software needs a soul again.
the real scarcity is meaning. and meaning cannot be reduced to a metric.
everyone talks about moats, but too often what they mean is features.
startups feel boring now because too many people are building for the startup ecosystem instead of the world.
for investors. for product hunt. for YC. for the screenshot. for the MRR post.
you can feel when something was built by people who actually care that it exists.
you can feel their heart.
people never had a moat. they had everyone else believing they had one. what a beautiful time to be alive. what scares those whose ego depends on the old order is that outsiders are starting to realise the walls were never real. plato’s cave but the shadows have cap tables
i find it hilarious how out of touch the average AI person on tech X is with how normal people feel about AI
tech x: “multi-agent orchestration is about to replace the entire software stack”
normal person: “why the fuck is there an AI button on my phone”
why do they keep throwing shit at each other?
you can win the AI race technologically and still completely misunderstand the human being standing at the finish line.
congrats btw @finkd@alexandr_wang for muse. normal people actually use it.
the current meta of software engineering might be accidentally optimised for ADHD.
rapid feedback loops.
constant novelty.
parallel agents.
jumping between abstraction levels.
externalising memory into tools.
turning vague ideas into something tangible in minutes.
automating the boring parts and keeping the interesting decisions.
the old workflow rewarded sitting still and manually grinding through implementation for hours.
the new one increasingly rewards:
notice → ask → build → inspect → change → repeat.
genuinely feels like the industry accidentally changed the game settings.
@droovg@cosign finally i can empirically quantify how hard everyone has been ghosting me. perhaps i’m too weird or my mutual do not believe in me at all
I’m looking to relocate to the US.
Building a startup in AI infra/model engineering, currently going through Imperial college’s incubator.
If you’re building, investing, or know where I should be looking?
@finkd The interesting shift is that intelligence is becoming easier to reproduce while distribution remains brutally hard to reproduce. Meta already has the latter. Enterprise AI is basically turning that distribution advantage into infrastructure.
@paulg The founder has usually accumulated far more emotional attachment to the name than the market has accumulated recognition of it. That asymmetry makes renaming feel much more expensive than it actually is. At that point, attachment becomes a failure mode.
@sinethxyz The unpredictability is doing more of the work here than the content itself, I think. A feed that gave you the exact same thing every time would get boring fast. It's not knowing whether this check is the one that pays off that keeps people pulling the lever.
whatever system we have, can it please be competent, rational, technologically capable, and not unnecessarily fuck with people? i just think incompetence should be illegal.
98% Aquinas. 93% Weber. 90% Karp.
the test thinks i’m going to spray fentanyl on my enemies with drones powered by Palantir Gotham when really i just like Lee Kuan Yew. a walking contradiction.