đĄThings I Learned In 2023
A bad plan is better than no plan: This is a very similar concept with the âaction-firstâ mindset I have written many times in previous posts. I have fundamentally changed my mind about uncertainty and ambiguity. I used to approach them as the worst possible enemies and productivity killers and tried to resolve them with extensive analysis and hesitation to take action. I now realize that only actions, even wrong ones, can generate outcomes and only outcomes can help us progress. When they are good we can either optimize them or build new ideas on top of them. When they are bad, they help us take different actions and rethink our plans.
Always seek for responsibility: One of the major differences I have witnessed after moving from Big Tech to the startup world is that accountability/responsibility is not an option in startups. In Big Tech, people are asked to be accountable because the modus operandi is that bureaucracy can abstract any idea of responsibility from individuals. In other words, if you are smart and you want to âstay in the shadowsâ you can very easily do this. Of course, we know that this usually leads to complacency, fake jobs and ultimately massive layoffs. But still, itâs an option. In startups, even if you wanted to, you canât really hide from responsibility. This happens mostly because of the small size and flat structures of startups but also because startups tend to attract talent that wants to ship fast, experiment, lead. Responsibility in the only way to professionally and personally grow. Donât hide from it!
Concise explanations accelerate progress: itâs widely documented how writing helps people sharpen and clarify their thinking and how companies that have a a writing culture (Stripe, Amazon) tend to overperform. When it comes to emerging technologies and science, concise, dense and simple writing can be a catalyst of progress. It helps the inventor/scientist/founder not only to onboard more people but also build a kind of intellectual hyperstructure upon which others will build/contribute new ideas/input/code. Bonus: Steph Angoâs post.
On the Law of Three Stages: if you arenât familiar with this law, you may have seen the famous IQ Bell Curve meme. I personally first encountered this concept in Proudhon who believed that every science develops in three successive periods: religious, sophisticated, scientific. The idea is that âmidwitsâ, aka sophisticated, consider themselves to be the smartest people in the room and look down to individuals who are either religious or donât overcomplicate their assumptions but history proves that itâs the âmidwitsâ who are usually the dumb ones. A good example is the Covid-19 lab leak theory which at the beginning was adopted only by Trump supporters (religious), instantly ridiculed by smart, college-educated people (sophisticated) but was proved to be right (scientific). You can apply this to almost every concept/debate out there. It has helped me tremendously to show intellectual humility and avoid (or expose) midwits.
On toddlers as LLMs: itâs still difficult for many people, especially for those without technical background, to understand how actually LLMs work. One analogy I am always using in my conversations and it seems to resonate a lot is that LLMs operate in a similar fashion with toddlers. Toddlersâ brains have been trained on thousands of years of past human experience (evolution), finetuned by the way they are raised and educated and produce output based on what you will ask them while demonstrating emergent behaviors parents didnât expect beforehand.
On the risk of becoming a âcompany manâ: I am frequently asked (from friends) why I left Big Tech (Microsoft) to join a startup (Matter Labs) and I am usually replying that one of the reasons is the parable of the âcompany manâ. The company man is best defined as someone whose identity is impregnated with the stamp the firm wants to give him. He dresses the part, even uses the language the company expects him to have. His social life is so invested in the company that leaving it inflicts a huge penalty. After working at Microsoft for 8 years and living in Seattle which is a predominantly monoculture city, I realized that I was clearly becoming a company man. And I didnât like it. Itâs worth noting that Microsoft, and Big Tech in general, is not to be blamed for this. Itâs an inherent feature of large organizations (and human nature?) as Taleb points out. Which leads me to another theory I have been thinking a lot about: Small is usually better at innovating and managing risk than Big. Small countries, companies, organisms have proven to be more resilient and agile than their bigger/larger counterparts. When in doubt, pick Small!
