
early stage product: 🪵 collect tinder, start fire before freeze later stage product: 🔥 keep fire alive, grow fire, defend fire very different h/t @christineluc

building findableapp.com also: - 🏝️ smol: agent - 🌌 Atlas: get most out of llms - 🐈⬛ Chess Cats: puzzles - 🍓 Jam: open source Clubhouse (@jam_systems)

early stage product: 🪵 collect tinder, start fire before freeze later stage product: 🔥 keep fire alive, grow fire, defend fire very different h/t @christineluc

P(@levelsio starts a hosting provider) went from 2% to 22%
How do I tell the @Hetzner_Online and @digitalocean CEOs to make new VPS servers: - have a default firewall that closes all their ports - by default have Tailscale installed as the only way to access it via SSH - by default have some web tunnel like Cloudflare Tunnel to serve web traffic safely That way VPS servers could get a better reputation instantly by being safe by default It's crazy in 2026 they let you create a VPS that by default is open to the entire internet!

One interesting bit of AI discourse: it's pretty broadly agreed that agents aren't great at software design. But...they're still good at getting tasks done. So, maybe software design doesn't matter? I think there's a lot of complexity baked into this question!


this
Skills should be made reactively when the agent struggles with something Not proactively And always ask the agent how it could've done better; often a skill isn't the answer Finally, ask the agent to eval your skill

in Poland "Suppengrün" (soup greens?) are called Włoszczyzna "Italian stuff" en.wikipedia.org/wiki/Mirepo…

Alex Martelli on Ruby, Python and Pasta in 2003 groups.google.com/g/comp.lan…

w/ @JensenHuang resigning how is $NVDA only down 6%?

openai decisions api (and clef) is available by now developers.cloudflare.com/ai…
luna decision api still not available via cloudflare unified billing

agents are the biggest 'upgrade' for operating system and computer usefulness in a long long long time (since web+browsers?) no need to upgrade to to the latest hardware even downgrading is a viable move I'd rather work on an intel macbook w/ agents than on an apple silicon macbook without unimaginable just a while ago

still amazed by the fact that a language model can drive a computer for multiple steps and achieve a goal I get how it works but I'm still amazed

programming language design has more leeway now pre llms breaking backward compatibility meant that you make it harder to build on top of the language (or stdlib etc) now changing things is still is not free but agents smooth over these changes that said: still not easy to come up with a new language that has a chance of seeing widespread adoption in some ways that is now even more difficult because fewer potential adopters will care about syntax, abstractions, api design or how 'easy' it is to do something

> I like Max Bittker's framing that models are very good at counterfeiting and the level of expertise it takes to distinguish a counterfeit from the real thing has been increasing as models improve via danluu.com/ambitious/

from Fable and Mythos 5.2 release announcement why omit GPT 6.1 Sol? just makes it more annoying to your target audience + lose trust in what way is this worth it? anthropic.com/claude-fable-a…

new twitter desktop web ui feels more 'reliable'

superpower in turbulent times: ability to hold multiple potentially conflicting thoughts at the same time