✨ New features for Berkshire Archive ✨
* Added all meeting transcripts, 1994-2023
* Added new 2023 letter
* Faster search
* Explore archive in 3D meaning space
* Navigate each doc via table of contents (top right)
The transcripts add a lot of value -- there is WAY more
1/ A recent lightbulb moment of mine was that competitive advantage can be represented visually as 1 or more feedback loops. These create an advantage "flywheel" that maintain and grow a moat over time.
✨ Just launched: A digital archive of all Buffett's letters to shareholders ✨
🔍 Search the letters by topic, question meaning, or similarity
📖 Browse each letter & search for similar sections
(link in thread)
Just launched:
Ask Warren Buffett* anything! Try it out and let me know what you think...
* Not actually Buffett, but an AI model with access to his writing. This is just a test/fun experiment + promotion for my Berkshire media.
🧵 1/n A quick primer on GPT-3 for anyone who's heard about it but doesn't know what it is.
Why? GPT is a game-changer in AI that has the potential to disrupt a huge amount of areas, potentially leading to truly generalized AI problem solvers.
I'm building a digital Berkshire Hathaway Archive.
You'll be able to read, search, and explore Buffett's letters like never before. Working on the search algorithm now.
1/ One of the models I use most in business analysis is tech stack trees 🌳
Every product is built on and enabled by 1 or more technologies.
Understanding where a product fits on its higher-level tech stack is an important part of any long-term strategy or investment thesis.
First sketch of an LLM idea maze ala
@balajis
The main branch I chose was what kind of interface: Chat, Search, Copilot. From there you have areas like assistants, self-help, guides, etc.
Where will the highest areas of value be?
Request for podcast series:
Every episode is one of
@patrickc
's Fast entries.
Radio-show-style storytelling series with interviews, narration, good sound production. Some episodes could compare the fast case with a modern-day slow case
Finished the "Arnold" doc last night. Really enjoyed it. One of my biggest takeaways was about the goals he had for each of his "lives". Each goal was:
1. Very ambitious
2. Fit his desires & abilities well
3. Timed right
The last one is key. Becoming Mr. Universe is crazy
From the archives: Microsoft S-1 from 1986.
1. Raised $42m, diluting existing shareholders by 8.1%
2. Prior year sales were $140m, pre-tax profits $41m
4/ This helps emphasize that the most successful moats have multiple flywheels that feed off of each other's momentum. Like Google's tech advantages enabling stronger brand allegiance. Or Coke's marketing-driven brand feeding off of the distributor based network effects.
5/ Using the analogy of a feedback loop helps to think of an advantage as a moving, changing system. A system that needs catalysts to get started, and will gain momentum at first but still be slowed by friction over time.
Planet Labs $PL is now trading at market cap of ~$1.3B. Given their likely dominance of earth observation in coming decade, this seems... low. Market and margins will continue to grow.
Proud to announce that Mashgin is profitable & we’ve raised a Series B of $62M from NEA and other great investors. We’re currently scaling to tens of thousands of locations worldwide, including C-stores stores across North America & Europe.
This is going to be a LONG thread on how I think about active investing.
They say all good investing is value investing. This is true in the sense that you're buying into a situation where the sellers are underpricing potential outcome. But…
I'd bet if we put advanced computer vision at every major traffic light, you'd get major non-linear improvements:
* Drastically reduced commute times
* Less emissions
* Many lives saved, indirectly (less traffic overall) & directly (warn when someone about to run red)
Cheaper,
1/ I've been revisiting the concept of "platforms" lately and their significance in tech stacks. What does a having a platform really mean? Here are some thoughts & resources 👇
Just finished the book "Organizing Genius", which I believe was a recommend from
@patrickc
. Authors recount & compare "Great Groups" like PARC, Disney, Manhattan Project, Skunkworks, etc. Thread 👇
Just finished “The Fish That Ate The Whale”. How has this not already been turned into a mini series on Netflix!? Such an amazing story and cast of characters.