Ideas are still powerful: we have come to believe that Humanities are dead and that students should invest time/efforts/money only on studying STEM but the power of ideas is still prevalent in the world. How else can justify the power and influence that Effective Altruism has in the tech industry? I am convinced that we will soon see a growing number of new intellectual camps/movements gaining traction and replacing EA and there are 4 reasons for this: a) itâs a cliche but people need a higher reason to exist/work/live b) religious attendance goes down c) last decadeâs obsession with optimization and secularism will backfire d) powerful people (VCs, politicians, elites) now understand the power of âsoft powerâ and wrap up their agendas in intellectual movements. Bonus: Nadia Asparouhovaâs post on Idea Machines.
We are dumb about intelligence: We still have no idea what intelligence actually is or how to construct it. I find this absolutely fascinating. We have reached the stars, we have built amazing civilizations, we have formulated theories about the universe but we still donât really know how the human brain works and what constitutes something as intelligent. I highly recommend you reading Levinâs latest research on algorithm emergent behaviors. In a nutshell, he proves that cognition exists in substrates weâd never imagine calling âintelligentâ.
Hype cycles are BS: Nobody can actually predict when a new technology will become mainstream. After spending 2 years in the Emerging Tech team at Microsoft and joining a startup that uses zero-knowledge cryptography to scale a blockchain protocol (!!), I am convinced that randomness plays a huge role on when something will stick. Utility, user experience, virality, status are all important factors but they are still ingredients of a recipe that nobody really knows how to make. I have been thinking a lot about this because I am trying to conceptualize what is the future of blockchain technology/crypto but I have now come in terms with the fact that you canât back engineer stickiness. You can enable it but not guarantee it. And thatâs actually very enticing on its own!
đ§âđ» Things That Excite Me
Palmer Luckey, Anduril and defense tech in general
Programmable Medicines
Zero-knowledge technology in general and its applications in GenAI in particular
Stablecoins
Decentralized Physical Infrastructure (DePin)
Autonomous Worlds
Collective Intelligence/Plural Organizations
Charter Cities (more than Network States)
AI/Git Books
đ Mind Crunches
The excellent Dwarkesh Patel on whether we can achieve AGI. Spoiler alert: 70% chances we will have AGI by 2040.
I cannot recommend Samo Burja and Bismarck Analysis enough. My most well spent dollars on the Internet.
The scientific breakthroughs of 2023. So much hope for the future.
Custom GPTs are here and I suggest you start using them as soon as possible. I have personally found some of them being much more useful than GPT-4 itself. For example, Universal Primer by the excellent Siqi Chen is my default option for GPT queries.
My latest obsession is Palmer Luckey. By far the single most innovative entrepreneur and original thinker out there. A real-life Tony Stark.
Another obsession of mine is the soft power of Stripe Press. I love each and every one of their books. I canât think of another tech company having a dedicated side media/content business. My theory is that Stripe Press and Works in Progress are both passion projects of Collisons and intellectual influence vehicles.
I wonât stop posting Luxâs quarterly investor letters. One of the finest examples of narrative/thesis-driven investing.
Great paper on the concept of NFTs as digital Veblen goods.
My opinion about the recent EU AI regulation.
On the importance of knowing your audience when launching a product. Mistral team has some cojones releasing Mixtral as a torrent! On a similar note, Midjourney is also a savage for using Discord as their primary interface.
a16z Big Ideas is an interesting read and a very good overview about the current state of most industries.
Balajis is more often right than wrong!
Fred Wilson on the rationale of monetizing a protocol.
A digital identity canon: probably the most comprehensive list of new primitives and startups building new digital identity solutions.
I have been truly impressed by Mistral and Mixtral AI. We are witnessing open source history in the making so weâd better look closer.
Recommended podcast: Stephen Jennings on building new cities, Conversations with Tyler.
Recommended Blogs: My favorite crypto blogs: Polynya and Emperor (you donât have to be a crypto degen to absorb them)
Recommended Book: Scaling People: Tactics for Management and Company Building, Claire Hughes Johnson. The GOAT book about running a company/leading a team.
Quote of the month: ââI donât have a talent, so I just get up earlier.â Henry Rollins
Photo of the month: Iceland Travel Project, Thomas Gravanis (you can learn more about his excellent work here)
Great EOY round-up, Vassili. Thanks for sharing your learnings.