Throughput trumps prioritization

One thing I've been reflecting on a lot as I read about various "theories of vaccination" is the key philosophical distinction between people and their proclivity to think temporally - to make decisions in the present which reflect a belief that the future is both predictable and mutable.

This comes up in a lot of theories of innovation: growing the pie vs slicing the pie, centralized management structures focused on "efficiency" versus artist colonies focused on "creativity," Thiel's "determinate optimism" vs "indeterminate pessimism," focuses on journey versus destination, and whether to emphasize economic growth over economic redistribution.

In this case, we see a fundamental tension between throughput and priority, between efficiency and effectiveness, between theory and data, and between the present and the future.

Like with most dialectics, I suspect there is merit on both sides, but by focusing so heavily on how to most efficiently distribute the vaccine we have now, knowing what we know now, to do the most good, we're missing the point - our goal should be to develop a decision-making framework that engages with an uncertain, but predictable, future in order to do the most good over the next 12 months.

More importantly - good prioritization with low throughput is almost always far worse than high throughput with bad prioritization.

• High-throughput systems can avoid decisions - rather than deciding between two high-priority goals, a high throughput systems can do both.
• High-throughput systems can pivot faster as priorities change - forward momentum is almost always easier to redirect than the inertia of a system at rest.
• High-throughput systems produce more information - because the system is doing more, it has the capacity to learn more quickly. It spends less time in an ivory tower, more time tinkering.
• High-throughput systems are more fun for the participants - every member of a high-throughput system matters and can individually impact the goal of doing more and moving faster. Rather than simply waiting for orders from above, when the system is focused on throughput, every participant can help on the margin.

In other words - when you don't know what to do, doing anything is better than doing nothing, and building a culture of "doing" is far more important than building a culture of "waiting." Maybe people get a few extra shots, maybe we accidentally miss a few second doses, maybe we accidentally give the vaccine to someone who didn't "need it." But vaccinating the population as quickly as possible is the goal, and we should be spending as many of our human, scientific, and financial resources to get that done faster, even if that introduces some randomness and inefficiency along the way.

Observer or Participant: Can the invisible hand sow the seeds of its own destruction?

As discussed in the previous post, companies in rapidly developing markets need to run effective idea factories - organizations of extremely talented and motivated humans arranged to bring new ideas to market. While markets often reach a "peacetime" state of equilibrium in which a few key winners dominate, the transitional "wartime" state is often marked by intense competition (e.g. the Browser Wars or Cloud Wars).

This competition happens between competing idea factories as capital gets injected into a company and intellectual property comes out the other side until one of the factories achieves some kind of "moat," at which point the war concludes and the winner has a defensible business which can produce attractive profits for a long time.

Different organizations have different inherent talent - as described by Matthew Ball, Disney and HBO regularly outperform Netflix in the efficiency of their content production. And that talent is distributed between employees and the organization. But regardless of the distribution, no matter how talented the organization may be, unlike a traditional company, IP-driven companies at war still rent huge percentages of their idea factory from their employees and depend on that talent retention to continue fighting.

So what happens as the market gets involved? Investors look into the future and pick a winner - the company which they believe will emerge as one of the dominant "peacetime" players. They do this, often, by looking at momentum and velocity: which organization is moving more quickly, taking more ground, developing more features - which idea factory is more efficient?

Then they discount that victory into the present (with a historically and unsustainably low interest rate) so that capital flows into the winner. This creates a uniquely low cost of capital for the presumed victor, all but ensuring their victory. Perception becomes reality as optimistic prognostications become the assumptions that drive present value calculations. In this model, victory itself is a self-fulfilling prophecy.

But underneath all of this lies a key question about incentives: in a war that depends on human capital, where that same human capital is compensated with stock in the enterprise, what happens when investors pay the factory workers so much for the factory's potential that the market shifts from playing the passive role of an observer into an active role, as a participant.

What happens when individual employees - not founders, or even executives, begin taking home 5-10x more than they expected? Do they continue to fight with the hunger that is necessary to win? Do they keep burning with the passion that is required for true innovation? Or do they, individually and collectively, take their foot off the gas? Does the hyper-capitalization of the idea factory perversely result in a factory that becomes less capable of victory even as it becomes more expensive?

And if so, what does that mean for capitalism - if capital can destroy value, how can we insulate small teams from capital and enable them to commit to long-term change, long-term success? What types of compensation mechanisms can we develop that are more stable, that result in more self-binding, more Ulysses Pacts where liquidity is available, but without the corrosive effect of hyper-liquidity? How could we design a compensation system in which the workers had stakes in their outcomes even in excess of their investors, where by deferring liquidity today, collective organizations could agree to provide liquidity only after victory was assured?

Today, we distribute the spoils of a future war to a present-day army, and too few people have thought through the unintended consequences as newly rich soldiers quietly leave the field in the midst of battle.

Human Capital Development: Renting an Idea Factory

The past thirty years have seen a massive shift in valuation, from an emphasis on value investing based on tangible assets towards momentum investing based on intangible assets. This, in turn, has led to broad based consternation within the investment community as conventional financial and economic models are outperformed by superficially less sophisticated approaches. Empirically, though, something is going on - it doesn't seem like our current models are working and in this scenario, new theories are required.

This essay is a reflection on the oddities of human capital, the production of ideas, and the complex relationships between employees, markets and firms in the modern economies. As the economy shifts from a focus on physical goods towards a focus on more abstract intangible goods, more and more companies are de facto becoming idea factories.

