Idea Compass #1: Overview
More resources can be mobilized behind bad ideas than ever before
Ideas matter
There’s a common piece of startup advice that ideas aren’t very important, and only execution matters in the end. I think this was never really true, and is even less true now. Good ideas are necessary but not sufficient to succeed.
First off, it’s possible to choose ideas that are technically infeasible. No amount of good execution could have saved Theranos, for reasons I’ll explore later. Second, it’s possible to choose ideas that are made obsolete by broader trends. No amount of business model innovation could have saved Blockbuster. It’s also possible to work on ideas that don’t solve a real problem. No amount of good execution could make NFTs or tokenized carbon credits useful. All of these kinds of bad ideas are in a “basin of futility” that cannot be escaped by execution, short of a hard pivot.
Leverage makes ideas more valuable
I would argue that ideas matter more than ever, because we live in a time of abundant leverage. A single person can scale software to millions of users. To paraphrase Naval Ravikant, all of us can summon a small army of robots in data centers to work for us while we sleep. Content can reach a wider audience than ever before. You can order just about anything and get it in a couple days. We can now have a conversation with a compressed version of all human knowledge. AI tools bring everyone’s ability to write, code, design, think, etc up to at least the median human level.
Execution is becoming cheaper, and ideas are becoming more valuable.
Bad ideas are easier to work on than ever before
Leverage is a double-edge sword: good ideas are easier to execute on, but so are bad ones! More resources can be mobilized behind bad ideas than ever before. Any MBA can now hire Devin AI for $8/hour to build an AI-copilot-for-X, while an ensemble of AIs and outsourced gig workers do the website, marketing, content, etc.
On a more serious note, if we do live in the most important century when many important problems will be solved, the opportunity costs of working on bad ideas is significant. You could spend 15-20 years building the next Segway while major advances in biology, healthcare, clean energy, food, transportation etc, pass you by.
If you’re optimistic about the future, it’s worth spending more of your time scrutinizing ideas.
The Idea Compass
This series of blog posts explores the common ingredients of bad ideas. To make sense of them, I’ve organized ideas into what I’m calling the Idea Compass (see below).
In the chart, I've plotted some ideas along two axes:
x-axis: Innovation. At the time, how significant of an innovation was this idea? Is it an incremental improvement, a paradigm shift, or somewhere in between?
y-axis: Value. How much value does this idea create for the world? High value could mean a marginal benefit to lots of people, or a huge benefit to a small number of people.
Top-left: “Faster Horses”
If you consume the same kinds of podcasts and blogs that I do, you'll probably recognize this as a reference to the apocryphal quote attributed to Henry Ford:
“If I had asked people what they wanted, they would have said faster horses.”
For any given problem in the world (e.g transportation), there are usually obvious, incremental solutions (e.g faster horses) and paradigm-shifting solutions (e.g internal combustion engines). People know what their problems are (or maybe they only think they do!), but they don’t necessarily know the best way to solve them.
Faster horses are not necessarily bad ideas. I imagine that a faster horse would have been very useful to a lot of people at the time. The ideas in this quadrant are incremental, but create lots of value: BlackBerry phones, Netflix’s DVD rental service, hybrid cars, and traditional AI algorithms (e.g SVMs).
Execution matters a lot in this quadrant, since the ideas are not novel. At the limit, you have something like a grocery delivery app, where there are 10+ startups competing over the same set of features.
Bottom-left: “Gimmicks”
Gimmicks neither innovate nor create value for people. They often are the status quo wrapped in better marketing or the appearance of innovation. For example, AG1 is a ~$100/month greens powder marketed as a nutritional panacea. Aspiration is a green bank investing more in airlines than renewables. Many Web3 ideas (like NFTs) fall into this quadrant as well.
Bottom-right: “Loonshots”
History is full of innovations that were supposed to be revolutionary, but didn’t pan out. Segways were supposed to revolutionize cities, the Metaverse was supposed to revolutionize work and social life, Makani was supposed to revolutionize renewable energy, and Ginkgo Bioworks was supposed to revolutionize bioengineering. It’s easy to make bold predictions, but hard to be bold, correct, and profitable.
Top-right: “Breakthroughs”
These ideas are technological paradigm shifts that bring significant benefits to the world. A very recent example is mRNA vaccines. mRNA vaccines have saved ~20M lives so far, can be designed in weeks or months (instead of decades) and leverage your own cells to manufacture the immune target. A historical example is the Haber-Bosch process, which allowed us to take nitrogen from the air and turn it into fertilizer. It was a breakthrough that has allowed us to feed billions of additional people.
Using the idea compass
Given a new idea, is it possible to predict which of the four quadrants it will end up in?
To disappoint you now rather than later, my answer is “sometimes.”
I don’t think it’s possible to consistently predict breakthroughs (top-right). If there was, investors would just buy these needles instead of the proverbial haystack. The bottom-right quadrant is a proof by contradiction: it contains a graveyard of paradigm-shifting ideas (“Loonshots”) that were backed by a lot of smart people. Once again, it’s easy to be bold, but very hard to be bold and correct.
Outside of the special top-right quadrant, however, I think it is possible to make predictions and catch common failure modes. We can’t consistently pick winning ideas, but we can rule out some bad ideas that aren’t innovative or impactful. That is the purpose of this essay.
First, we can detect incrementalism masquerading as innovation, which almost always ends up in the bottom-left with the other “gimmicks”. Once you peel away a few layers of marketing, these ideas and companies are plainly unexciting.
Second, we can separate the faster horses from the breakthroughs, though they exist on a spectrum. Some cases, like electric cars, are debatable; are EVs an improvement on the concept of a car or a paradigm shift toward an electrified world? The point is not to start an endless philosophical debate, but to just be honest about the level of innovation involved.
Third, we can also identify novel technologies being applied to the wrong problems, which are often called “solutions in search of problems” (SISPs). SISPs are an interesting case where the underlying technology might be a breakthrough, but the product isn’t useful. SISPs can end up in the bottom-left or bottom-right, but never the top-right.
Fourth, I think there were warning signals for some, but not all, of the Loonshot ideas in the bottom-right. One red flag is betting against powerful external trends that seem unlikely to yield. Another is having a nebulous argument for why people should adopt a new way of doing things. For some ideas, the math simply doesn’t check out.
In the following series of posts, I’ll provide some salient examples of each quadrant and try to extract some useful patterns. And to have some skin in the game, I’ve added some of my own predictions in the final post.