CC
@EricJorgenson
@david_perell
From 1 of my experiments with generative neural nets, this time basic style transfer (tweaking params of an open source repo). Goal was to emulate style of WSJ portrait drawings -- result turned out not too bad:
@elidourado
* Universal living modules, capable of sustaining life in any circumstance (underground, ocean, arctic, space)
* Permanent World's Fair revival
* Interactive map of time/space w everything we know of that's ever happened
* Programmable GE trees to grow & maintain infrastructure
Check it out 👉
The idea came about when doing a bit of "customer development" with some folks at Berkshire HQ. Turns out they wanted a good way to find things in the letters! My book originally did this job for them, now with LLMs there's a better way.
Really interested to keep an eye on this. Not just generative ai in particular, but a company that utilizes *all* the latest tech (like Pixar did) to tell amazing stories.
I created a vector database of all my Readwise highlights using OpenAI’s embeddings API.
It's an amazing way to semantically search anything I've highlighted in the past. I can also look at "similar" quotes that may not even have any overlapping works, but similar concepts.
Always forget how much leverage Buffett had Berkshire in the early years. Over 2-to-1 for much of the 70s, with >100% of equity in stocks for 5 years up to 1983. Went down again & peaked again in 1992 at 129%(!) equity.
Grateful today for our new baby girl, born 7 weeks early but healthy and already surprising everyone with her strength! And thankful to mom who she got it from 😊
🤯 Watch this demo + read the logs as it goes.
Multiple GPT4 agents in a recursive loop to solve a problem. This is what I mean by "the existing models are all we need". When these patterns start to get productized... watch out.
Massive Update for Auto-GPT: Code Execution! 🤖💻
Auto-GPT is now able to write it's own code using
#gpt4
and execute python scripts!
This allows it to recursively debug, develop and self-improve... 🤯 👇
Obvious application of generative agents is games & other sims. Non-obvious:
These ephemeral "worlds" may be spun-up on command w/ different characters to solve specific problems, similar to getting diverse inputs from people IRL. May produce better output then 1 big LLM chain.
🤖Generative Agents🤖
Last week, Park et all released “Generative Agents”, a paper simulating interactions between tens of agents
We gave it a close read, and implemented one of the novel components it introduced: a long-term, reflection-based memory system
🧵
Optimism time! Starting a thread on all the major innovation breakthroughs and milestones that happened in 2020.
I'll add to the thread over the next few days ⬇️
I'm not the best investor in my family. That honor goes to my grandma, who died in November at the age of 100.
In 1942 she spent one of 1st paychecks to buy ~$40 of shares in a local railroad in Virginia.
After M&A over decades it became $NSC. 80 year CAGR = ~11% before divs
🌎 Inspired by
@pmarca
's rec of "Whole Earth Discipline" here are some highlights on my notes on
@stewarts
book. See link full post at end of thread 👇 1/n
Best product strategy tip: Build a product for 1 customer, then 10, then 100, and so on. This is *very* important. Don't think about the next stage until the customers in current stage love you.
Ahead of the annual letter tomorrow, here's a thread with a few interesting
#brk
stats I learned from making this data visualization 👇
(all of the numbers apply only to the 50 year data from 1965 through 2014)
The new larger-format, print-on-demand version of the Berkshire Wall Print is now available. This uses thicker, higher quality paper. You can also purchase a pre-framed version or frame the print yourself.
Amazing that in 40 years, through multiple new tech waves and platform transitions, the dominant technology companies are still:
Apple and Microsoft
Both founded in the late 1970s by hackers in their early 20s. And seems like they'll continue to dominate for next 10 years.
Work in spartan, shabby environments ("low road" as Stewart Brand calls them). The right tools are essential, but fancy digs aren't. "The tendency of great things to be accomplished in dreadful spaces should give architects and decorators pause."
I see these anecdotes all the time now, and it’s implied that this is a *California* thing. But really it’s just an SF thing.
In the Peninsula and South Bay for example, no one seems to be leaving and there’s crazy demand for houses. Prices up 10% in past 3-5 months.
Amazing that people think businesses maintain dominance primarily via gov’t enforcement. Similar to other conspiracies, I guess it stems from lack of understanding of how world works... easier to point to 1 person/entity to blame.
Bill Gates made his money by extinguishing free software and forcing us to pay him a tax just to turn our computers on. This was only possible because the State was ready to violently enforce his "intellectual property".
