Welcome to Edition 7 of Superfluid: Flow State where I curate the smartest takes on startups, AI, and capital allocation from the 75+ articles I read every week. I’ll give you the key insight from five of the best pieces that I consumed over the last week that you can immediately apply to your business (in <2 minutes).
Feel free to hit reply and let me know what you thought of this week’s article. I respond to every email.
Today’s word count: 579
Read time: 2 min 27sec.
Who decides what AI can do? by Byrne Hobart
Byrne’s framing is really sharp in this article. The question of AI governance isn’t necessarily about which regulator gets to write the rules, but about who has the legitimacy and technical competence to make those calls in the first place.
Governments are reactive by nature, labs are incentivised to move fast, and the public is largely outside the room where decisions get made.
Situational Awareness: The 25 year old taking on Wall St by Harry Wruck
Leopold is such an interesting person in that he is incredibly young, yet very accomplished. I first came across him on Dwarkesh’s podcast here a year ago, and it’s interesting to see his AI bets inside Situational Awareness. Spoiler: it’s heavy on electricity generation for data centres and semiconductor supply chains, true infrastructure investing for AGI. It’ll be interesting to see whether his second-order effects thinking leads to outsized returns.
Venture Capital Doesn’t Exist by
Something I’ve said for a while is that the current era of VC isn’t really traditional venture capital. Capital allocators have sought to move higher up the risk curve, and we’re now at a stage where growth-equity sized rounds are happening at “Seed.” Kyle articulates this well. The risk/return profile that justified the VC model in the first place has been fundamentally disrupted by the sheer volume of capital chasing a relatively fixed number of breakout companies.
The market consensus has settled on bigger rounds being a competitive necessity if your rivals are raising large, you have to as well. But my view is that the best returns are going to come from people who see things differently: those who aren’t reliant on the VC conveyor belt and can either make alternative capital allocation models work that defy the power law, or find more flexible funding structures suited to businesses with distinct risk profiles. Consensus thinking rarely generates outsized outcomes.
This was pretty inevitable given the impact AI has had on people’s livelihoods (e.g., higher competition for jobs, lower job security). Beyond all the doomerist chat on AI safety, alignment etc, the real impact is being felt by broader society and these voices will only get louder following the high profile layoffs at Atlassian and Block.
Whilst Rex puts forward his thoughts on fixing AI’s PR image, I think the most interesting questions to ponder are:
-
Is the backlash against AI actually a proxy for deeper anxieties about inequality? This has certainly been brewing for a long period of time.
-
Can AI companies genuinely rebuild public trust, or is the damage structural in that it is baked into the way these systems were built and deployed from the start?
The AI Signal to Noise Curve by
It’s so easy to always feel like you’re drowning in the latest developments in AI. Whether its new products, features etc, the overload is very very real. This is a framework shared by Casey that I’ve intuitively followed since my time in Crypto when it feel pretty similar to AI now.
Whilst I’m definitely up to date with what’s happening in AI, I don’t always experiment with things that I don’t necessarily think will give me outsized leverage on the things that I care about.
It’s easy to sink hours and days in optimising your multi-agent framework to doing something, but that doesn’t really matter if that complexity gets abstracted away in a months time.
Forward this to a founder who needs better signal.
Thanks for reading! I’d love to hear what you thought about the format. Feel free to hit reply and let me know, I respond to every email.
Abhi