Info Diet offers a peek into the personally curated feeds and media habits of the people shaping the future. In each installment, a different builder spends two days chronicling all the content they consume in order to stay ahead of the curve. This time: Elena Burger, a deal partner on the a16z crypto team, where she focuses on games, NFTs, web3 media, and infrastructure. Elena lives in New York City.
Sunday, July 10
10 a.m.: It’s an uncommonly lazy Sunday for me — usually I wake by 7 or 8 a.m., caffeinate, go for a run in Central Park, and skim Twitter/my email. But today I happen to be visiting a family member’s beach house with a friend, so my unflagging morning peppiness has been tempered a bit.
I sip a coffee at the kitchen table and pop open an issue of The New Yorker sitting in front of me. This is another uncommon thing: I used to read a lot of print media, but most of my media diet is digital nowadays. I skim a profile of Mavis Staples, the gospel/blues singer, as people stream into the breakfast room.
12 p.m.: I’m on the Jitney bus heading back to Manhattan when my friend pops one of his airpods in my ear and asks, “Have you ever watched Channel 5?” No, I hadn’t.
Channel 5, I discover, is YouTuber Andrew Callaghan’s man-on-the-street interview show, where he seeks out potentially controversial scenarios (a pickup artist bootcamp, an anti-vax rally) and films the participants doing their thing. We watch the Utah Rap Festival episode (I’d recommend lowering your volume to slightly below “max” before clicking the link), which features some truly masterful rapping from a 15-year-old. I enjoy the video — it’s kind of similar to shows like Nathan For You or How To with John Wilson. I don’t watch a ton of YouTube, but as with the Mavis Staples profile, if something is in front of me, I’ll typically consume it.
I then pop in my airpods and listen to the Lex Fridman interview with Grimes from a few weeks ago. I come away from it absolutely loving her. I actually hadn’t realized how deep into crypto she’s gotten and that she has a very techno-utopian view of web3. She discusses stuff like using DAOs to fund motherhood and open-source research, and also talks about how web3 or AI can automate the function of music managers and other middlemen. It’s so interesting to me that artists can immediately grok crypto use-cases a lot more practically than many people in finance or traditional tech do. I’m also now super curious about what Mavis Staples would think about DAOs.
As an aside, there’s a moment in the podcast where Lex mentions that he lives next door to a well-known anarchist, which makes me wonder: If you’re an anarchist and you do own private property, what tenets of anarchism do you actually bother to uphold? If someone knows, feel free to ping me.
3:30 p.m.: I’m home. I open up the SITALWeek (Stuff I Thought About Last Week) newsletter on my laptop, which I read every Sunday. The newsletter is compiled by Brad Slingerlend, who focuses a lot on topics in traditional finance and tech: AI, semiconductors, the streaming economy, inflation, and oil/natural energy. As someone who spends 95% of my time thinking about crypto, it’s a great “touch grass” leveler for me.
This issue has a great piece on the impact of robotic surgery on the skill levels of trainee medical students, China’s dominance of the lithium-ion battery market, and a few interesting links about looming austerity spending among various streaming services.
4:30 p.m.: I go for a walk and call my brother. He’s an electrical engineer, and for the past few months I’ve been trying various tactics to crypto-pill him. I’ve probably been a bit too aggressive recently because the first thing he says when he answers the phone is “Elena, I’m not coming to the Stanford Blockchain Conference.” Alas.
We begin talking about all the recent advances in AI and image generation, and we discuss DALL-E, which I recently got access to and have been having a lot of fun playing around with. My brother sends me an Astral Codex Ten piece called “A Guide To Asking Robots To Design Stained Glass Windows,” in which the author describes his process of guiding DALL-E asymptotically closer to the image he wants rendered.
When I use DALL-E, I toggle between trying to dominate it with meticulously specific prompts, and deliberately feeding it something open-ended and a bit vague. As an example, a few weeks ago, I prompted it with “Thomas Edison selling bootleg DVDs on the street in the 90’s.” The first two images it produced look like they’re from the 1990s (saturated colors, Tom-Ed flanked by a pickup truck), but the remaining four look like they’re from the 1890s — black and white, grainy images, and the DVDs are unnaturally large and oblong, like the only way you could watch them is if you ran them through a visual phonograph. This was a creative leap I don’t think I would’ve been capable of, so it’s fun to think that DALL-E accidentally (or intentionally…) got there on its own.
After reading the Astral Codex Ten piece, I realize that I can sort of answer the question I had before (“What would Mavis Staples think about DAOs?”) if I run it through GPT-3, which I also have access to via the OpenAI playground. I give it the prompt: Write a paragraph from a New Yorker article about Mavis Staples talking about cryptocurrencies and decentralized autonomous organizations.
Here’s the result:
Mavis Staples has been singing gospel and soul music since the nineteen-fifties, and she’s as revered as a living legend as she is loved as a performer. This past fall, she released her latest album, “If All I Was Was Black,” a collaboration with the producer and musician Jeff Tweedy. Staples is also a venture capitalist. In early December, she participated in a panel at the Web Summit conference, in Lisbon, on the subject of cryptocurrencies and decentralized autonomous organizations. “I’m not a techie, but I’m learning,” she said.
