iiinova — where your digital self keeps evolving

Reposted by Sean Scott

vegavandal.bsky.social @vegavandal.bsky.social · 27/09/2026

“Why does OAI keep unleashing dangerous models on the world with zero cybersecurity guardrails?” is still a better question than “is the code thinking”.

04014

Reposted by Sean Scott

Olivia Guest · Ολίβια Γκεστ @olivia.science · 09/08/2026

Useful generally is Ryle's distinction between knowledge-that vs knowledge-how. This allows one to notice that knowing how to ride a bike is unachievable through knowing that it has two wheels etc. No number of books can ever be read such that then one knows how to ride a bike. It's not possible.

25710

Reposted by Sean Scott

Anthony Moser @anthonymoser.com · 26/01/2026

"ai" is not a genie that can't be put back in a bottle, it's a pillar of smoke that dissipates unless you keep burning money

5507167

Sean Scott @oraclesean.com · 27/09/2026

Feminize a band name Deva

010

Sean Scott @oraclesean.com · 26/09/2026

Seeing multiple stations with $10+/gallon diesel in Los Angeles today.

000

Sean Scott @oraclesean.com · 26/09/2026

Not rogue activity. Design failure. These are systems engineered by people not working as intended. If this was a building we'd condemned it. If this was a single 737, Boeing would ground the entire fleet. If this was a nuclear reactor, we'd evacuate a perimeter & fill the crater with cement.

073

Reposted by Sean Scott

Caræsten 🇵🇸 @cara.city · 24/09/2026

guys please stop falling for this, this is not “rogue AI” it’s “environments with lax cybersecurity controls”

431651428

Reposted by Sean Scott

Tressie McMillan Cottom @tressiemcphd.bsky.social · 25/09/2026

Please try to reclaim your precious precious attention, at all costs. Do puzzles. Take walks without a phone. Dear god, try to read, almost anything that isn’t a manifesto will really do. You are going to need your brain.

52519654

Reposted by Sean Scott

Oregon 🕎🎲 @oregonthedm.bsky.social · 25/09/2026

Most riveting thing I've seen all week

383112783050

Sean Scott @oraclesean.com · 25/09/2026

There's scientific proof that eating ice cream actually produces weight LOSS, not gain. A thread.

110

Reposted by Sean Scott

Aaron Helton @aaronhelton.com · 25/09/2026

"Oh, but LLMs can do that!" you say, breathlessly. But they can't. No matter how truth shaped their answers, that truth shape is a simulacrum. It's built on a high probability model of correctness, but as we've seen this model can be hijacked pretty easily. LLMs have no relationship to truth.

152

Sean Scott @oraclesean.com · 24/09/2026

I for one know where I live. I know every address/phone number I've had since age 3. I've also committed passages of my favorite literature & poetry to memory & recite it to myself to help me sleep. Not sure how Mr Huang sleeps. Maybe his intellectual dexterity debt washes his conscience clean.

000

Reposted by Sean Scott

Dr. Damien P. Williams, dread portent down from a mountain cave @wolven.blacksky.app · 24/09/2026

If we let the current crop of "AI" people stay in charge, we are not gonna make it; not because of "superintelligent AI," but because of what the MIT study terms "cognitive surrender": offloading and deterioration of critical skills. And it increasingly feels like that's exactly what they want.

931078

Reposted by Sean Scott

Ryan North @ryannorth.ca · 23/09/2026

Pompeii is a UNESCO world heritage site not because it was remarkable in its prime, but because it survived to our era untouched: one moment, preserved for all time. But future generations won't know what OUR world was like unless we do the same. This is why we MUST bury Toronto with a volcano

Picture of the streets of Pompeii, from https://commons.wikimedia.org/wiki/File:Pompei_-_panoramio_(26).jpg. Taken by Tanya Dedyukhina, used under CC

11867551616

Reposted by Sean Scott

AP Stylebook @apstylebook.com · 23/09/2026

Artificial intelligence systems do not think, feel, want or understand. Avoid language that gives them human characteristics.

apnews.com

https://apnews.com/article/openai-safety-ai-framework-089e75b95bc935af092da7b79d92706d

5542231619

Sean Scott @oraclesean.com · 24/09/2026

Putting "Never burned popcorn in the office microwave" on my CV.

011

Sean Scott @oraclesean.com · 23/09/2026

Nested threads but if there was ever a case of court proceedings containing foreshadowing, this is it. The attorney here is either doing suicide-by-judge or irrepressibly stupid.

020

Sean Scott @oraclesean.com · 23/09/2026

What could possibly go wrong?

020

Reposted by Sean Scott

Acyn @acyn.bsky.social · 22/09/2026

Child: This is my first time meeting a mayor. Mamdani: Wow, have you ever met a prime minister? Child: No Norway PM Jonas Gahr Støre: *walks over*

775285744855

Reposted by Sean Scott

Alastair Somerville @acuity.design · 22/09/2026

Value Capture as described by C. Thi Nguyen. The expert loses their expertise and thus the institution loses its ability to see AI as slop

• Your values are rich and subtle - or developing that way.
• You enter some social (typically institutional) setting that offers you simplified, often quantified renditions of your values.
• The simplified versions take over.

13612

Sean Scott @oraclesean.com · 22/09/2026

This explains why I have seventeen 1990s-era medical textbooks on neurosurgery, neurology, critical care nursing and radiation oncology.

