Shut it down #001

I have a habit of keeping quite a lot of browser tabs open. I know I am not the only one.

On my laptop, my tabs tend to be related to what I am currently working on. I can deal with that. On my phone they tend to be a mixture of things I hope to read in the future, and some things which I have read already and thought were good, but which I didn’t know what to do with. That is a bit more bothersome, because it is accretive over time, so now I have loads.

The good news is I now have a place to find closure, right here on the internet. Thank you for witnessing my decluttering, where I Shut It Down to browser-tab-zero.

12 Things I learned from Chris Dixon about Startups

This has been in my tabs for a while; it’s an article I have re-read a couple of times. The toy / hobby / weekend metaphor for good startup ideas is pretty well-known by now, but it is always worth revisiting. I like the idea of a good startup idea being predicated on a secret, and of moving from uncertainty to value creation. I feel like the transition of technology from bits to atoms is still pretty nascent, which is exciting. Meanwhile the market-size narrative challenge for startups is evergreen. The importance of getting rejected often is hard to internalize, but I see the value (in setting ambitious goals). While the parting message for entrepreneurs (“get ready to feel sick to your stomach for the next five years”) can certainly be considered fair warning.

The Psychology of Human Misjudgement by Charles T. Munger

A long one but a good one, a chimera of several talks Munger made in the early 90s, where he highlights biases which lead to bad decision-making. Far from an academic study (as he himself acknowledges somewhat gleefully), it nonetheless (or perhaps as a result?) is very thought-provoking.

He starts with the always underappreciated power of incentives (“I think I’ve been in the top five percent of my age cohorts almost all my adult life in understanding the power of incentives, and yet I’ve always underestimated that power”), before making his way via the biases of loving and hating, through man’s dislike of inconsistency and doubt, touching on the dangers of optimism, and loss aversion (“Deprival Superreaction Tendency”), as well as ways in which “leaders … display followership akin to that of teenagers” due to Social Proof Tendency. The Twaddle tendency has an entertaining bee-based comparison, but it all comes together with the “Lollapalooza Tendency”, which considers the potential dangers when many different tendencies are brought to bear at once. An irreverent waltz, and worth a read on a long train ride or equivalent.

Do things that don’t scale by Paul Graham

Something of a classic, and certainly a phrase that is now part of Silicon Valley lore. People were presumably talking about Delight prior to 2013, but I like how it is described here (“You should take extraordinary measures not just to acquire users, but also to make them happy… existing conventions are not the upper bound on user experience”). Pithy warnings too, about relying on Big launches and partnerships

It’s interesting to see companies like Superhuman push the “do things that don’t scale” approach even further in 2019 (1 to 1 onboarding for a consumer internet product). Paul hasn’t written an essay since 2017, which is a shame (though he is alive and well on Twitter).


Not an article, but a collection of intro-to-AI notebooks. It recommends using Google Colab, which I had last used when it was very much an internal-only Google tool, so cool to see it in the wild. A few topics still waiting to be covered, but a nice starting point nonetheless.

The art of decision making

Much is made of the importance of good decision-making. In this piece Joshua Rothman breaks down the things that make that difficult, with his decision to become a parent as a case-study.

Decisions are often more gradual than they are discrete (“it’s a momentous choice, but I can’t pinpoint the making of it in space or time”), are often are limited by past choices (so-called “bounded rationality”), and can’t be understood on a single scale. There is certainly scope to improve decision-making (“scenario planning … seeking out diverse perspectives on the choice, challenging your assumptions, making an explicit effort to map the variables”), as we “ask ourselves what we value, then seek to maximize that value”. But it becomes more difficult once you realise that what we value might shift over time in a way that isn’t always predictable (“Why should today’s values determine tomorrow’s?”).

This is probably why people “are in fact more casual and cavalier in the way they handle their big decisions than in the way they handle their ordinary decisions”, in the words of Edna Ullmann-Margalit. Certainly food for thought as you think about your own decisions and those you observe others making, both before and after the fact.

MakerDAO Sponsor Bounties for ETHDenver

A bit rogue I suppose, but I am always interested by the sponsor-suggestions for hackathon-type projects, as it tends to capture what is top-of-mind in a given industry or companies. It’s been cool to see the evolution and growth of the MakerDAO stablecoin project, so I was interested to see their areas of focus – in this case wallets, lending and zero-knowledge proofs.

