In what is the least spectacular fundraising announcement you'll read this month —  I'm amped to finally share the close of @EarthlingVC Fund I, a $4.5M vehicle investing first round/first check in next-generation computing surfaces and mediums.

Startup city ranking that shouldn’t trigger anyone: S: Wherever you have access to your customers, wherever you have a unique edge in recruiting the best talent (big gap) D: anywhere else F: Dubai

Mistake #1 that I made a lot with my early angel checks was over-indexing on ideas I was biased to from my engineering day job. Learned from bad investments that just a “good idea” is a bad reason to invest (although it can make you sound smart in a memo)

I’m pretty happy in hindsight that I tried a bunch of stuff and made a ton of investing mistakes on my own dime before raising LP money (fully expecting to be roasted by all the naturally-talented “great pickers” for this take)

Announcing REK AMERICA🇺🇸 Road show bringing humanoid robot fights to America. LA to NYC stopping at boxing gyms across the country in between. Nov 11th - Los Angeles Nov 13th - Vegas Nov 17th - Austin Nov 20th - Miami Nov 25th - NYC Buy LA Tickets below. Let's REK AMERICA!

What this rhetoric misses is that AI is much more about “physical AI” (robotics) than LLMs (small piece of the pie), in the medium to long term and China is already actively flaming the US in robotics

i.e. I almost feel it's easier to raise $4M right now by starting the raise asking for $1.5M, building momentum fast, and scaling up than it is to ask for $4M upfront. Concentration into momentum deals in VC is wild rn

imo there is a lot of alpha for founders to fabricate scarcity in their raises: > ask for less $ than you intend to raise > fill up allocation fast (creates momentum/scarcity) > VCs start offering money at higher valuations > Take more money on a higher cap Happening a lot rn

Pulling blocks out of the AI Jenga tower

Now that my theory is closer to reality, this is why I'm concerned about what's going on with "Big AI": > as a trained ML engineer I was mostly taught to think about how to do things efficiently (i.e. increase inference quality with limited or less resources, even at Meta) > The problem with all these big AI programs (e.g. OpenAI, Nvidia etc.) is that they require excessive spending to become profitable > The only way to be profitable as an AI science company is by charging for compute. E.g. OpenAI's CFO has already stated that the current business model (selling inferences/APIs) won’t be profitable and they need to pivot to selling compute (this is also where Nvidia comes in) > This basically means these companies need to scale inefficient systems to make money. Excessively power-hungry AI systems will sell more compute (i.e. they don’t care about how “good” the science actually is). Its all about selling more machines. > This whole setup is like a Jenga tower. Eventually, some scientists will figure out how to maintain high inference quality with much, much less compute (thus cost), and when that happens, the whole tower blows up > tl;dr the way they are trying to grow to profitability is anti-science. A precarious position to be in a scientific field imo > This leads to 2 potential outcomes: OpenAI blows up and the economy tanks big time, or there is regulatory capture (which OpenAI's CFO clearly advocated for yesterday). Both outcomes suck btw it should be clear that I'm very bullish on AI as a science. I'm not bullish on the current underlying economics

this also explains the recent excess of grumpy tweets on my timeline btw

investing pre-seed is an eternally optimistic job, but it's difficult to be thrilled when the entire startup chain (from pre-seed rounds up to advocacy for regulatory capture a la OAI) is bloated with unsustainable excess I don't think anyone can argue the market is healthy atm

This might be one of my all time bangers ngl

I’m obv not surprised this happened (see original tweet). Amazed at how quickly though

“Backstop” yep

Yep

I'm just gonna say it, I think the market is oversaturated with VC scouts atm Feels like a pissing contest between firms more than anything else

btw everyone knows we have a data problem with training robots best way to solve this is by deploying robots into the real world (even if they suck for now), not keeping them in closed lab settings (which is what US companies are currently doing)

Yes, but say I have a $2B fund I toss $500k into a fringe pre-seed, end up being right, and return 1000x on my investment (extreme outlier); but only return 1/4 of my fund. The economics make no sense With huge AUM I make money by moving money (i.e. money on money investing)

That's to say, from their pov, I think OP is right. Someone who quits SF probably "lost" the tournament and would thus be less interesting to a megafund (not exclusive to Lightspeed)

Note that VCs are not altruists and have motives behind opinions they share (incl me) Lightspeed's business model (as a megafund) is to pick early winners in consensus startup verticals (i.e. expected highest value). SF *is* the best place to compete in the consensus tournament