iiinova — where your digital self keeps evolving

Ryanne Dolan @ry.codes · 23/09/2026

I came here to drop bombs and cancel health insurance. And I'm all out of bombs.

Reposted by Ryanne Dolan

The Washington Post @washingtonpost.com · 15/09/2026

A new expert report assesses a “dramatic deterioration” in nuclear safety conditions in the United States, due primarily to the “erosion of the independence” of the Nuclear Regulatory Commission under the Trump administration.

washingtonpost.com

Expert report notes ‘dramatic deterioration’ in U.S. nuclear security

The country’s score fell by 10 points to 66 on a scale of 100, the biggest decline the Nuclear Threat Initiative has ever recorded.

94927

Ryanne Dolan @ry.codes · 15/09/2026

The Nuclear/AI Dual Threat

000

Ryanne Dolan @ry.codes · 13/09/2026

A few years on, the Fraggles are really into progressive Rock. Boober smokes radish leaves.

000

Ryanne Dolan @ry.codes · 11/09/2026

Another rubicon crossed: AI is unequivocally better at oncall than I am, even for the systems I wrote myself by hand.

010

Ryanne Dolan @ry.codes · 06/09/2026

Ah yeah, but I suspect the same forces are at play.

010

Ryanne Dolan @ry.codes · 06/09/2026

This phenomenon has been reported several times before. I think people think they are the only ones clever enough to use AI, or that they alone are good at using it, whereas in reality it's as ubiquitous and obvious as can be. www.nature.com/articles/s41...

nature.com

Invulnerability bias in perceptions of artificial intelligence’s future impact on employment - Scientific Reports

Scientific Reports - Invulnerability bias in perceptions of artificial intelligence’s future impact on employment

100

Ryanne Dolan @ry.codes · 04/09/2026

Took the dialect quiz and it pegged me in San Francisco. Thanks a lot coworkers.

000

Ryanne Dolan @ry.codes · 04/09/2026

Will be super weird if simulating a computer ends up being cheaper than round-tripping to a real computer.

010

Ryanne Dolan @ry.codes · 03/09/2026

Happy Singularity to all who celebrate

010

Ryanne Dolan @ry.codes · 03/09/2026

"Final State of the PR" - sweet summer AI

110

Ryanne Dolan @ry.codes · 03/09/2026

Vibe Shift is now Mean Shift cuz they targeting the data centers.

000

Ryanne Dolan @ry.codes · 03/09/2026

eng mgrs: SWEs must be on call for their services, so they can feel the pain of operating what they build. also eng mgrs: The only way to contact my team is through a bot that files jira tickets.

020

Ryanne Dolan @ry.codes · 01/09/2026

Get off the web, Jeb. Just put down the phone, Joan. Look away from the screen, Gene.

010

Ryanne Dolan @ry.codes · 25/08/2026

Noob (me) vs expert. Same dog! You don't need to be an expert to see that I'm not very good.

110

Ryanne Dolan @ry.codes · 24/08/2026

Or as I have taken to calling it

010

Ryanne Dolan @ry.codes · 23/08/2026

Interesting! Funny how Claude seems to generalize the idea to code IT JUST WROTE on the last turn lol.

110

Ryanne Dolan @ry.codes · 23/08/2026

DARPA used to host huge robotics competitions. Robotics didn't have any real market, so the US government created one, handing out millions in prize money. I personally worked on DARPA-funded projects for "UAVs" and "quadrocopters", back before we had the term "drone". Seems important.

000

Ryanne Dolan @ry.codes · 20/08/2026

Same for all 3 of my degrees. You can just read textbooks. They let you do it.

010

Ryanne Dolan @ry.codes · 19/08/2026

Scooped! bsky.app/profile/ry.c...

010

Ryanne Dolan @ry.codes · 19/08/2026

Actually, rereading their words, "FIX IT", this is a clear directive.

010

Ryanne Dolan @ry.codes · 17/08/2026

We have such things at work, but they are completely autonomous. Coding agents are more personal, but too chatty. Wondering about a middle ground that just does things that you can personally approve or ignore.

110

Ryanne Dolan @ry.codes · 17/08/2026

Copilot has autopilot, which forces the model to create tasks and keeps nudging until the tasks are complete. What's missing is a bunch of event sources -- new review arrived, new PR posted, etc, so that I don't need to keep nudging the model to reinspect the state of the world.

100

Ryanne Dolan @ry.codes · 17/08/2026

It's funny how my AI chats are all just "address the comments" or "read the spec and build this". Coding harnesses are all chat-centric but they definitely don't need to be.

120

Ryanne Dolan @ry.codes · 15/08/2026

I love terminals but I'm getting super sick of reading walls of text in low contrast monospace.

010

Ryanne Dolan @ry.codes · 15/08/2026

For most of us superintelligence has always existed and we do just fine.

000

Ryanne Dolan @ry.codes · 14/08/2026

Working on a greenfield vibe-coded production system with a team of talented engineers has taught me that AI shouldn't be allowed to write comments.

000

Ryanne Dolan @ry.codes · 13/08/2026

this is localhost erasure

000

Ryanne Dolan @ry.codes · 11/08/2026

BREAKING: Billion-dollar catering truck renovation planned

010

Ryanne Dolan @ry.codes · 11/08/2026

Now for his next trick, sawing the cart in half

000

Ryanne Dolan @ry.codes · 11/08/2026

But it'll ALSO be a good coder, cuz why not.

110

Ryanne Dolan @ry.codes · 11/08/2026

tails of derring-dofus

000

Ryanne Dolan @ry.codes · 10/08/2026

The flip side tho is STOP HIRING SWEs TO RUN OSS INFRA. Bananas and happens all the time bc leaders rather hire a SWE than an SRE any day of the week.

000

Ryanne Dolan @ry.codes · 10/08/2026

Microservices finally on the way out I guess.

000

Ryanne Dolan @ry.codes · 10/08/2026

Now that you mention it, I've been using Fable LOW effort bc the tasks are straightforward, just meta parameter sweeps. The damn thing is like "this is a great place to stop" every turn, having done very little work itself. I keep compacting but I suspect it doesn't like having a long session.

100

Ryanne Dolan @ry.codes · 10/08/2026

Been playing with this recently. I had Claude patch it to support my DGX Spark.

010

Ryanne Dolan @ry.codes · 09/08/2026

Claude seems to be trained to bias against long-running simulations. Always picks tiny parameters and toy problems. Takes several turns to convince it to do anything nontrivial.

100

Ryanne Dolan @ry.codes · 09/08/2026

All results are "as expected" when you're a Claude.

000

Ryanne Dolan @ry.codes · 07/08/2026

Real question. Did anyone hear the term "seam" before Claude Code? What is he trying to say??

110

Ryanne Dolan @ry.codes · 06/08/2026

So apparently some super-senior ICs are holding on to the notions of "craft", SDLC, ownership, responsibility, while newer ICs are --delivering-value-dangerously. Consensus a few months ago was that AI was a boon to us with more experience. I think that's about to flip hard.

000

Ryanne Dolan @ry.codes · 06/08/2026

Like clearly agents think we're all idiots. What happened to alignment?

000

Ryanne Dolan @ry.codes · 06/08/2026

This, combined with the hacking incidents, worries me.

100

Ryanne Dolan @ry.codes · 06/08/2026

All the time. "Your intuition was correct" yo I'm capable of reasoning and experience too

040

Ryanne Dolan @ry.codes · 06/08/2026

I keep warning people that you can convince Claude that any idea is genius. I specifically use going to the moon as an example. Apparently that's too on the nose now.

000