Heuristics to Generate Startup Ideas

A well trodden path, but some good thoughts in this one. I particularly like “Tools inside a big company” and “Revisit ideas that were too early”. “Understand how teens communicate” is sadly probably beyond me at this point (I am not sure I was that in touch when I was a teenager myself…).

What is Amazon by Zack Kanter

A tour de force. On the evolution of the retailer algorithm, from the “bounded” physical world of Walmart to the “unbounded” digital world of Amazon, and the importance of the third party-marketplace in the new world (vs the old world’s vendor selection model). On how Amazon’s need to grow at internet scale made it necessary for them to become a “platform; that is, an aggregation of resources made available through a series of interfaces“. On the edict from Jeff Bezos, for all teams to build as if for external customers, to enable the business to transform into, and scale as a platform (“Platforms became part of the algorithm”). A dissection of Ads, designated a misstep (but one that may be “impossibly addictive”, given the revenue). And a riff on the risks of its “Wild West for sellers”. I definitely share Zack’s parting sentiment: “I remain fascinated to see what will happen next.”

Open Jobs: Making labour markets smarter and more empowering for jobseekers

I have been thinking quite a lot recently about the future of work and closing supply-demand gaps in labour markets. Nesta have a bunch of interesting articles on the topic, and this one is right on:

“although 75.3 per cent of adults in the UK are in jobs, this headline figure masks some deep inefficiencies and problems of stagnant pay, social mobility and productivity and major failures in the transition to work”

The sell is for their Open Jobs project, which aims to address issues with data, policy and pilots.

How not to fail by Jessica Livingston

Taken from a talk given to the Female Founders Conference, it’s all good stuff. Make something people want, know if you are default alive, keep expenses low, and more. YC content is pretty pervasive these days in startup-article-land, but that’s probably because they know what they are doing.

How the UK lost Brexit

Fascinating dissection of the negotiations between the UK and the EU in the aftermath of the EU referendum, and how the UK were outmanoeuvred time and time again. Bloody red lines, cherry-picking and Barnier.

Holloway About Page

I don’t know much about them, but I like what they are doing (“We publish fully digital Guides to high-stakes, complex subjects”). One to keep an eye on.

The Rawlsian Diagnosis of Donald Trump by Samuel Scheffler

On Rawls and the importance of reciprocity in liberal societies: a proposal that American society has in recent decades seen “the failure to achieve—or even to strive seriously to achieve—an ideal of reciprocity”. A suggestion that this failure has contributed to the rise of Trump in a manner predicted by rather than contradicting Rawlsian liberalism, as “U.S. institutions have come closer to maximizing the position of the best-off group than to maximizing the position of the worst-off group”. Interesting, but definitely on the academic end of the spectrum.

Invisible Asymptotes by Eugene Wei

I first came across Eugene’s Status-as-a-Service article in February, which I also recommend. This piece introduces Invisible Asymptotes, where a single factor limits growth beyond a certain point unless addressed appropriately (if it can be), starting with Amazon’s first: shipping cost (which people hate to “to literally an irrational degree”) . A whistle-stop tour through tech Royalty’s varied and different invisible asymptotes, some more practical words of advice, plus some reflections on the asymptotes in all of our lives (“In my experience, the most successful people I know are much more conscious of their own personal asymptotes at a much earlier age than others”). An interesting way to think about potential and future growth in all walks.

And that is that. My phone feels lighter already, thanks for closing down my tabs with me. Until next time.

Life Capital Conference

An Income Share Agreement (ISA*) is a personal financing instrument where repayments are determined as a percentage of someone’s income for a specified period, an equity-type alternative to personal debt for individuals. It’s something of a hot topic, having been adopted by some high profile coding bootcamps, and a lot of the interested folks came together on Wednesday for the Life Capital Conference, hosted by Slow Ventures, Village Global, The Information, and Cooley.

It was a packed agenda in a packed auditorium. It is still a nascent industry, so the discussions were mostly forward looking, and were variously practical (how do we make this work right now) and expansive (what is the possible future for this thing). It highlighted some of the progress already being made, as well as some of the challenges.

Equity for individuals?

While “equity for individuals” is a catchy tagline, it doesn’t necessarily describe how ISAs are actually working at the moment.

From a practical standpoint, ISAs today predominantly provide funding for education in return for a fixed percentage of personal income (revenue) for a set period of time, where the income is above a certain annualised threshold, and where the total amount repayable is taxed. This is different to equity investment. I will leave aside the troublesome ownership implications of “equity” when applied to an individual.

