🏗️ Love to build stuff (@runwayco, @sandboxvr, @postmates, @zynga) people love. 💸 Investor @amplitude_hq, @mercury, @slope_finance, @italic
Jun 19, 2023 • 6 tweets • 3 min read
zaha hadid architects trained a generative model that renders architectural designs in the style of renowned architects ... from randomly crumpled pieces of paper.
this one is in the style of frank gehry:
🧵👇
Daniel Libeskind
Apr 2, 2023 • 4 tweets • 2 min read
had multiple conversations this week with builders who think it's already possible to build self-improving autonomous agents through GPT4.
key belief shared is that we are just scratching the surface of what it can do through better prompt engineering:
e.g
you can pretty easily imagine an opportunity to use an LLM for meta reflexion / self improvement to improve its own architectural loop / prompts so it can discover things like this autonomously:
🤯 this paper demonstrates you can improve gpt4 performance an astounding 30% by asking gpt4 to reflect on “why were you wrong?”, and generate a new prompt for itself taking that reason into account until it is correct.
i have been really struggling to figure out what to build today that is going to not be obsolete within 3 years and it’s distressing tbh
physical things like @SandboxVR is a good answer
Mar 14, 2023 • 4 tweets • 1 min read
gpt-4 (visual) is multi-modal.
which means it can take images as input.
so it can explain memes like this:
Mar 13, 2023 • 4 tweets • 2 min read
unpopular opinion today but f it:
you can find the style distasteful but @DavidSacks and @jason stuck their neck out to do everything they can to prevent a massive bank run knowing they would get skewered for it.
this is fair nobody was more effective here than @garrytan
Today I learned that when Twitter acquired @uenodotco in 2021, @iamharaldur quietly negotiated for the acquisition price to be paid as salary so he could pay 46% in taxes instead of 22%, in gratitude to his education in Finland.
Founders - don't use that standard 10 slide template from Google for your fundraising deck.
Here’s how to craft a fundraising story of your own that actually closes:
🧵:
These are the two most important principles of a good pitch:
1. Focus on the size of outcome if successful, rather than the probability of success. 2. Your deck is a vehicle for the delivery of emotions, not of facts.
Most fundraising advice you'll find online can usually be traced to some famous investor or other, so it might be tempting to just do whatever investors tell you to do.
This is a mistake, because investor incentives are very different from your incentives as a founder ...👇🧵
As an investor, your job is to invest in the very best startups and filter out shitty ones as fast as you can.
As a founder, your job is to get investors to invest in your shitty startup no matter what.
Those are very different incentives.
Sep 20, 2022 • 6 tweets • 2 min read
A lot founders (myself included) see Steve Jobs as this autocratic product genius who ruled Apple with an iron fist, but that's just a caricature.
A few less well known and under appreciated management gems from his 1992 talk at MIT:
🧵👇
Steve Jobs on resolving conflict:
"I've never believed in ... 'We're on all on the same team, you don't agree, but buy into it!' ... I've always felt the best way is to get everybody in a room and talk through it until you agree."
Nov 9, 2021 • 10 tweets • 3 min read
1/ We've been working on something new @runwayco, and we're pretty fired up about it.
Goal: build something 100X better than a spreadsheet, but just for the financial planning use case.
We think we're getting pretty close.
Here's why we're building this:
Excel is amazing. Excel will never die.
But for financial planning, we have had to put up with ... a lot.
Let me know if any of these problems resonate with you:
Aug 21, 2021 • 8 tweets • 3 min read
If you’re going to build something to replace Excel, it better be 100X better than Excel because 10X ain’t gonna cut it.
I like how this is framed a lot
Georgia: Biden behind by 23K votes, with 50K votes left to count. Votes have been breaking roughly 80% for Biden, which if holds puts Biden over the top if barely. This will be in recount territory either way.
Arizona: Biden ahead by 200K votes, with about 400K votes to go. Remaining votes is a mixed bag but leaning R. This is going to be pretty close but still looking good for Biden.
Nov 4, 2020 • 13 tweets • 3 min read
Biden is now leading in Wisconsin.
Michigan is looking promising.
If Biden takes MI and WI I think we're done.
Now up on @DecisionDeskHQ
I worked for a startup called Powerset in 2007 that tried to build something like this. We spent $100M and it was bought by MSFT to become Bing in 2008.
A single engineer can now build something that is literally an order of magnitude better in less than an hour with GPT-3.
It's a whole new world. I don't think we're ready for it yet.
Jun 25, 2020 • 9 tweets • 8 min read
The way business finance works today doesn't just predate computing, it even predates the popularization of negative numbers in the Western world.
What if we rethought financials for today's world where computers actually exist?
It’s software! I know this.
Thread 👇
12 years ago, I started my first company. I didn’t know how to read an income statement.
12 years later, I was running @SandboxVR. I still barely knew how to read an income statement.
Financials make me feel stupid, and I know I’m not alone in feeling this way …
May 11, 2019 • 7 tweets • 2 min read
1/ This is the sad story of how I didn’t invest in $UBER after seeing it prelaunch. Or: why early stage investing is hard. Or: why I’m an investing idiot. Thread 👇:
2/ About 9 years ago, I was out with @kamalravikant for drinks at the St. Regis in SF. He starts telling me about this magical app that can summon a luxurious chauffeured black car to pick anybody up, and drop them off wherever you want. He called it: Ubercab.
Daniel Libeskind 
