Oliver Rochford
cyberfuturists.com
Oliver Rochford
@cyberfuturists.com
Cybersecurity Futurist, Neo-generalist, Researcher, Skeptic, X-Gartner Analyst. #Cybersecurity & #Cyberpunk - hi-tech/lo-life, techno-realist
Mixture-of-experts is when the LLM models take turns to hallucinate.

clip.cafe/whisky-galor...
'We don't have a village idiot, we just take it in turns.' - Whisky Galore
Macroon: We don't have a village idiot, we just take it in turns.
clip.cafe
November 28, 2025 at 5:56 PM
Stop being surprised about modern fantasy and sci-fi having been dumbed down. It is a reflection of the readership.
November 28, 2025 at 2:45 PM
Having just used Gemini for a day: benchmarks lie
November 28, 2025 at 12:28 PM
💥 OUT NOW! CURIOUS AI 97 💥

This week - we've released AuNoo AI community edition - the only Open Strategic Intelligence Platform.

open.substack.com/pub/thecurio...
The Curious AI 97
For pragmatic decision-makers & foresight pros. Techno-realism, second-order effects, and a touch of Stratechery wit.
open.substack.com
November 28, 2025 at 11:57 AM
Reposted by Oliver Rochford
I do find it disappointing to learn when a person doesn't read science.
November 25, 2025 at 6:35 PM
Reposted by Oliver Rochford
Anyone that's followed me for any length of time knows how much I love the Netherlands, I really enjoyed working there

Dutch security people are some of the smartest people I've ever worked with (in a bunker) and they're really good at intelligence.

And now I love them even more because this:
BREAKING:

NOS, the largest and most important public news organization in the Netherlands announced that they'll no longer post on X

You can follow them at @nieuws.nos.nl and @nieuwsuur.nl

Well done 👏

More of this please, you'll do better on Bluesky anyway 🤷‍♀️
November 25, 2025 at 9:14 PM
I am slowly coming to the conclusion that the Pacific is not a European concern. We should trade with China: they stop supporting Russia in Ukraine, we turn a blind eye to the South China Sea.
November 26, 2025 at 8:17 AM
I think the jury is out on whether LLMs make people more productive, but I think it's pretty clear it makes many more dishonest.
November 24, 2025 at 11:23 AM
Anyone else getting tired of people on Bluesky posting content from X? If I wanted to still see mental diarrhoea I wouldn't have migrated
November 24, 2025 at 11:09 AM
All of the anthropomorphization of AI is the blowback from "you don't need to be technical to work in tech". Now we have a bunch of marketers and sales people in charge who think AI is sentient.
November 24, 2025 at 7:44 AM
Now I know why I've always hated poets.....

Adversarial Poetry as a Universal Single-Turn Jailbreak Mechanism in Large Language Models

arxiv.org/abs/2511.15304
Adversarial Poetry as a Universal Single-Turn Jailbreak Mechanism in Large Language Models
We present evidence that adversarial poetry functions as a universal single-turn jailbreak technique for Large Language Models (LLMs). Across 25 frontier proprietary and open-weight models, curated po...
arxiv.org
November 23, 2025 at 10:09 AM
💥 OUT NOW! THE CURIOUS AI 96!! 💥

This week, Tech firms over estimate the demand for AI, and human-shaped robots, are we asking for too much, too early?

open.substack.com/pub/thecurio...
Curious AI Weekly Digest — Issue 96
For pragmatic decision-makers & foresight pros. Techno-realism, second-order effects, and a touch of Stratechery wit.
open.substack.com
November 21, 2025 at 2:00 PM
I think that the more intellectually bankrupt Silicon Valley is becoming, the more desperate the quest for superintelligence, because we always seek what we don't have.
November 21, 2025 at 8:55 AM
In 2026, we can expect the emergence of cost-maximizing attacks -techniques that deliberately inflate the operational cost of AI agent-based SOCs by generating adversarial traffic, ambiguous alerts, or high-complexity workloads.

cyberfuturist.medium.com/the-emergenc...
The Emergence of “Cost-maximizing” Adversarial AI Attacks.
Prediction Restatement
cyberfuturist.medium.com
November 20, 2025 at 11:32 AM
Reposted by Oliver Rochford
These are not serious people.
November 19, 2025 at 9:05 AM
Reposted by Oliver Rochford
POV: you are a young woman celebrating a recent academic success
November 17, 2025 at 7:20 PM
This explains so much! I guess lots of tech ceo's are dogfooding AI...

Increased use of AI tools can lead users to overestimate their own skills and knowledge, potentially causing inflated self-confidence.

www.livescience.com/technology/a...
The more that people use AI, the more likely they are to overestimate their own abilities
Researchers found that AI flattens the bell curve of a common principle in human psychology, known as the Dunning-Kruger effect, giving us all the illusion of competence.
www.livescience.com
November 17, 2025 at 4:56 PM
💥 OUT NOW! THE CURIOUS AI 95! 💥

This week - when AI superstars do Indie, and what if this time it's different, and efficiencies don't unlock wage growth?

open.substack.com/pub/thecurio...
Curious AI Weekly Digest — Issue 95
November 14, 2025
open.substack.com
November 14, 2025 at 2:36 PM
Sam Altman sounds a lot like WeWork did

OpenAI says it plans to report stunning annual losses through 2028 and then turn wildly profitable just two years later

fortune.com/2025/11/12/o...
OpenAI says it plans to report stunning annual losses through 2028—and then turn wildly profitable just two years later | Fortune
The company anticipates burning through roughly $9 billion this year on $13 billion in sales, a cash burn rate of approximately 70% of revenue.​
fortune.com
November 13, 2025 at 10:28 AM
Russian Robots seem much like Russian Soldiers
Watch Russia's AI robot fall seconds after being unveiled

bbc.com/news/videos/crre8g5e45jo
Watch: Russia's AI robot falls seconds after being unveiled
Footage shows the moment Russia's first anthropomorphic robot fell just seconds after its debut
bbc.com
November 13, 2025 at 10:09 AM
It's not a "Digital AI worker", it's just a computer program.
November 13, 2025 at 9:15 AM
How are you going to afford a robot when you don't have a job? Who owns the means of production will be the biggest point of contention in the near future.
November 13, 2025 at 8:56 AM
Japanese Woman Marries AI Character She Generated on ChatGPT

www.tokyoweekender.com/japan-life/j...
November 12, 2025 at 7:09 PM
Reposted by Oliver Rochford
They’ve laid off so many people that the government is now getting its economic data from DoorDash.
November 12, 2025 at 1:09 AM
Full on rogue state territory
November 11, 2025 at 3:35 PM