1. optimization/FIRE mentality
2. most of technical talent are immigrants or children thereof, who adopted frugality/scarcity
3. startup salaries are built on shaky foundations (illiquid equity, vc raises, negative margins, unproven R&D); wise to save for periods of unemployment
san francisco is the only city in the world where the more money people make the less they spend
the things people who make $1k an hour are ready to do to save $10 is shocking
Is there a term for introspection by seeking out media that affirms your lived experience? For discovering unknown knowns (blind spots) by finding words that articulate your beliefs better than you could have yourself?
One app I really miss: @Detour
It was an immersive podcast experience that blended storytelling with location- you’d walk through a city while an amazing narrator (like Ken Burns) guided you, syncing their voice to your exact position... you could do them at your own speed, the audio would work even if you walked faster.
It felt like being inside a documentary, with local voices, hidden histories, and personal anecdotes unfolding as you moved.
It was founded by @andrewmason (post-Groupon)... at the time (10 yrs ago) it was one of the best demos I’d ever gotten. Intimate, place-based audio.
I think it probably struggled as a business (not sure how you acquire customers for a travel app like this), was acquired and shut down by Bose in 2018. No replacement has captured that magic for me since.
They spun out the tech that they used to build it, and it became @descript (audio & video editor)
You could make a better version today with AI, personalized to the user.
Imagine: NYC, by Lin-Manuel Miranda
Mumbai, by Suketu Mehta
the sounds of Reykjavík, by Bjork
A foodie tour of Rome, by Massimo Bottura
Mexico City Taco Truck Tour, by Rick Bayless
Not sure why more feeds/recommendation engines don’t support online pre-caching.
You should be able to preload x hours of content just before a flight. Save any likes and comments to be publish when reconnecting, along with evicting the cache.
@drummatick I remember for my freshman year EE class we had to SSH into the lab computers and transfer our files to submit our assignments. By the last year everything had switched over to Canvas/Gradescope. I think about what was lost there.
Told you watching movies and appreciating art are crucial for scientific & technological breakthroughs.
Self-attention was inspired by the alien language in the 2016 movie Arrival, based on Ted Chiang’s Story of Your Life, it would be a fascinating connection if true. @kyosu
In the age of symbolic AI, the preferred AI programming language was LISP, or C (Deep Blue). For statistical ML, this is arguably R. For current deep learning, this is Python with CUDA bindings. Will there be a next revolution and what language(s) will we use to code it?
Idea: a bank for your stuff. You pay a fee to store what you don’t need right now, like self storage but everything is catalogued and others can borrow it for a time (like a library). If you stop paying, anything you don’t pick up is sold or recycled. Great perk for apartments.
I agree. Until we add constraints or inpainting abilities, it’s just a toy. But at this pace, you could use depth inference to capture camera pos, textured 3d entity models explicitly.
I’m also in its ability to do “unconditional outputs.” How can we explore Sora’s latent space?
The reason I'm not scared (yet) of the Sora vids as an animator is that animation is an iterative process, especially when working for a client
Here's a bunch of notes to improve one of the anims, which a human could address, but AI would just start over
What client wants that?
The irony of ad streaming tiers is that the customers who will likely be viewing the ads self select for frugality, and are therefore less likely to engage. This incentivizes the streamer to make the ads annoying so they just upgrade to a higher tier.
Future LLMs will allow programmers to work in their own custom pseudocode, translated in the background to/from idiomatic HLPL code. Also: codebase rewrite/switching costs will decrease, the new major languages will come from leading LLM shops as intermediate representations.
@sethbannon Loved this book. We’re beginning to see an uptick in affordable lab space in the upper peninsula thanks to decline in commercial real estate, esp brick and mortar retail. I’d rather lose a mall if it means lower cost per bench!
Curious to see how good LLMs are at Systems Thinking. In my experience, ChatGPT is not good at extending or reasoning over graphs, but hopefully this is just a prompt engineering issue.