Single Points of Failure
ClowdStrike, Agrobiodiversity, and physical schools: What happens when they collapse?
What’s Happening in the Market
An Indian court has forced Byju’s into insolvency, removing Byju Raveendran and the board of directors from control of the company. This is a huge fall from grace from the company valued at $22B just two years ago. The insolvency proceedings were triggered by claims from creditors, most notably–the Board of Control for Cricket, which claims Byju’s owes them $19M. We’re going to be using Byju’s as a case study for many years (poor governance, the mismatch of VC and slow-and-steady growth that education necessities, etc.) but perhaps the biggest lesson is this: You don’t mess with cricket in India.
Andrej Karpathy, co-founder of OpenAI, recently founded Eureka Labs, an education and incubation company focused on building & supporting AI native startups (shoutout to Audrey Ellis, host of AI Fishbowl, for helping us understand this topic more). The distinction between AI native companies and companies who tack on AI features is really pronounced already, and will probably determine winners as the adoption curve plateaus. Why? Because there is a very real, very tangible benefit to building a company with AI ingrained in its DNA, vs. one that simply adds on a chatbot. For all the criticism I lob at MagicSchool, the company rolls out new features incredibly quickly, and with a much leaner team than their sluggish peers. Do all of those features land? No, but as AI native startups and companies refine their products and find product market fit, the survivors will be able to deploy code much faster, they’ll be able to surface documentation immediately, and they’ll essentially have three times the employees staffed at a marginal cost (we’re already seeing this trend of companies referencing their “AI employees”).
This isn’t related to education, but has learnings for us nonetheless. CrowdStrike is used by over 50% of Fortune 500 companies, and Microsoft accounts for 85% of the federal government’s productivity software. The CrowdStrike debacle reminds me a lot of the destruction of agrobiodiversity: Since the 1900s, we’ve lost about 75% of plant genetic diversity, which has made crops so susceptible to diseases. On the OS side, we basically live in a duopoly or oligopoly, and that makes us all the more susceptible to dumb mistakes or nefarious hacks breaking everything.
One Big Idea: Single Points of Failure
Speaking of diversification, 60% of students in the US use Clever to log in. Higher education isn’t much better: Canvas has about 41% of the market share, while Blackboard has seen its market share decline from 18% to 17%.
These companies will say this consolidation is good for customers. That’s the ostensible defense of any monopoly. But as we saw with CrowdStrike and Microsoft, that isn’t necessarily true, and it is deceptive to say that it is.
The downside to single points of service: They can become single points of failure. When physical schools were closed because of the pandemic, 50 million students were sent to remote learning, and half of those students wouldn’t re-enter a school for a year. We relied on a single modality for learning, and when it failed, our educational system collapsed. Zoom school and remote school, as any teacher will tell you, was an abysmal failure (even ignoring Zoom’s own major outage in 2020). Even today, I’ve met with several teachers who work at remote public schools, and they do not function: Their students don’t attend class or sit with their cameras off and don’t engage.
The importance of diversification
We often speak about the downsides of schools/districts swimming in an overabundance of tools, which is true (especially unvetted tools, like AllHere’s Ed). And certainly, the fragmentation of tools is onerous for teachers. But it’s important, too, for schools and districts to diversify in case a company fails and because different companies offer different value propositions and areas of expertise. This is especially true given the line in the sand between AI native tools, and legacy tools which are trusted but not as future-oriented.
As in all things, nature has opportunities for learnings: The broad-breasted turkey now accounts for 99% of all turkeys in the US, but is incapable of breeding on its own. More than 7,000 apple varieties have gone extinct in the last century, and only two varieties account for 50% of the crops. Our turkeys and apples could vanish if one gene or generation fails.