100

Reposted by Sean Scott

ProPublica @propublica.org · 22/09/2026

Sen. Susan Collins helped steer millions in federal defense contracts to Navatek, whose CEO illegally funneled thousands of dollars to her campaign. The FBI sought to launch a probe and was eyeing Collins and other lawmakers. Then Trump gutted the unit in charge. THREAD 🧵

410113446033

Reposted by Sean Scott

Mother Jones @motherjones.com · 22/09/2026

NEW: OpenAI took 20 million ChatGPT responses and had one of its experts search for chunks of text—specifically chunks of 25 words in a row- that were identical between ChatGPT’s responses and a news article. They found more than 400,000 of these 25-word overlaps between ChatGPT and our articles.

motherjones.com

AI exec: We may have pulled off “the largest theft of labor in human history.”

Documents from our lawsuit reveal how tech leaders talk when they think no one is listening.

12395192

Reposted by Sean Scott

Microplastics Sommelier @leastactionhero.bsky.social · 21/09/2026

“It’s 12 million metric tons of potash, Michael. Where could we get it, Belarus?”

19014

Reposted by Sean Scott

Mariana Z @mariana057.bsky.social · 21/09/2026

My wife just informed me that I own every board game except one. I had no Clue.

942136185

Sean Scott @oraclesean.com · 21/09/2026

TRANSPLANT (v) - replacing a thing that's stopped working thing with something that does. TRANSPLANT (n) - a working thing that replaced something that stopped working.

000

Sean Scott @oraclesean.com · 21/09/2026

Hey babe, a new season of To Catch a Predator just dropped.

020

Sean Scott @oraclesean.com · 21/09/2026

As someone who grew up in the 70s-89s, maybe I can shed some light on the issue. (This is recollection not research, so grain of salt here.) In the 70s we knew there were gay men. We knew who they were. It was no secret. But we were polite about things and didn't talk about it so it was ok.

200

Sean Scott @oraclesean.com · 20/09/2026

The existence of Real Madrid implies there's an Artificial Madrid.

210

Sean Scott @oraclesean.com · 20/09/2026

The Chick-fil-A chicken-waffle sandwich commercial is the whitest thing ever created. Blinding. The opposite of Vantablack. An albedo of 1,000,000. Do not watch without welding goggles. m.youtube.com/watch?v=ORdb...

m.youtube.com

Chick-fil-A Chicken & Waffles Sandwich Commercial #2 (2026)

YouTube video by Commercial Archivist

010

Sean Scott @oraclesean.com · 20/09/2026

The new Austin Powers film is awful.

120

Reposted by Sean Scott

Dr. Damien P. Williams, dread portent down from a mountain cave @wolven.blacksky.app · 20/09/2026

I say again, any "pause/pacing/contemplative silence where AI CEOs think about what they've done and promise to be very good boys from now on" that doesn't result in ACTUAL enforcement mechanisms to hold them Several-Hundred-Billion-Dollars accountable any time their product fucks up is nigh useless

Screenshot of posts from me:

Chris, they trained a system to hack & got scared when it hacked. This is on human beings not being able to logically think through what they've built & what they then trained & told it to do; they'll never "solve alignment," because they can't get what human values they're trying to "align" "Al" To.

Hold the HUMANS accountable for what their systems do. The humans' companies and individual humans themselves.

7428133

Reposted by Sean Scott

Steven (with a PH) @sjksalisbury.bsky.social · 20/09/2026

Christening a baby is nothing like christening a ship. I know that now. Once again I can only apologise.

159191003840

Sean Scott @oraclesean.com · 20/09/2026

"Why are you trying so hard to touch the hot stove? Has no one yet told you how painful it is to be burned?" is rarely couched in politeness, nor should it be.

100

Reposted by Sean Scott

Max! @littlepolygon.bsky.social · 19/09/2026

Real pet peeve for podcasters saying "well programmers tell me LLMs are useful for coding" because it fundamentally misunderstands what "coding" is. It's the result of a process of problem-solving, and when one-shotted produces artifacts without solutions.

29311

Reposted by Sean Scott

Problem Machine @problemmachine.bsky.social · 20/09/2026

This is getting a bit of attention so I'd like to expand on it a bit. What the compiler offers to the programmer is a way to a) describe their mental model and b) reify it into a behavior for the executing machine. The process of programming is the process of unambiguously describing this model.

142

Reposted by Sean Scott

Problem Machine @problemmachine.bsky.social · 19/09/2026

Quoting myself from a day ago - "The problem is thinking that the product of a programmer is code. It isn't, really, the code is just the final stage. The actual valuable thing the programmer produces is the underlying mental model. Without that, the code is just tech debt"

524373

Sean Scott @oraclesean.com · 20/09/2026

One of these things is not like the others.

A photo of 12 cars in a parking lot. 12 are Porsches. One is a white Subaru.

021

Reposted by Sean Scott

Faine Greenwood @faineg.com · 18/09/2026

It becomes a LOT easier to argue to the general public that we shouldn't let the AI industry regulate itself when you can ALSO produce clear evidence that much of the AI industry appears to be steered by genuinely crazy weirdos with anti-social and anti-human ideas.

659687

Sean Scott @oraclesean.com · 18/09/2026

Hear me out... Electric sharks.

020

Reposted by Sean Scott

Kathleen Bachynski @bachynski.bsky.social · 18/09/2026

“On Thursday, scientists reported that they had found another gene-editing defense in nature, this one in viruses… This system, called VIPR, appears to be more than four billion years old, and it seems be the evolutionary ancestor of CRISPR… It may prove to be the more powerful gene-editing tool”