flâneur

surya

28 followers · 36 following · 2836 views

on the atlas — 106

highlights — 8

  • "No one can tell you what you can or can not do. With no rules to follow, this adventure is up to you" -- Official Minecraft Trailer 2011
    Hero's Journey
  • Technological improvements continued, but change became more evolutionary than revolutionary.
    AI Will Not Make You Rich - Colossus
  • In the mid-1980s, many [venture capitalists] experimented with investments in such low-tech industries as specialty retailing. Results were generally disappointing, in part because the financiers often didn’t understand what they were getting into. (Wall Street Journal, “Venture Capitalists Find Low-Tech Firms Appealing”, 6/20/1991.)
    Heat Death: Venture Capital in the 1980s | Reaction Wheel
  • Unfortunately, this theoretically nice thing about the internet has been defeated by a sadly definite thing about the internet: it is great at creating hiveminds, at funneling 99% of attention to 1% of things. Now not only can the people who bought the guidebooks learn about That One Restaurant Only Locals (And You) Go To but you can learn about it, too, and so can anyone else with access to the internet, and probably you will each hear about it from a different influencer saying something different, though equally exaggerated, about it.
    Unconventional Adventures - Quarter Mile
  • I tell founders after we lead their seed round, that the most important decision they will make for the first year of the company, is their choice in office lease. If you don’t have a top-tier office, top-tier talent will subconsciously be averse to joining you.
    Thirty Observations at Thirty
  • Arriving in Silicon Valley in 2012 I was overwhelmed by how much there was to learn about who actually built large tech companies and how it was done. Now, a decade later, I realize that the set of players relevant today has a massive overlap with those a decade ago. Not a whole lot changes.
    Thirty Observations at Thirty
  • In the world of steel manufacturing, historically, steel was made in massive integrated mills. They made high quality steel with reasonable margins. Then came along electric mini mills. These mills were able to make the lowest quality steel at a cheaper cost. The large steel manufacturers saw this, shrugged, and focused on making high quality steel at a (relatively) high margin. Over time, the electric mini mill operators figured out how to make higher and higher quality steel, moved upmarket, and killed the massive integrated mills (US Steel— once the 16th largest corporation by market cap in…
    The evolution of the LLM API market - by Finbarr Timbers
  • In speculative decoding, one has two models: a small, fast, one, and a large, slow one. As the inference speed for a modern decoder is directly proportional to the number of parameters, with a smaller model, one can run multiple inferences in the time it takes a large model to run a single inference.
    Transformer inference tricks - by Finbarr Timbers