Benefits of ISAs

The benefits of ISAs were well understood and articulated throughout the panel conversations. One of the core premises is incentive alignment: everyone involved wants the ISA recipients to do well. A significant benefit for the individual is risk reduction: if things don’t work out, they have to pay less back (or nothing at all). There are also opportunities for improved accessibility, affordability and outcome transparency not offered by conventional borrowing.

Happy even if it goes well

As a student attending a coding bootcamp, downside protection is a significant advantage, when compared to a conventional loan. Aanand Radia of University Ventures described an ISA as “a loan with an insurance wrapper”, which resonated, but Barry Cynamon’s “insurance where you pay the premiums afterwards” captured the behavioural challenge of ISAs, where if you have a good outcome (i.e. you earn a good salary), you have to pay more, after the fact. One of the core challenges with ISAs is this cross-subsidising across a cohort of students, where the more successful students pay more (often a lot!)

Ensuring people are happy “even if it goes well [and they have to pay a lot more]” is one of the key communication challenges facing ISA providers. In this case the successful students are paying for the downside protection they didn’t end up using, but it was pointed out by Ashu Desai of Make School that this was not an easy thing to put a value on. Multiple panelists highlighted the importance of students being well-informed and “bought in” to the collective nature of ISAs because of this dynamic, understanding that the ISAs work as a kind of income tax (Sam Lessin, who hosted the event, even noted that as taxpayers, we are already participating in an Income Share Agreement by another name).

This has lead to the natural concern that ISAs will suffer adverse selection, where students with a lower chance of success are more likely to take an ISA. Purdue University’s Mary Claire Cartwright indicated that their research showed that there was no adverse selection, having compared their ISA students and those taking regular student loans. I understand that this would be reassuring, but I find it quite surprising that this would be the case: companies that take equity financing are different (and often more risky, with more volatile outcomes) than companies taking debt financing. As an individual with predictable outcomes, a loan is probably a better option, while for someone with a bit more volatility, an ISA is a much more attractive option. If this is not being reflected by those taking ISAs at Purdue, it says something about human psychology, or a lack of appropriate information. Perhaps both.


ISAs are a significant enabler for accessibility, as they can be offered to people who otherwise might not be able to access finance (and might therefore be excluded if the ISA is for an educational course). This is an undeniably powerful benefit, particularly in the case of coding schools and the tech industry, which has an acknowledged problem with diversity and access.

Though a word of caution – while some credit checks used for conventional loans are exclusive (based on historic income and credit worthiness), affordability decisions on the basis of significant existing debt are still likely relevant for individuals taking on an ISA.

Incentive alignment

The incentive-aligning power of ISAs was much discussed throughout the day. It is definitely very powerful for vocational schools’ payday to be aligned to the paydays of their students. However I worry that it might drive a one-dimensional money-focused approach to student success. With a conventional loan, the lender’s concern is that the borrower has a certain threshold of income where repayment is affordable. Anything beyond that is just upside for the borrower. While for an ISA, higher and higher salaries are even more rewarding for the ISA provider. I wonder whether that might motivate schools with ISAs to focus purely on placing into positions with higher salaries, rather than thinking about other things that contribute to happiness and fulfilment in a job (particularly given research that indicates that incremental income above a certain threshold doesn’t contribute any more to happiness).

We do definitely understand the power of the outcome-based nature of ISAs. In the work we are doing at Vested we are excited by the possibility of an outcome-based loan, providing downside mitigation without requiring significantly more from the highest performers.

Headline risk and regulation

One of the prevalent concerns amongst participants was the risk of one or several bad actors taking advantage of students with usurious ISAs, and the associated headline (and existential?) risk for the industry. This likely drove the excitement over potential ISA regulation, a strange occurrence at a tech conference (as Austen noted on Twitter). Local and national legislators were represented on panels, in Adam Boman of California State and former congressman Luke Messer, who both indicated positive progress on that front. As a Brit, I am familiar with the quasi-ISA used for university financing by the government, and I definitely see the potential for state-enabled post-secondary ISA-funded education. I wonder whether legislature in the States will make moves in that direction, or remain focused on privately funded ISAs.

Sam Lessin was one voice in favour of a considered approach towards regulation, thinking carefully about any legislation that is too prescriptive based on ISAs as they currently exist, for example mandating caps or minimums or terms. “Let’s not make short-term trades that might handicap us in the long-term”.

