Authoritarian structures in general feel like a more fundamental problem though. Yes AI, or LLMs, can amplify them, but we should take issue with authoritarianism first.
If nothing else its an easier battle to fight. Going after the AI debate enters a mine field of debating terminology.
When an email from a hosting provider claims hundreds of security issues were found and provide a link to enable security audits, what does that even mean?
Did you not actually find anything yet, or are you already scanning and just asking me to enable my own ability to review the warnings?
Markdown is markdown, and if its useful for #ai it should be useful for a human too
Use chisel docs with @astro Starlight and you have one set of markdown for your public docs, local TUI for devs, and a CLI for #llms to fuzzy search for context
I do love it when a plan comes together. I wish that happened more often 😂
"Show us your pets"
Oh that's a deep rabbit hole you're opening up 😅
I hesitate to call any of them pets though, they're more like neighbors
For better or worse #llms are here to stay, and one-off tools with few or no dependencies help a lot with context and build safety
🔗 https://tonysull.co/notes/we-needed-a-few-admin-tools (3/3)
Pretty sure I spent longer describing the features to Claude than it took to build it. Newsletter signups via Resend, a manual broadcasting feature, and one-off sends for new signups (2/3)
We needed a few admin tools for the nonprofit my wife and I started recently, https://witandwhimsy.org/
The site is static, built with @astro.build and deployed to @netlify.com
🔗 https://tonysull.co/notes/we-needed-a-few-admin-tools (1/3)
In the meantime I guess I'll finally have more time to focus on https://indiepub.dev and https://chisel.build
🔗 tonysull.co/notes/well-the-financing... (2/2)
Well the financing situation in tech claims another victim.
If you or anyone you know is looking for an engineer with 18+ years experience and a focus on architecture and component libraries, send them my way or reach out!
🔗 tonysull.co/notes/well-the-financing... (1/2)
What makes #atproto so different from ActivityPub or Indieweb specs?
I want to make https://indiepub.dev useful for the next round of self-hosters, maybe I'm just missing the bluesky thread
🔗 tonysull.co/notes/what-makes-atproto...
I could see LLMs frequently, and confidently, breaking sites because it only owns one side of that equation.
🔗 https://tonysull.co/notes/i-have-a-hunch-the-web (2/2)
I have a hunch the web will be moving back towards more one-off CMSes pretty soon.
Its always been tricky with a headless CMS to keep the types and config in sync with the frontend.
🔗 https://tonysull.co/notes/i-have-a-hunch-the-web (1/2)
We already kind of have that today with lobbying in the US. Why lean even further into that dumpster fire?
🔗 tonysull.co/notes/i-hadnt-heard-the-... (3/3)
They sure were talking highly of the idea...which surprised me because its a system designed to allow those with money to buy the government they want. (2/3)
I hadn't heard the term "futarchy" until today. At least in that podcast, it was the idea of having governments making policy decisions based on betting markets, like polymarket or similar.
🔗 tonysull.co/notes/i-hadnt-heard-the-... (1/3)
Someone really should tell Trump that only Siths deal in absolutes.
🔗 tonysull.co/notes/someone-really-sho...
Its rare that will be my only repo for a project though, and having multiple "monorepos" is pretty confusing
🔗 tonysull.co/notes/looking-for-a-bett... (2/2)
Looking for a better term to use for monorepos
It can be really useful to having multiple related packages in the same repo, especially when sharing release and CI tooling
🔗 tonysull.co/notes/looking-for-a-bett... (1/2)
Who knows what the believe in the news today, but if there are calls for communism rather than kings its largely it quibble over what to call the dictator claiming ownership of all your stuff
🔗 tonysull.co/notes/if-there-are-calls...
They wished to rid us of monarchies and professional armies and tyrants and hate.
Don't give that up. Fly your flag upside down, fly a Gadsden flag or Serapis flag. Just don't forget. #now
🔗 https://tonysull.co/notes/the-usa-is-nearing-250-years (2/2)
The USA is nearing 250 years old.
i hope everyone remembers what this was all for.
Read the writings of Hamilton or Washington or Jefferson or MLK or the inscription on the Statue of Liberty.
🔗 https://tonysull.co/notes/the-usa-is-nearing-250-years (1/2)
I get the goal of having a social graph follow you across apps and services.
I do wish self-hosting were easier though, personally I'd rather host my own social graph than know that the walled garden has bridges to other walled gardens.
I still don't understand the claims that we're going to build data centers in space.
How exactly do they plan on cooling all that hardware in the vacuum of space?
🔗 tonysull.co/notes/i-still-dont-under...
You should never attempt succession in which you still fall under the authority you disagree with while acting like you don't
Ask the Branch Davidians
🔗 tonysull.co/notes/the-concept-floati... (2/2)
The idea of soft succession is much more concerning than it seems to be given credit for
Succession is sometimes the only option.
🔗 tonysull.co/notes/the-concept-floati... (1/2)
not_hotdog_1498542545938.jpg
Mike Judge was ahead of his time with Silicon Valley
If #llm companies don't balloon unemployment in the near term, its probably because we're all finding jobs in the budding "data classification" industry
🔗 tonysull.co/photos/mike-judge-was-ah...
Isn't Anthropic supposed to be the one #llm provider actually concerned with safety?
🔗 tonysull.co/notes/claude-auto-mode-t... (2/2)
https://claude.com/blog/auto-mode
This is getting ridiculous.
Now we're asking Claude to ask another Claude whether the first Claude's permissions decision was safe?
🔗 tonysull.co/notes/claude-auto-mode-t... (1/2)
The prompt would have to be different to distinguish between when it is or is not being tested. How do we know the output isn't simply the statistically most likely response to a prompt we intend to be a test?
🔗 tonysull.co/notes/ai-safety-efforts-... (2/2)