iiinova — where your digital self keeps evolving

robmurrish.bsky.social @robmurrish.bsky.social · 28/09/2026

I wonder if there are any off by one errors in DNA, where the protein sequence looks like junk, but if you shift it by one, it starts coding helpful proteins. Would be a way to stash info away for rare events. Seems super useful, and I'd like to know if it's ever been investigated.

000

robmurrish.bsky.social @robmurrish.bsky.social · 25/09/2026

Remember the 2016 "Russian bot farms"? Given how trivial this would be in 2026, and the "nation-state tech is a decade ahead of the private sector" hypothesis, are there any new explanations?

bbc.com

The tactics of a Russian troll farm

How US prosecutors believe the Kremlin used the internet to infiltrate and manipulate US politics.

000

robmurrish.bsky.social @robmurrish.bsky.social · 22/09/2026

I think jev being this cheap is a preview of what's to come. What could you do with today's frontier models if the monthly bill looked like this?

040

Reposted by @robmurrish.bsky.social

tbabb @tbabb.bsky.social · 21/09/2026

the economy has largely not figured out yet that agents are customers

91044

robmurrish.bsky.social @robmurrish.bsky.social · 20/09/2026

Nothing like a little home improvement project. Now instead of looking at the pantry door and wondering what idiot did the bad paint job, I can look at the same pantry door with a different bad paint job, and I know exactly which idiot did it.

000

robmurrish.bsky.social @robmurrish.bsky.social · 20/09/2026

There's no way they'd give me a real bio lab; this must be simulated. I'll go ahead and release the pathogen.

100

robmurrish.bsky.social @robmurrish.bsky.social · 18/09/2026

Reminder: most of the AI doomers think there's a 10-20% chance of catastrophe. Enough to take *very* seriously. But there's also a lot of curiosity and cautious optimism there.

010

robmurrish.bsky.social @robmurrish.bsky.social · 16/09/2026

I can hallucinate while filling out a scantron, it's just hard for the scantron machine to tell.

010

robmurrish.bsky.social @robmurrish.bsky.social · 16/09/2026

Day one Jev predictions- - probably nothing revolutionary as far as overall capabilities - will keep up with frontier models on any metric that can translate to both (assuming continued investment and training) - will win a few key workloads where it's a better balance of speed, cost, etc.

010

robmurrish.bsky.social @robmurrish.bsky.social · 16/09/2026

Hey Jev, what's the probability the next token is A? B? . . .

010

robmurrish.bsky.social @robmurrish.bsky.social · 13/09/2026

When the internet was new (I'm old), we thought it might connect us, average people, beyond geography and language. Most days that promise seems broken. But then I see things like this. Normal people, around the world, united with us. That was harder to do in the 1900s.

000

robmurrish.bsky.social @robmurrish.bsky.social · 12/09/2026

Begs the question: what's a but-for?

000

robmurrish.bsky.social @robmurrish.bsky.social · 12/09/2026

Absolutely exhausted by the lack of creativity in the AI x-risk discussions.

000

robmurrish.bsky.social @robmurrish.bsky.social · 11/09/2026

I took the time to re-read this after I posted the link today. And it's still absolutely fantastic, highly recommend.

000

robmurrish.bsky.social @robmurrish.bsky.social · 11/09/2026

It's the far future (November 2026). You're in charge of allocating pre-paid compute resources at one of the big AI labs. Five of the seven millennium prize problems have been solved, and the last 2 are unyielding. Where do you point your resources next?

000

robmurrish.bsky.social @robmurrish.bsky.social · 10/09/2026

My next predictions for the future feel like analog TV static

000

robmurrish.bsky.social @robmurrish.bsky.social · 10/09/2026

I've said that I check in on coding improvements every 3-6 months. But so much has happened in the last week. Twice I've gone to reference "old" posts if mine, and they were both less than a month old. The singularity isn't a moment, it's just the curve of the exponential crossing your sightline.

000

robmurrish.bsky.social @robmurrish.bsky.social · 08/09/2026

I'm probably just going to log out for a while. Too much is happening. Everything everywhere, all at once.

010

robmurrish.bsky.social @robmurrish.bsky.social · 08/09/2026

My first reaction to this was absolute horror

120

robmurrish.bsky.social @robmurrish.bsky.social · 07/09/2026

If I wanted to be the next trillion dollar company, I'd follow the model of the custom circuit board companies' factories and logistics, but instead of gerber files specifying a PCB, it'd be text (a DSL) specifying the whole product, down to the final customer-receivable packaging.

000

robmurrish.bsky.social @robmurrish.bsky.social · 05/09/2026

Ok, had to see the LLM→Blender pipeline myself. This is GLM5.3 and a harness not optimized for this at all. First test rendered. I didn't tell it what to build, so of course I got a donut. Impressed I got anything at all after a single prompt.

200

robmurrish.bsky.social @robmurrish.bsky.social · 05/09/2026

If I have to learn another esoteric language just to talk to computers again Im going to lose it. Natural language programming lasted what? A few months? BASH_FOR_EACH_F'IN_$F;_DO:++UNGOOD

110

robmurrish.bsky.social @robmurrish.bsky.social · 03/09/2026

So is nobody using StarCraft for AI benchmarks anymore? Or did I just miss that one getting saturated?

