The Queue #26

Myopia over wrappers, GPT-3's moment in the sun, the stickiness of structural advantages

Short reads ⚡️

Yours truly, with some observations on ‘wrappers.’ I’ve seen a pattern in startups that target technical users wherein their product is dismissed as ‘just a wrapper around X.’ This view is severely myopic and misses enormous opportunities! Taylor Clauson


Chris Keller, writing on the persistence of returns (or lack thereof) in venture firms. He does a great job of pointing out that finding structural advantages in their managers is one of the only viable strategies for LPs wanting to achieve durable returns. Chris Keller


In his parting gift to his country, John Lewis wrote a beautiful and challenging essay to be published on the day of his funeral. Sad to lose a hero, but a moving message. New York Times


Long read 🎇

Cloudflare is fantastic, and its Workers product is fascinating. They had some big announcements this week; it’s going to be fun to watch edge compute continue to make its way into mainstream applications and open the door for new types of applications. This post is particularly worth reading because of CEO Matthew Prince’s “Hierarchy of Developer Needs”:

I'd propose instead that what developers on any platform need, from least to most important, is actually: Speed < Consistency < Cost < Ease of Use < Compliance. Call it: Matthew’s Hierarchy of Developers’ Needs. While nearly everyone talking about edge computing has focused on speed, I'd argue that consistency, cost, ease of use, and especially compliance will ultimately be far more important. In fact, I predict the real killer feature of edge computing over the next three years will have to do with the relatively unsexy but foundationally important: regulatory compliance.

I am unsure about the relative positions of the last two. A nontrivial number of developers, particularly as they are getting something new started, optimize for Ease of Use as their greatest need and then attempt to layer in Compliance down the road. I might go so far as to rename ‘Ease of Use’ in this context to ‘Speed of Implementation,’ and that’s precisely why I think it jockeys for that most important position. The rate of iteration in producing software is paramount.

Graphic I love: GPT-3’s Parameters 📊

I know, I know… no one wants to hear anything more about OpenAi’s GPT-3 language model right now. Honestly, I don’t either. The hype train has achieved escape velocity. That said, this graph showing how many parameters were used in its training (175 billion) is a marvel to review. Too frequently, we let phrases like orders of magnitude and exponential improvement slip into our vernacular when we are describing something impactful. But it’s far rarer to pause and look at something representing orders of magnitude improvement over the steady-state. GPT-3 has limitations, but coupling its flexibility with its few-shot learning feels like a refreshing example of what’s possible when you lower the barriers to build and provide a better starting point. I’m particularly bullish on some of the applications that involve making software development easier for ‘shallow’ technical users (those without a CS degree or extensive programming background).

Wikipedia rabbit hole 📖

The wrong type of snow. The British version of “you’re holding it wrong.”

Parting thought 🤔

The only thing that could make pandemic life worse would be a new Nickelback album.