seth

AI driven dev work be like

I really misunderstood the strength of lexicons for decentralizing power in the Atmosphere. Atproto isn't an ecosystem, it's a foundation for creating parallel ecosystems.

"By letting lower court rulings stand, the Court effectively solidified a 'Human Authorship' requirement."

"If the code is truly a 'new' work created by a machine, it might technically be in the public domain the moment it’s generated [...]."

Kind of wild

All that said, if I’m being honest, I don’t know that I have the time to dedicate to seeing this project through. Just got curious and spewed some ideas to Claude 😅

Feel free to take and run with this if you’d like!

Would be curious how representative your approach is overall. Internally at Zapier, inline comments are preferred where feasible.

Our internal review agent often leaves suggestion comments that can be directly applied from the UI even, which is a very nice touch.

Because phones are better? 🙃

Another hard part is owning and operating the software, being responsible for keeping it secure and reliable for its users.

AI helps there too but it’s still a different ball game than spitting out code for a prototype that never goes to production.

Personally I think it’s less about taste and more about exploring the design space to understand second/third order effects of the things being built.

Not all projects need that level of deep design thinking. But many do, and that’s one of the non-coding “hard parts” I generally refer to.

True! There are pieces that need more tending and maintenance than others though and that is where buy can still be good.

Lower level primitives that you can build your custom use cases on top of, as opposed to heavy all-in-one solutions.

Not quite ready to launch, but if you're a decentralized identity and/or AT proto nerd, you might want to take an early peek at this freeq.at

A section of the OpenAI Symphony readme that says “tell your coding agent to build symphony in a programming language of your choice” with a link to a detailed spec

We have reached a moment where instead of releasing software you simply release the detailed spec for software and tell people to prompt their agent to build it themselves

From the README of OpenAI’s new Symphony orchestrator: github.com/openai/symph...

If your job was to know when to hit the semicolon key on the keyboard, I'm sad to say that's not an employable skill anymore.

If you shape ideas into tools and art via code, it's really so much better than ever now, and I don't see that changing soon.

Oh interesting, people who don’t know how to build software are getting mad at my post about building software. Cute.

Let me be clear, over the next year, the job of software engineer will shift dramatically to no longer have typing syntax into an editor as its primary time sink.

Somebody, apparently 😩

There's a better security story with MCP as well. I say this as someone who doesn't love MCP, but is designing a background agents platform that needs to meet some solid security requirements.

bsky.app/profile/seth...

The idea is that it can be posted to an actual PR in github/gitlab/etc, so those headings with filenames and line numbers get mapped to inline comments on the PR.

The more I thought about it though, the more it felt like this markdown version should just be an optional rendering of the data model.

Also not all tools you'd want to integrate with support short-lived or tightly scoped credentials. MCP lets you put in place whatever agent-specific controls into those systems you could want.

I don't love MCP, but it has undoubtedly the best security and auditing story right now.

Yes. The alternative being CLIs and other tools running in the same agent sandbox, which means opening up network egress, and potentially having secrets in the sandbox as well.

BookHive | There Is No Antimemetics Division See There Is No Antimemetics Division by qntm on BookHive, a Goodreads alternative built on Blue Sky

Finished There Is No Antimemetics Division! So good, such a wild ride. Feeling like a second read through is warranted just to turn the situations over in my mind a few more times.

I need more like this. Open to suggestions!

bookhive.buzz/books/bk_60v...

You couldn't even eval this given the incredibly high variance of agent context and soul documents its expected to operate in.

Truly nothing but thoughts and prayers 🤞🙏

This is the way.

I like how in 2026 a common security paradigm is writing a strongly worded letter to the guy in your computer

In retrospect, this feels like a darkly obvious attractor for a network society. If we're all neurons in a network, then the network will want to condition us to fire reliably when poked. Gambling is a crude conditioning mechanism.

Musk, worth $829 billion, owns X.

Bezos, worth $234 billion, owns The Washington Post & Twitch.

Zuckerberg, worth $231 billion, owns Facebook, Instagram & WhatsApp.

And now Larry Ellison, worth $202 billion, is about to control CNN, CBS, TikTok & HBO.

Yes, this is oligarchy.

We are working on a Discovery page on Margin where you can find new reads on the Atmosphere using @standard.site 😉

skyreader/cli at main · disnet/skyreader An RSS reader for the AT Protocol. Contribute to disnet/skyreader development by creating an account on GitHub.

Skyreader now supports RSS reading from the command line!

npm install -g skyreader

then you can do cool things like:

skyreader feeds --all --json | jq '.feeds[].items[].title'

or

claude -p "summarize today's feeds using skyreader"

github.com/disnet/skyre...

Was not familiar with MOQT before this but this is really cool. Applying it to atproto is galaxy brain.

my text editor should have a stats bar that tells me how long I've been thinking for and how many tokens I've emitted

13 years later and the story of Aaron Swartz still fills me with anger and sadness