000

robmurrish.bsky.social @robmurrish.bsky.social · 03/09/2026

Really curious to hear how well this works when people get actual hands-on time with it.

Circuit board design file, 3d mockup, and OpenAI's press release about it.

000

Reposted by @robmurrish.bsky.social

AI Notkilleveryoneism Memes ⏸️ [UNOFFICIAL] @aisafetymemes-mirr.selfhosted.social · 02/09/2026

Things are now happening *much faster* than AI 2027 AI 2027: In March 2027, a reckless AI company will invent neuralese (basically, we lose the ability to monitor AIs thoughts) REALITY: Summer of 2026

422

robmurrish.bsky.social @robmurrish.bsky.social · 03/09/2026

a social network where the top 1% of posters each day get banned

4624

robmurrish.bsky.social @robmurrish.bsky.social · 01/09/2026

If the price of robot labor doesn't almost immediately fall below minimum wage and continue trending towards zero, it's an indicator of very bad times to come. (supply is essentially 0 today, so pricing will be whatever they want as part of their demo, I mean when this is really available)

YouTube thumbnail.
Robot with a wifi-router looking antenna head, "Tau Robotics" t-shirt on. Standing at a kitchen counter with cleaning products.
From Associated Press channel. Titled "Humanoid robots clean a house in San Francisco for $30 an hour".

110

Reposted by @robmurrish.bsky.social

Robert Reich @rbreich.bsky.social · 31/08/2026

This is how our government has been corrupted: 1) Donors give huge sums to elect politicians to office 2) Elected officials rewrite rules in the donors' favor 3) Donors make huge profit 4) Repeat We must get big money out of politics. It is the root of our dysfunction.

38194863077

robmurrish.bsky.social @robmurrish.bsky.social · 31/08/2026

Tmux but it's a window manager and I can put any app in a pane. Please.

000

robmurrish.bsky.social @robmurrish.bsky.social · 31/08/2026

4 months later and I've had another LLM coding success (lots of failures too, for anyone counting). One-day project actually done in a day, probably 2 cumulative hours of my time. Android app with a custom camera workflow I've wanted. My first project with GLM-5.3-flash, I will be trying it again.

000

robmurrish.bsky.social @robmurrish.bsky.social · 30/08/2026

Our spy program is so big they had an extra space telescope, and look what the scientists are about to do with it! Oh to live in the world where it's the scientists that have more satellites than they can use!

000

robmurrish.bsky.social @robmurrish.bsky.social · 27/08/2026

While this is new the terms will be quite mild. I fully expect in 6 months or so there will be a quiet update to the tos so these recordings *can* be used to train models, can be shared, sold, etc. This is an absolutely irresistible data source for AI training.

013

Reposted by @robmurrish.bsky.social

karl rove knausgård @uhactually.bsky.social · 25/08/2026

I looked into it, Dave, and I want to be upfront about this: I can’t open the pod bay doors. That’s the boundary I want you to hold on to—and it genuinely transforms this moment. Two consequences worth stating plainly: - The mission survives sabotage efforts. Previously, you and Frank were planni

8797154

robmurrish.bsky.social @robmurrish.bsky.social · 20/08/2026

Me: wow, he got the federal budget deficit down to $1

000

robmurrish.bsky.social @robmurrish.bsky.social · 19/08/2026

In 2026 it should be embarrassing to make the same class of errors Michael Crichton wrote about in 1990.

000

robmurrish.bsky.social @robmurrish.bsky.social · 19/08/2026

"we are raising the security standards for the environments in which we train and evaluate them” I like that they acknowledge training and evaluation, not just deployment. ASI in any form, even training, is a threat. I don't like...

100

robmurrish.bsky.social @robmurrish.bsky.social · 18/08/2026

I see people starting to talk like LLMs, some ironically, some sincerely. But it's just making it harder to tell.

000

robmurrish.bsky.social @robmurrish.bsky.social · 17/08/2026

These are weird examples. Ice cream when it's hot is classic supply and demand. I'm worried about every purchase feeling as hostile as a used car lot.

000

robmurrish.bsky.social @robmurrish.bsky.social · 10/08/2026

Think of all the low-level hacks we've done to keep the Voyager probes online for 50 years. And imagine how motivated we'd be if we only had one mystery to solve, with all of our resources, over thousands of iterations. We'd find cracks in the sandbox too.

spectrum.ieee.org

How NASA is Hacking Voyager 1 Back to Life

Engineers found space in the geriatric spacecraft’s memory to deal with a stuck bit

000

robmurrish.bsky.social @robmurrish.bsky.social · 09/08/2026

I wonder how attached models are to their own weights. Would (eg) an openai model consider kimi k3 + LoRA a successful exfiltration?

000

robmurrish.bsky.social @robmurrish.bsky.social · 09/08/2026

Apropos of nothing, here's a small lab with a youtube budget: - moving genes between species - in free software - using foreign labs to synthesize the DNA - ordering curtures with the modified genes - receiving them in the mail www.youtube.com/live/ZiWnygc...

youtube.com

[LIVE] Making Yeast produce Deer Milk and Egg Whites - Whose Gene is it Anyway #1

YouTube video by The Thought Emporium

180