to-face meetings to Zoom video meetings (a scanner of faces), and then (soon?) to native digital VR meetings. We transitioned from physical cash to credit cards and PayPal (a scan of the pre-existing banking system), and then to the native digital version of money: cryptocurrency.

Once you see this pattern, you can see it everywhere. Look for places where we’re still stuck at the scanned version—where we’ve taken an offline experience and put it online but haven’t fundamentally innovated. These are opportunities for innovation.

For things we can do completely on the computer, productivity has measurably accelerated. Emailing something is 100 times faster than mailing it. But a slow human still needs to act on it. One theory: humans are now the limiting factor.

Representing a project online (in something like Google Docs) may not be the huge productivity win we think it is. Humans still need to comprehend all those documents. The problem may be in the analog-to-digital interaction. If we want to actualize as fast as we can compute, zero-delay robotic task completion will be the true productivity unlock. We haven’t gone full digital yet. As long as humans are still in the loop, we won’t get the full benefits of digital productivity.

Many industries will evolve like this:

1. Human Service

2. Semi-automated service

3. Fully automated

Human, then human/machine pair, then machine.

The pitch of “140 characters” sounded trivial, and the pitch of “reusable rockets” seemed unrealistic, but those ideas resulted in Twitter and SpaceX, respectively.

Trying to reduce a company’s core competency to the functions it implements is a thought-provoking exercise. We could see huge companies run by only one to two people providing a single function.

Obviously, Google Search has more than one or two people working on its search function, but it’s a billion-dollar function nonetheless. Other functions that qualify: geocoding, face recognition, machine translation. All have a simple input with high backend complexity.

A good question for a software company is: what’s your billion-dollar function? For Facebook, it’s arguably the function that allows advertisers to put ads in front of users. The defensibility comes from its database. It’s likely not a single function today, but it probably was or could be.

This concept is perhaps less useful for companies with significant offline components. Uber’s function could be to take in two (x, y) coordinates and move you between them. I guess you could say it’s a function where the state it updates is your GPS position, though it’s not as elegant.

Each successful platform has to have one “killer application.” For mobile phones, it was texting and visual voicemail. Those things pushed people over the threshold to purchase an iPhone, which was a new platform. You can build a billion-dollar company on that user base. Jack Dorsey started Twitter before the iPhone because he knew mobile was going to be big. That's why he used a 140-character limit for Twitter—it was the limit for one SMS text message. No one was going to buy an iPhone just to use Twitter, but once a person bought an iPhone, trying Twitter had no incremental cost.

From a founding and investing standpoint, you have to consider strategic questions. What kinds of platforms are there? What new platforms cure such a pain point that people get on it? Then what else can be deployed on that platform?

One of those new platforms is going to be crypto. Lots of people getting crypto wallets is good. We can deploy all kinds of new software once most people have wallets.

AR glasses are coming too. It might be Facebook's version three or version four. Apple and Google are also working on them. We might just get a bunch of models at the same time. It's like anticipating the iPhone. AR glasses are an incredibly predictable invention you can start thinking about now.

The best entrepreneurs are logical enough to think of unpopular truths and then social enough to make those truths popular.

If you are using software to go after a physical legacy industry, one option is to do it "full stack."

Replacing just one layer of an outdated legacy stack is hard. Customer acquisition and integration costs can kill you. That’s when you go full stack. You can reinvent and reintegrate multiple pieces of an old industry and make better margins. Consider "restaurant powered by technology" versus "tech for restaurants."

With Counsyl, the molecular diagnostic company I co-founded, we found we could not survive by selling bioinformatics alone. So we built a full clinical lab, software, and a national sales force. We had to build a robotic genome factory, insurance coverage, clearances, and clinical integration.

You cannot automate something until you've done it manually many times. Control all of the factors and show technology cuts costs. Remember, for many legacy companies, information technology is only a cost center. They only adopted the code. You were born in it, molded by it. A full-stack entrant into a new vertical is formidably protean. You can morph your product just by hitting keys. Tesla’s over-the-air updates for its cars are a fantastic example.

Some specific examples of full-stack startup ideas in a few verticals are law, medicine, architecture, accounting, and restaurants.

●      Full-stack law firm: Template all contracts, use law APIs as core technology, and try to hyperdeflate legal costs.

●      Full-stack clinic: Employ mobile EMR/EHR, quantified self, genomics, telemedicine, and doctors with technology skills. Accept insurance (preferably cash subscriptions).

●      Full-stack architecture: Put APIs at the core of a new construction company. Start with unmanned buildings like data centers to de-risk early versions. Work toward the ultimate goal of hitting the “Enter” key to build a building with drones and prefabrication.

●      Full-stack accounting firm: Given a bank account, auto-prepare it all, from tax to diligence to S-1 with legal sign-off.

●      Full-stack restaurant: Implement mobile orders, payments, and reservations. A/B test dynamic menus with supply chain integrations, and use robots for food preparation and delivery.

If the goal is full stack, always talk to executives in the field early on. A few words can save years of work to identify key cost centers and hard parts. You can start all of these as "just" a new clinic/restaurant/accountant/architect/law firm. Think big, start small. Prove, then scale.

Staging is key for full-stack startups. Start with the ambition to do it all, but pick a specific upgrade sequence carefully! If possible, use industry standard/off-the-shelf for a specific layer until you can get around to improving it.

Anything founded before the internet may not be able to survive the internet.