Bundling and unbundling

The ascent of Lambda School has done much for the profile of ISAs. Austen Allred’s discussion with Will Quist highlighted the interesting way in which the Lambda School model is both an example of bundling and unbundling. Austen acknowledged that traditional universities layer in things that Lambda does not (the housing, the extra-curriculars), while Lambda School just focuses on the “I want a job” part of the university experience. The best choice for the student depends “which bundle is the student choosing”. In this sense, Lambda School is an example of unbundling education, but at the same time, Austen described their vertically integrated “underwrite, teach, place” model as key to their success – bundling up three pretty distinct competencies.

Must be the money

Given the (relatively) embryonic stage of the ISA market, it is not surprising that source of investor capital was a common topic, with schools from Holberton to Lambda talking about funding ISAs with their own equity. Tonio DeSorrento of Vemo described the outside capital evolution that almost all ISA businesses would go through, from philanthropy to novelty to impact to hedge funds to municipal money, based on the evolution of rate of return and track record. University Ventures are currently the largest provider of ISA financing in the world, but as they themselves noted, for the success of the industry that couldn’t remain the case.

Future for ISAs

A lot of participants were clear that ISAs were not a replacement for student loans, and both would likely co-exist in the future. There was a prediction that in 5-10 years, ISAs would be the same size as private loans, in dollar terms. Richard Lee of Leif predicted $500M in ISAs originated in five years time. Austen suggested that, however guaranteed their outcomes became, they would never switch to debt (“Debt is fundamentally opposed to the lambda model”), and will instead just work to drive the cap down.

At least offering an ISA was expected to become table stakes, in terms of offering affordable post-secondary education. “300 universities will offer ISAs in five years time” per Andrew Platt of Vemo. “If your vocational school doesn’t have an ISA, you are dead in five years” per Austen. We are certainly currently seeing a chain reaction in coding schools that is a move in that direction.

It will be interesting to see if that continues to other vocational skills – sadly Austen and Will’s discussion didn’t have time to fully cover Lambda’s ambitions beyond bootcamps, which are substantial. Austen did note that the model would likely look similar but slightly different in nursing, for example.

There was general agreement that growth of ISAs in the future could contribute to much improved outcome data for educational institutions, to enable applicants to make better informed decisions about their education.

The focus of the day was very much on educational financing, though the “Imagining the future” session took a more expansive view, with Dani Grant, Sam Lessin and Erik Torenberg. Were current models “too practical” – why have a cap on repayment? Why can’t individual teachers be motivated and remunerated in this way? What if individuals in a cohort or community (Stanford grads, say) cross-invested in one another to normalise our outcomes? This is apparently something which apparently already happens in high-stakes poker games. Dani Grant of USV wondered wether we would we be kinder to each other, if we were cross-invested. But there was general acknowledgement that we maybe were not ready for the “out there” version quite yet.

So what, for the future of ISAs? Tonio DeSorrento was optimistic, anticipating “a Cambrian explosion of ISA businesses”, though he wryly noted that the Cambrian explosion took 25M years. Those in the room seemed hopeful it might all happen a bit sooner than that.

* British readers will be confused by this acronym, which in the UK refers to tax-free Individual Savings Accounts


Amir Efrati of The Information has an interesting piece on Waymo robotaxi customer satisfaction (or lack thereof). The article highlights existing services (Uber, Lyft) as a key benchmark for automated vehicles, and suggests that the comparison is currently unfavourable:

“Waymo rides received five-star ratings for 62.6% of 2,498 rides it completed in the first 10 weeks of the year… By comparison, as of 2017, users of Lyft’s ride-hailing service gave a five-star rating in about 94% of rides”.

I am not sure this comparison is a perfect one. When rating my Uber rides, five stars is my default. That is sadly not because I anticipate a blissful ride every time I ride with Uber, but because I am aware of the importance of a driver’s average score to their continued employment, and maybe livelihood, thanks to Uber’s policies and algorithms. I am therefore willing to put up with a swerve or two (or three), and still select a five star rating. I know I am rating the driver, not just the ride.

I am sure I am not alone in this. I expect that a lot of the riders in Waymo’s automated vehicles had a lower “default” score in their minds, once you took the humanity out of the car.

That being said, the article also highlights some specific customer feedback. My favourite: Jan. 16 “Mr toad’s wild ride! Go left then right then right then left then left the to far then not far enough……”