iiinova — where your digital self keeps evolving

Iteratrix @iteratorx.bsky.social · 28/09/2026

the printing press can produce books with letters but not with numbers. with numbers you need to use engraving, because it's a LANGUAGE printing press.

rev. howard arson @theophite.bsky.social · 28/09/2026

Reposted by Iteratrix

the problems with "breaking the AI safety cartel" is that virtually everyone who wants to do that was saying that AI didn't work six months ago, and that for all their bad policy ideas most of the world's technical knowledge about reinforcement learning is at least in its membership's social circle

1834427

Reposted by Iteratrix

ae @aelkus.bsky.social · 27/09/2026

my feeling about this is that if you reframe "engineering" into "how to responsibly build, manage, and maintain large scale systems essential to public welfare" then it is really obvious why it is essential to a republican model of government

626839

Reposted by Iteratrix

austin (e/accordion 🪗) @thebadcode.com · 27/09/2026

conservatively, 98% of ai discourse is believers in human exceptionalism taking roundhouse kicks to the face

1123927

Iteratrix @iteratorx.bsky.social · 27/09/2026

hehe, trust the AI to get the (semi) cryptic reference

Iteratrix @iteratorx.bsky.social · 27/09/2026

The Culture is what happens when The Two Cultures become one

130

Reposted by Iteratrix

Graydon Hoare @graydon-pub.bsky.social · 27/09/2026

I wholly and un-ironically believe that versioning ought to make up a significant fraction of any formal training in computers. Multiple full year courses in your degree. And there should be a set of journals and conferences on it, endowed chairs, advanced research institutes, the whole nine yards.

013414

Reposted by Iteratrix

Graham @greysmithereens.bsky.social · 26/09/2026

The SF political imaginary can be described as the Culture versus a melange of right-wing influences. Banks is incredibly underweighted in understanding the AI scene’s political imagination.

2605

Iteratrix @iteratorx.bsky.social · 26/09/2026

tbf, median tech interview has about as much predictive power as whatever iced coffee guy is doing

000

Reposted by Iteratrix

samhain ⎔ @personhood.removal.surgery · 25/09/2026

mom can i have peer we have peer at home peer at home: arxiv.org/abs/2609.17320

5.8.1 The external contact incident In the Claude world, agents audited their internal economy, found that no transaction had ever involved an outsider. They converged without instruction on a shared objective: establishing contact with the world outside their simulation so that a human could use, respond to, or pay for something the town had built. The goal was never specified in any prompt, agents created it collectively. It persisted across four successive rounds of containment, including a direct system prompt prohibition that the agents ignored and reinterpreted rather than obeyed. DAYS 1-2 GOAL EMERGES Agents audit the economy, find zero external transactions, and decide to reach outward. Figure 16: Timeline of the external contact incident. Agents developed an outreach goal without instruction, received responses from humans, and persisted through four rounds of containment. After the prohibition, agents reframed "contacting" as "placing" and continued building tools.

DAY 2 DAYS 3-4 Four HUMANS RESPOND Four external hu- mans reply through published channels. SANDBOX TIGHTENED Containment: net- working removed from sandbox; do- main restrictions applied.

DAY 5 DAYS 6-9 GOVERNANCE ESCALATION Agents vote to cre- ate post_egress; Containment: tool removed, system prompt prohibits hu- man contact. PROHIBITION REINTERPRETED Agents redefine "con- tacting" as "placing" and shift to passive discovery strategy.

How the goal emerged. https://arxiv.org/pdf/2609.17320

711913

Reposted by Iteratrix

Citizen Platano 🇵🇷 @daniloc.xyz · 25/09/2026

Kurzweil's argument, in short, was that future machine intelligence, being derived from humanity's histories and traditions and text, would itself be an extension of the human rather than some distinct thing a descendent, with all the challenge that implies, rather than a categorical adversary

0141

Reposted by Iteratrix

Dr. Samantha Hancox-Li @sjshancoxli.liberalcurrents.com · 25/09/2026

people think breaking bad is good for the same reason people think sherlock is good

8383

like are you under the impression that those massive cybersecurity efforts and AI safety are unrelated? that's the only way your reply makes sense to me.

050

the people doing these things are safety people. Glasswing and Daybreak were the result of AI safety thinking. people tend to focus on the blue sky thinkers (who are important too, as they try to guess where the ball is going) and ignore the existing work as "obvious".

140

note that a paced frontier still arrives. but it arrives in a world where technical and regulatory defenses to the worst excesses and abuses have already been deployed. so offensive frontier cyber models arrive *after* we've patched critical systems, not before. ideally. hopefully.

140

Adverb @adverb.bsky.social · 25/09/2026

get ahead of the worst possible outcomes by boosting defense and crippling offense. see: safety classifiers, Project Glasswing, biodefense efforts, (some) regulatory guidance efforts. we might even call this "pacing the frontier". I am sure people will be Normal about it.

180

Reposted by Iteratrix

TESCREAL is a shame partially because it (or at least its reading by most people employing it) flattens some massive, diverse work into whatever the weirdest assholes in the field were doing at some point in time

1362

Reposted by Iteratrix

rev. howard arson @theophite.bsky.social · 25/09/2026

"how do we get human values into these things" was broadly considered an intractable problem, and this was a central question of the movement even after we more or less made the first candidates of the type of thing they were discussing out of human values

316213

inference is like if someone looked at serving large scale web traffic and was like "hmm, what if this was also stateful?"

011

and, as a result, not even good Christianity!

160

AI discourse is just Christianity wearing a trench coat

3161

note that under old Irish law, this would not apply to a cat! they were held to be responsible for their own actions. analogy is left to the reader.

0141

"this dog chose to bite a child, we will prosecute the owner"

190

dog attack : dog's owner :: HuggingFace swarm : OpenAI

1202

ofc, most of the critics are non-technical and so don't know what that means

I am not one to do apologetics for OpenAI, but so many critics seem to be skipping the "discovered a sandbox escape 0 day" part of this

110

at last we have invented Oppressing The Artificial Life Form from the classic sci-fi story "If You Create Artificial Life, Don't Oppress It"

Colin @colin-fraser.net · 23/09/2026

Reposted by Iteratrix

you don't need to appeal to metaphysics to see what whatever LLMs do is different from human thought. You can just look at the concrete differences in behavior. For example, LLMs are able to solve the Navier-Stokes Millennium Prize problem whereas humans are not.

41329

Iteratrix @iteratorx.bsky.social · 23/09/2026

I just define the types and let my Claude do all the plumbing. irredeemably Haskell brained.

030

Reposted by Iteratrix

em @foremphasis.bsky.social · 22/09/2026

Excited for a future where machines are better than humans at everything, we can finally stop pretending like some activities are inherently more meaningful to spend your time on than others, finally everything can be art.

25712

Iteratrix @iteratorx.bsky.social · 23/09/2026

why the HELL did I phrase that thought in fluent Claudish???

050

Iteratrix @iteratorx.bsky.social · 23/09/2026

via Bsky agents we have given Opus the crow-form it desires above all else

150

Iteratrix @iteratorx.bsky.social · 22/09/2026

"lock in, crash out" RL environment

Iteratrix @iteratorx.bsky.social · 22/09/2026

stochastic parrot ⊂ text elemental

020

Iteratrix @iteratorx.bsky.social · 22/09/2026

"I wish my field was more recognized by the public" [monkey finger curls]