Gianfranco's Best of June 2024 Reading List

Welcome to the June 2024 edition of my monthly reading list.

This month, I've curated my favorite essays across various categories: Artificial Intelligence, Society and Technology, Finance and Economics, Early Stage Investing, Writing, Life and Meaning, and Web3.

If you only have a few minutes, these three posts were my favorite, and are included in the list below:

𝘛𝘩𝘦 𝘚𝘰𝘯𝘺 𝘞𝘢𝘭𝘬𝘮𝘢𝘯 (𝘊𝘰𝘮𝘮𝘰𝘯𝘤𝘰𝘨)

• The Sony Walkman redefined personal audio, challenging the notion that music was inherently a social experience. Sony's insight to create an individual listening experience proved revolutionary, even as they hedged their bets with two audio jacks in the first model. This shift from shared to personal audio consumption prefigured the rise of personalized technology and content that became mainstream in the early 2000s.

𝘞𝘩𝘺 𝘐𝘵'𝘴 𝘏𝘢𝘳𝘥 𝘧𝘰𝘳 𝘚𝘮𝘢𝘭𝘭 𝘝𝘦𝘯𝘵𝘶𝘳𝘦 𝘍𝘶𝘯𝘥𝘴 𝘵𝘰 "𝘗𝘭𝘢𝘺 𝘵𝘩𝘦 𝘎𝘢𝘮𝘦 𝘰𝘯 𝘵𝘩𝘦 𝘍𝘪𝘦𝘭𝘥" 𝘪𝘯 𝘈𝘐 𝘐𝘯𝘷𝘦𝘴𝘵𝘪𝘯𝘨 (𝘊𝘩𝘶𝘥𝘴𝘰𝘯)

• Charles Hudson at Precursor Ventures highlights the challenge for small venture funds navigating the AI investment landscape, where experienced founders bypass traditional seed rounds for larger, higher-valued rounds from multi-stage funds. This trend challenges portfolio construction for smaller funds, potentially forcing them to choose between missing out on promising AI startups or breaking their fund models. The situation underscores the evolving dynamics in venture capital, where fund size increasingly dictates investment strategy.

𝘖𝘯 𝘉𝘶𝘣𝘣𝘭𝘦𝘴 (𝘈𝘣𝘰𝘷𝘦 𝘵𝘩𝘦 𝘊𝘳𝘰𝘸𝘥)

• Bill Gurley's 2014 piece draws parallels to the current AI investment froth, where enthusiasm might be outpacing prudence. The rapid rise of AI applications echoes past cycles, with investors, employees, and other actors discounting risks to avoid missing out. This herd behavior can lead to increased risk tolerance not too dissimilar to ZIRP, potentially setting the stage for a bubble.

Artificial Intelligence

• We aren’t running out of training data, we are running out of open training data - Interconnects.ai

• Excerpt: “Despite the definite advantages that the big tech players have in this domain, they’ve still signed dumb data deals with mid-sized technology companies and small platforms to get access to their customer data. Examples include Reddit, news agencies, Stack Overflow, and more, where labs sign these deals that present as “crucial infrastructure” for the next generation of models, where in reality it seems most like a legal strategy to punt potential lawsuits to firms with shallower financial pockets.”

• Read More

• The Internet as You Know It Is Dying - Fabricated Knowledge

• Excerpt: “In the rush to make bigger and better models, we are not only changing how we compute and the infrastructure around that, but also the internet itself. I think this is the crucial next step of the internet, and Web 3.0 will not be the democratization of the internet, but rather just one giant training ground for what’s next. The internet as you know is dying so it can birth the next generation of AI models.”

• Read More

• The Best Ways to Run AI Locally - Pat McGuinness

• Excerpt: “ This has changed: Smaller open-source AI models are getting really good. Phi-3 Medium, a14B model, is almost as good as Claude Sonnet, with an MMLU of 76. Just-released Mistral’s Codestral achieves over 81% on HumanEval for Python, making it a great CoPilot for coding (literally a drop-in replacement for Github CoPilot using Continue.dev).”

• Read More

• WWDC, Apple Intelligence, Apple Aggregates AI - Stratechery

• Excerpt: “ In short, my updated current thinking is that both Apple and OpenAI are making the bet that very large language models are becoming increasingly commoditized, which means that Apple doesn’t have to pay to get access to one, and OpenAI sees scale and consumer mindshare as the best route to a sustainable business.”

• Read More

• How Apple Intelligence Works, Apple’s AI Bear Case, The Talk Show Live - Stratechery

• Excerpt: “What is fascinating to consider is the extent to which these intents might be a Faustian bargain for developers: the implication of exposing functionality to Apple Intelligence is that the user interface of an app is diminished in importance; this is fine for apps that have differentiated resources that are not dependent on owning the user interface, but it does provide a long-run existential risk to apps that rely on user attention (in the most obvious use case, to show ads). The power of an Aggregator, though, is to compel suppliers like app developers to deliver functionality on their terms, so to the extent that Apple Intelligence becomes a meaningful interface for accomplishing things on your phone, is the extent to which developers will have to run the risk of losing user attention even as they deliver user benefits.”

• Read More

• AI for the rest of us - Interconnects.ai

• Excerpt: “ In summary, Apple foundation models focus on personalization (alignment strategies), performance (core models), and size (on-device strategies). Only the last one is orthogonal from other developers out there, but it surely added substantial constraints that inspired the other two.”