He didn't created a computer revolution, he destroyed one.
@TurnerNovak
He’s also very dyslexic and from a mutual friend I know he doesn’t read books or much financial info as a result.
Shows the upside to relentless focus and grit.
@Ben_Reinhardt
Just read & enjoyed your post on difference between ARPA projects vs FROs.
I’m curious your thoughts on where each falls in this framework?
Thread: Just finished 2 books on the history of Polaroid 🌈. A remarkable tech company with enormous success in consumer & industrial applications for decades. Also remarkable how much Apple was influenced by Polaroid.
@WilliamAEden
Agree with a lot of this, although I think many areas are much more likely to accept 99% accuracy (or lower) than you say. Alternative is high priced humans that also have <1 accuracy.
My experience is in AI self-checkout. We haven’t seen much difficulty convincing <1 is ok.
Thread: Marc Andreessen
@pmarca
tweetstorm from 2015-02-16 on how to expand the amount of startup founders: [RE-POST]
1/ More Montessori & Montessori-style, free-form, and/or project-based K-8 public & private schools.
Damn this is a beautiful plane. This angle really makes it clear it's really just a cockpit attached to 2 enormous jet engines. And no weapons needed if you can just outrun any missile or bullet!
@gammatrader09
@wolfejosh
The numbers don’t add up for this. These are not huge positions for Berkshire. All their airline holdings combined were only 2% of book value at end of 2019, even less now. If all went to 0 it still wouldn’t outweigh a huge deal that works out.
10/ The generalization of the model is the real killer here. The SAME model can output customer service chats, music/movie recs, legal docs, summaries of sporting events, poetry, code functions, and complex descriptions from simple text prompts.
@farnamstreet
"There's no time to lose," I heard her say.
Catch your dreams before they slip away.
Dying all the time,
Lose your dreams and you will lose your mind...
Ain't life unkind?
-- Rolling Stones
Interesting episode. Very similar to the premise of John Kay's book "Obliquity" -- the indirect route is often the best, especially for higher-level, more complex goals.
See clips from a post I wrote on this a few yrs ago:
My conversation with
@kenneth0stanley
about why you shouldn’t pursue big hairy audacious goals
I hope this stirs up some controversy b/c the implications of the concept are large
That he “discovered” this idea via AI research fascinates me
Enjoy!
Thinking about making a SpaceX equivalent of the BRK "letters to shareholders" book, with chronological, primary-sourced history of the first 10 years (2003-2013), mostly written by
@elonmusk
. So many amazing lessons on startups and rocket engineering.
100,000 Americans dead from a global pandemic. Race riots in cities across the country. Astronauts launched past lower earth orbit, inspiring the world. 2020? Nope, 1968.
@EricJorgenson
I think you’d be proud: Spent a few hours this weekend coding a bot that checks all pharmacies in area for vaccine appointment availability every 30min and texts a family member if there is.
🔼🔼🚲🚲
(are these good emojis for leverage??)
Amazing all the negative responses to this thread, many seemingly from doctors. “But false positives!!!” Yeah no shit, so you factor that in and get better over time. More data can lead to erroneous conclusions but fixing that by not collecting at all is laughable.
On Monday I went to
@prenuvo
for a full-body MRI. The idea is to catch diseases like cancer early, when they are still treatable.
Wild numbers: out of 1,000 people scanned, 44 will have cancer, often in a treatable stage. Worth doing.
@TaylorPearsonMe
“The Art of Strategy” for general game theory and “Prisoners Dilemma” for great stories on game theory in history (nuclear weapons, Cold War etc.)
7/ What are examples of platforms that emerged pre-computing? My list:
🧭 Silk Road trade route
🛤 Railroads
🔌 Electric utility grid
✉️ Rural mail delivery
🛣 Interstate highway system (cars, trucks, retail)
🚢 Intermodal shipping
☎️ Telecom system
📻 Radio
📺 Over-the-air TV
No doubt that if Trump was the CEO of a large co he'd be fired by any reasonable board for similar comments and actions. Exactly the opposite type of leader that Buffett looks for in all his investments.