However, compared to actual factories, in which labor is rented from employees to operate high-cost machinery owned by a firm, idea factories rely on high-cost labor carefully arranged in order to produce finished products (e.g. software and content). The output of this idea factory is "owned" by the firm, but the firm itself is also owned by humans, many of whom may be key employees in the firm. This fact - that individual humans play many different roles in the context of a modern idea factory (founder / factory manager / idea producer / owner / investor) - is wildly under-discussed.

Too often, valuation models focus too much on a given firm's stock of intellectual property with too little emphasis on the half-life (depreciation) of that IP - here, we attempt to reflect on the flow of intellectual property; the capacity of a firm to create novel IP, which is a function of the organization's ability to recruit, retain, and grow talent (atomic human capital) as well as the organization's ability to organize that talent productively (emergent human capital) - in other words, it's stock of human capital.

Unlike other forms of capital, atomic human capital is owned by the individual and rented by the corporation. Its productivity is wildly variable, hard to measure, and subject to power-law dynamics: the productivity of a single human within a complex idea factory varies tremendously. More importantly, though, individual humans have non-linear atomic growth. This growth, over time, increases the capacity of a given human to produce value. Perversely, though, this growth also increases the bargaining power of that human to lobby for higher wages. As such, the return on investment in human capital development is much different and complex than the return on investment in traditional capital - labor, not capital, should be able to extract some large percentage of that created value, and firms end up competing with one another for access - in some sense, the value of human capital investments accrues to the individual, not necessarily to the firm.

But aside from Substacks / Newsletters, there aren't that many great projects where the value is around the sum of the parts - most great idea factories require the management of brilliant individuals together, not to make them replaceable, as in a traditional factory, but to make them irreplaceable as in a great championship winning team.

On the acknowledgement of messages

 Recently, I've been hit with a virus - I'm in a twenty-person text thread which was initiated, for purposes unknown, by a some kind of spam genius organization. Frankly, I love it - I'm proud of the spammers: they've created a perpetual spam machine that has no clear end in sight. Every day or so, sometimes multiple times a day, an exchange like this occurs:

A frustrated member of our self-created prison lashes out, demanding that we all stop texting, and then a helpful good Samaritan replies, explaining that texts sent to the group do, in fact, get sent to the entire group. Of course, the ultimate irony is that by chastening the first texter, goodie two-shoes is the prime mover of the next exchange, ad nauseam, ad infinitum.

Hats off to the spammers, this is top notch spam!

Spotify Podcasts

I was recently listening to the Invest Like the Best interview with Daniel Elk, the CEO of Spotify, and was stuck by the dissonance between the theoretical articulation of the product vision and direction and my lived experience with the product, specifically as it relates to podcasts.

While Daniel coherently walked through the Clayton Christensen "job to be done" framework (also popular with the instagram founders in an earlier episode), I was shocked about how poorly the Spotify product team seems to have adopted this lesson in their product development strategy.

Specifically, while there's an overlap between the times I'm hiring spotify to find and curate music and when I'm hiring spotify to find and curate podcasts, after that initial decision, it feels like the team retrofitted the existing spotify app and user experience without critically thinking through how users actually want to interact with a podcast-specific experience.

In particular, I was shocked to see that they tried to smash podcasts directly into the existing Spotify app rather than having a podcast-oriented application geared specifically towards podcast users. As an episodic medium with repeat consumption and curation, podcasts are just very different than traditional music. Additionally, once I've decided to listen to a podcast, the likelihood of me wanting to listen to music is extremely low - podcast-specific search, prioritization and discovery is much worse when mixed up with musical mediums (https://community.spotify.com/t5/Live-Ideas/Podcasts-Split-out-podcasts-into-separate-app/idi-p/4938692).

This is the approach Facebook has taken with Messenger, and Gmail took with Inbox; as a user, I get a tailored experience without being distracted but content unrelated to my general user-intent.

As a final aside, while I understand the complexity of the b2b partnership side, as a consumer, if spotify is trying to move into podcasts, the lack of audible / deep integration with books-on-tape feels like another way to control the user experience.

Data / Model-driven writing tools

For ~5 years, I've continuously returned to this blog post:

http://worrydream.com/ClimateChange/

I think that there's something so subtly powerful about being able to take data / logic and surround it with rhetoric and prose. Most tools separate these out between Word / Excel / Powerpoint, but I think that successes of things like dashboarding tools and other interesting kinds of data-driven visualization tools make me really excited about the kinds of tools we should be focused on building as technologists.

Incident Response

Part 1 and Part 2 and are here – main takeaways below, but recommend reading the raw papers for more detail. Overall, I think that we do a pretty good job of this, but especially as we move towards more Operational Responsibilty, Baseline, and PRX working closely together, worth reviewing for some ideas of where things fail in a distributed incident response environment. Some of my notes below:

• It was really cool to see the rigor with which they analyzed the response rather than the incident itself – there are a lot of papers about “bad bugs” but less about “collective multi-engineer debugging.”
• Organizational complexity often mirrors system complexity (Conway’s Law) conflating incident response.
• How important decentralized agency / autonomy is in incident response – “plans / actions” are often poorly decomped and require high-trust operators to go off script.
• Reminds me a lot of patterns in immunology with antibodies / tcells / general sympathetic nervous response, as well as trauma triage principles in emergency medicine.
• Willingness to achieve “immune response” and bring in expertise rather that fearing the repercussions of over-escalating.

Thematically, I think that the most important takeaway for me is that preventative design is not enough: while it’s really important to design our software and try to prevent catastrophic failure, it’s probably equally valuable to design our human / organizational response systems in a way that quickly resolves these catastrophes.