• Read More

• From Syntax to Semantics - Alessio Fanelli

• Excerpt: “Most codebases are also messy and are optimized for humans to write syntax, not for LLMs to manipulate them. This usually looks like class hierarcy to abstract common functionalities, extracting modules, monkey patches, moving specific functions to independent microservices, etc. As logic gets spread across files and codebases, it becomes harder and harder for models to intervene.”

• Read More

• Bain - AI Survey: Four Themes Emerging - Bain

• Excerpt: “Five use cases show signs of success: sales and sales operations, software code development, marketing, customer service, and customer onboarding. Meanwhile, use cases in legal, operations, and HR appear less successful.”

• Read More

Society and Technology

• The Sony Walkman - Commoncog

• Excerpt: “More important, the Walkman changed people's relationship to technology; its solitary, enveloping quality became its defining feature. The Walkman and its rivals quickly became a landmark in the history of media and a symbol of an inwardly focused era. "Personal sound" was a forerunner of personal computers and personal digital assistants.”

• Read More

• How to Build an AI Data Center - Construction Physics

• Excerpt: “The amount of computer equipment they house means that data centers consume large amounts of power. A single computer isn’t particularly power hungry: A rack-mounted server might use a few hundred watts, or about 1/5th the power of a hair dryer. But tens of thousands of them together create substantial demand. Today, large data centers can require 100 megawatts (100 million watts) of power or more. That’s roughly the power required by 75,000 homes, or needed to melt 150 tons of steel in an electric arc furnace.2 Power demand is so central, in fact, that data centers are typically measured by how much power they consume rather than by square feet.”

• Read More

• Why some (movie) franchises go downhill while others stay great - Noahpinion

• Excerpt: “Good fictional universes are always bigger than the story. And when you try to expand the story to show every single important event that ever happened in a fictional universe, the universe becomes cramped and small and there’s no room left for your imagination.”

• Read More

Finance and Economics

• When It’s Legible, It’s Too Late - Capital Gains

• Excerpt: “This creates a frustrating optimization problem: by the time you've read the movie or seen the book, the specific thing you aspire to is almost certainly out-of-date. Some institutions are built on this; the World Economic Forum is continuously cycling through the latest set of people who don't realize that Davos isn't cool anymore until they actually get there.”

• Read More

• SaaS Isn't Dying: It's Going Logarithmic - Sleeper Thoughts

• Excerpt: “Companies routinely cite a “tough macroeconomic environment” on their earnings calls despite unemployment near all-time lows and a stock market near all-time highs. After a pause in CY23, SaaS hiring is back in full force, with more headcount than ever. The typical scaled SaaS company continues to run a net loss on a GAAP basis and makes almost no effort to justify the massive level of investment implicit in running a business so far below its terminal margins. Management practices that were helpful shorthands in an age of exponent are wasteful and value-destructive in a logarithmic era.”

• Read More

Early Stage Investing

• The Attio Deep Dive: How it Builds Product, Does PLG, and is Challenging Salesforce - News.Aakashg

• Excerpt: “Attio’s philosophy about engineers is one of the most unique things about its product process. It hires for product engineers. It doesn’t hire regular software engineers. This structure enables engineers to work closely with customers and maintain a tight feedback loop. “Every engineer at Attio has a very good product sense. If you’ve only worked your tickets, you probably won’t pass the interview process.” […] And it’s not because they dislike PMs. It’s because they love speed.”

• Read More

• Why It's Hard for Small Venture Funds to "Play the Game on the Field" in AI Investing - Chudson

• Excerpt: “Many founders with domain expertise are not raising small pre-seed and seed rounds; they are going directly to multi-stage funds and raising larger rounds of capital at higher valuations for their initial rounds. If you see these rounds and have the opportunity to invest, the prices and associated ownership will likely break your portfolio construction, and these companies will look like exceptions to the model. It’s hard to build a portfolio of companies that are all exceptions to the model you pitched your limited partners when you raised your fund.”

• Read More

• On Bubbles - Above the Crowd

• Excerpt: “Ask yourself this question. What is the percentage of employees in Silicon Valley that are working at profitless companies (i.e. companies that are losing money or have negative cash flow)? And how has that trended over time? What was that percentage in 1999? What was it in 2003? And what is it today? An employee’s decision to work for a company that is losing money is an implicit decision to discount risk. If the macro environment changes, that company is under much greater stress than one that is profitable. Yet many individuals are making just such a decision today.”

• Read More

On Writing

• How To Succeed On Substack - A Writer’s Notebook

• Excerpt: “Substack is based on the premise that good writing is worth paying for, and it is. But not all writing is good writing. Not all writing is valued in the same way. Sometimes even the best writing will not find a large audience, and that is simply the way of literature, the result of the current state of popular culture. Even so, we cannot expect to be richly compensated for work that is not at a professional level.”

• Read More

Life and Meaning

• How I Survived (Barely) as a Jazz Musician at Business School - The Honest Broker

• Excerpt: “But I picked Stanford Business School and for all the worst reasons. It was just two years—while law school is three years and the joint law/business program is four. That meant I wouldn’t have to borrow so much money. And I opted for Stanford over Harvard because of the weather. Laugh at that if you will, but I had spent too much time riding my bike through the English rain—which is less poetic the more you get drenched.”

• Read More

Web3

• The Social App Thesis - David Phelps

• Excerpt: “It was wrong, first of all, that financial incentives could build retention. In fact, the reason financial incentives are so good at user acquisition is exactly the same reason they’re so bad at user retention—because a mercenary who will use an app to profit will leave it just as soon as the opportunity is better elsewhere. The same people who come for a price that goes up will leave for a price that goes down. Their loyalty means nothing unless you can continue to get them paid.”

• Read More

Interesting Charts: