OpenAI, Parental support for flexible K12 funding, new college grads struggling to find work, and a new report on the future of work
What a weekend! A few announcements before we check in on OpenAI:
Boston Happy Hour Tuesday, 11/22: Tomorrow (or today, depending on when you are reading this)! Hit reply to this email for details!
End-of-Year Survey: Thank you to all the folks who have taken the time to fill out the State-of-EdTech survey. The survey will remain open until the end of November - I’d love to hear your thoughts on 2023 and what you think is in store for 2024! Survey here.
End of Year Posting Schedule: There will be no Funding + M&A Update or Weekend Reading this week due to the Thanksgiving holiday in the US. There will be posts on 12/3 and 12/10 before I pivot to end-of-year posts and a holiday break in the second half of December.
With that, on to the update!
The business news of the weekend, across industries, is that OpenAI fired their CEO, Sam Altman for…reasons. There are lots of rumors swirling, but it is not my place to speculate on what happened.
There are two lessons that are worth reflecting on for EdTech folks though, as they can (and should) be applied to almost any organization: 1) aligned governance and 2) platform risk.
1. Aligned governance: OpenAI started life as a non-profit funded by a number of rich tech people. In 2018, there was a power struggle for control of the company, which led to Elon Musk leaving the company’s board and Sam Altman taking over as CEO. From that point forward it became clear that the company would operate with a much more commercial bent; less than a year later the company transitioned to a convoluted “capped profit” structure and Microsoft invested $1B in the entity (Microsoft has since invested north of $10B in cash and processing capacity).
There is nothing inherently wrong with this sequence of events. Board battles happen. Non-profits become for-profits all the time - occasionally for questionable reasons, but usually for fairly reasonable ones.
The strategic error here is that OpenAI’s then-new CEO, Sam Altman, never fully re-aligned the company’s board with the company’s new direction. Altman wanted to accelerate the pace of progress towards artificial general intelligence (AGI), while a (slim) majority of the board feared the societal implications of this progress.
Again, this is not an argument for or against acceleration or deceleration towards AGI - that is beyond my scope. It is an argument that company governance - particularly at the board level - matters. Lack of alignment, or plain old bad governance, leaves a company exposed to enormous destruction of value.
This argument extends far beyond OpenAI. The dominant company-building narrative of the 2010s was placing as much decision-making power in the hands of the CEO as possible. This (probably) allowed entrepreneurs to take bigger risks, and build bigger companies than they might have otherwise. It is also what enabled WeWork founder Adam Neumann to buy a wave pool company for his office subleasing business; no person or collection of people could tell him not to.
OpenAI’s board did, technically, have the power to tell Altman to slow his roll. Where this lesson gets complicated is that having the power to do something is different than actually doing it. The social contract between Altman, the board, the employees, and OpenAI’s investors does not appear to be strong enough to weather the actions of the stakeholders involved. The result is, at least, a weekend of chaos, and probably much more substantial damage to the organization.
My hope is that organizational leaders will be taking the last few weeks of this year to consider their own teams’ alignment, using this weekend’s doomscrolling as the impetus to take action where appropriate.
2. Platform risk: On Friday, I was workshopping a headline concept along the lines of “That sound you hear is the footsteps of 1,000 companies pivoting from Tutoring to Web3.”
It was a little too dark for me to have actually run with, but I think it helps quickly illustrate this point: Building on someone else’s infrastructure allows you to move quickly, but puts you at the mercy of their decisions.
For the past 3 years, building on top of OpenAI infrastructure was an amazing decision. The company shipped product faster and better than anyone else on the market. So much so that Microsoft - a company worth more than a trillion dollars - started laying off internal AI experts in favor of a deeper relationship with OpenAI. Thousands of other companies made similar decisions, including virtually every big EdTech company.
Leaders at all of these companies spent this weekend with no control over whether their AI products would work today, considering alternative providers, and wondering whether they should invest in their own models.
I have argued that there are very few use cases that demand a fully customized AI model. It is tempting to walk back that stance in light of this weekend’s events, but I think it is still true.
The costs - upwards of $10M per hire and millions more to train and maintain a model - and risks of building a viable proprietary model are simply out of reach for most companies.1 Like cloud services, payments, and legal documents, the risks of trying to build something yourself almost always outweigh the rewards.
Funding + M&A
Bloom raises $10M / UK, Career Coaching / Octopus Ventures, MMC Ventures
Retorio raises €9M / Germany, Career Coaching / Square One, Porsche Ventures, Storm Ventures, Basinghall Partners, Passion Capital, Sofia Angels Ventures
xUnlocked raises £5M / UK, Content Provider / BPP Education
EarnBetter raises $4.5M / US, Hiring Platform / Andressen Horowitz, Abstract Ventures
BrightBreaks raises CA $2.5M ($1.8M) / Canada, Employer Benefits / Build Ventures, Concrete Ventures, Invest Nova Scotia
Xueyan Wang joins Outschool as COO / via PRNewswire
Physicswallah lays off 120 / via moneycontrol
Survey of US parents shows broad bipartisan support for flexible funding for K12 education. I think this is going to be one of the most important topics of the 2024 US election cycle. / via Odyssey
Education Department eliminates cap on Innovative Assessment Demonstration Authority, which will allow states to pilot new assessment approaches. While it reads like a small change, it feels like it could have a big impact if AI helps reduce the cost of producing efficacious assessments. (A big “if”, but not an unreasonable bet to make.) / via K12 Dive
Retrospective on 20 years of school choice vouchers in DC. / via K12 Dive
Stride Inc. guidance continues to be on record growth. Stride is one of few public barometers we have for alternative K12 pathways in the US. Obviously, this guidance should be taken with a grain of salt. But it says something that the company continues to project substantial growth. / via businesswire
New college grads struggling to find employment. / via Washington Post
It says something that the University System of Maryland is “concerned about” and lobbying against a proposed change to federal overtime rules. If the financial health of your $6B+ annual revenue business is dependent on an army of low-paid salaried employees working more than 40 hours a week, it might be time to rethink the business model! / via Inside Higher Ed
College students took 16% fewer language classes (excluding English) from 2016 to 2021. Language learning seems to be one of the early strengths of AI, and language learning companies (both startups and big companies) have gotten a lot of hype this year. Duolingo sizes the market for English language learning at 2 billion, which is big enough for a lot of companies to occupy space. But it still feels notable that learning other languages has fallen off at the college level. / via Axios
Following last week’s earnings report, 2U’s CEO, Chip Paucek, stepped down. Several news outlets tied the leadership transition to the company’s announcement that they were cutting ties with USC. The more plausible explanation for the transition is that the company is going to have to pull off some serious financial maneuvering to stay afloat and the company’s stakeholders felt that their CFO was the best fit for that job. / via Higher Ed Dive, USA Today, and PhilonEdTech
Reach Capital’s Reimagine Work report on how the Workforce segment is evolving. / via Reach Capital
This email, ETCH Weekend Reading, is ETCH’s free newsletter providing links to the week’s EdTech Funding, M&A, People moves, and a curated list of Links to relevant industry news. If you enjoyed this edition, I hope you will subscribe and/or forward to your friends!
Meme of the Week
I’m cherry-picking a little here to make a point. You can find AI talent for less than this figure, especially if the scope of what you’re building is not a global-scale LLM. You can train a model for $600. However, building and supporting an in-house AI function is a serious undertaking. Very few companies have the team and/or credibility to pull it off relative to starting with something off the shelf.