Pretty amazing!! The only inaccuracy I notice (apart from the fictionalized details about Mavis Staples’ second life as a VC popping up at Lisbon crypto conferences) is that Mavis’ most recent album is “We Get By,” released in 2019. It’s quite possible that this is because the training data used by GPT-3 isn’t quite up to date, though I’m not sure.
7:30 p.m.: I don’t really recall how I ended up in this rabbit hole, but I’m reading about Minecraft servers. I find this great Rock Paper Shotgun piece from 2015 on the emergence of Hypixel, the largest independent server by CCU (concurrent users). It hit a peak load of 216k players in 2021, and regularly averages around 45-70k players/day. I’m still trying to figure out how Minetrack quantifies CCU’s in the context of Hypixel games, because the server itself directs players into much smaller “parties” in which the games are played (so I don’t think you’re actually seeing tens of thousands of players rendered around you). But I could be wrong. Scaling games to tens of thousands or even millions of CCU’s is still a massive engineering hurdle — Water & Music recently did a very good writeup on this topic in its “9 design principles for a musical metaverse” (skip to the “Scaled Synchrony” section for more on this).
Monday, July 11
7 a.m.: I wake up and do a few laps around Central Park. I only listen to music when I run (I tried podcasts for awhile, but my mind wanders when I jog), and right now my rotation is a lot of Meek Mill, artists like SOPHIE, and aughties-era producers like Ryan Leslie.
10 a.m.: Again, I’m doing my “non-crypto” news rotation. At this point even non-crypto news is crypto related: chip shortages, inflation, entertainment, etc — it’s all related. One of my other go-to newsletters is Liberty’s Highlights, which is a great roundup of news related to the earthlier matters of the world: semiconductors, chips, traditional media, etc. This issue actually cites the SITALWeek link from yesterday about surgical automation — there are also some interesting links from The Wall Street Journal and Compound Capital Advisors CEO Charlie Bilello, respectively, about the energy-efficiency of electric vehicle batteries and Eurozone inflation.
8 p.m.: Most of my day is spent on calls or answering emails/Slack messages — not a lot of time for consuming news (other than tweets). I do read a lot of pitch decks and emails from founders, but my time for deeply reading articles or falling into rabbit holes is reserved for later at night (or on weekends).
Tonight, I decompress by catching up on the Reddit AMA with the Ethereum Foundation research team from a week ago. I definitely try to stay as up-to-date as possible on discourses surfacing from the EF (especially as we approach the Merge). It’s pretty interesting to read through the AMA and understand that there isn’t always complete agreement among the core devs (a good case-in-point are the two responses to this question about the gap between research and implementation).
The biggest area of consensus seems to be around the kinds of applications that people are looking forward to in the post-rollup, scaled-Ethereum future: the agreement (among Vitalik, Justin Drake and Danny Ryan, at least) is that applications around identity, reputation, and privacy will see a lot of innovation.
(Another great newsletter that I regularly check is What’s New in Eth2, compiled by Ben Edgington at ConsenSys. Also while we’re on the topic, a necessary shoutout to Tim Beiko for his extremely detailed notes from the recurring Ethereum AllCoreDevs calls.)
1 a.m.: After dinner, I continue to catch up on messages and scroll through Twitter. Then, I’m somehow back in a Minecraft rabbit hole. Before drifting off to sleep, I read this OpenAI research post on training a neural net to find and craft items in the game. Impressively, only a small portion of the videos used to train the neural net were labeled with corresponding mouse and keyboard actions. The rest was just an enormous feed of videos uploaded by players online (in other words, the neural network was able to extrapolate and learn from a small set of labeled data, and apply it to a much larger set of unlabeled data).
The more I pay attention to things happening in the AI space, the more I’ve been thinking about how the field will converge with crypto. Will we see more protocols that seek to implement on-chain AI as a community-owned tool that anyone can audit and upgrade? 0xPARC had a proof-of-concept a few months ago demonstrating an on-chain numerical image classifier inside of a zkSNARK, and a few years ago Ali on our team published a great Twitter thread on how one might go about bootstrapping an autonomous neural network on-chain. I hope we’ll continue to see more of these kinds of applications in the future.
Get Elena’s Info Diet
- “The Gospel According to Mavis Staples” by David Remnick [The New Yorker]
- “A Guide to Asking Robots to Design Stained Glass Windows“ [Astral Codex Ten]
- “The Future of Minecraft’s Biggest Servers” by Alex Wiltshire [Rock Paper Shotgun]
- “9 Design Principles for a Musical Metaverse” [Water & Music]
- “[AMA] We Are EF Research” [Reddit]
- “Learning to Play Minecraft with Video PreTraining (VPT)” [Open AI]
- “ZK Machine Learning” [0xPARC]
- “Grimes: Music, AI, and the Future of Humanity” [Lex Fridman Podcast]
- “Utah Rap Festival” [Channel 5 with Andrew Callaghan]