Utopiah
@benetou.fr
Pinned
Utopiah
@benetou.fr
· Nov 8
Naive heuristic : avoiding gadgets or services (free or not) that increase inequality by design, through technology or business model or both, would be a good starting point.
Sorry to be this grim but... ALL those negative posts about Meta, its XR downsizing📉 and the closure of its business division are basically unrolling the red carpet to the @valvesoftware.com Steam Frame 😎.
2026 the year of the Linux desktop and VR headset! 🐧🎆
2026 the year of the Linux desktop and VR headset! 🐧🎆
January 16, 2026 at 4:14 PM
Sorry to be this grim but... ALL those negative posts about Meta, its XR downsizing📉 and the closure of its business division are basically unrolling the red carpet to the @valvesoftware.com Steam Frame 😎.
2026 the year of the Linux desktop and VR headset! 🐧🎆
2026 the year of the Linux desktop and VR headset! 🐧🎆
Reposted by Utopiah
"This isn’t even venture capital anymore. Venture capital means you take a risk. High risk, high reward. If there’s no risk, that’s not venture capital." —@anildash.com sf.gazetteer.co/theyve-pickl...
‘They’ve pickled each others’ brains’
Longtime entrepreneur and critic Anil Dash speaks with Gazetteer on where the tech industry has been and where it may be going.
sf.gazetteer.co
January 14, 2026 at 11:06 PM
"This isn’t even venture capital anymore. Venture capital means you take a risk. High risk, high reward. If there’s no risk, that’s not venture capital." —@anildash.com sf.gazetteer.co/theyve-pickl...
"only about an hour's worth of licking"
Oh well.
Oh well.
January 15, 2026 at 2:28 PM
"only about an hour's worth of licking"
Oh well.
Oh well.
Wow... even wondering WHICH trick to learn in what order while watching somebody doing mind blowing ice skating trick?
What if there was a method to the madness?
Check out freestylepedia.org and its Trick Tree!
What if there was a method to the madness?
Check out freestylepedia.org and its Trick Tree!
January 15, 2026 at 1:22 PM
Wow... even wondering WHICH trick to learn in what order while watching somebody doing mind blowing ice skating trick?
What if there was a method to the madness?
Check out freestylepedia.org and its Trick Tree!
What if there was a method to the madness?
Check out freestylepedia.org and its Trick Tree!
Pure player with short runway, no income, nothing to loose openai.com/index/introd...
Incumbent with established revenues dominating markets arstechnica.com/ai/2026/01/g...
Very different behaviors.
Incumbent with established revenues dominating markets arstechnica.com/ai/2026/01/g...
Very different behaviors.
Introducing ChatGPT Health
A dedicated experience in ChatGPT designed for health and wellness.
openai.com
January 13, 2026 at 9:43 AM
Pure player with short runway, no income, nothing to loose openai.com/index/introd...
Incumbent with established revenues dominating markets arstechnica.com/ai/2026/01/g...
Very different behaviors.
Incumbent with established revenues dominating markets arstechnica.com/ai/2026/01/g...
Very different behaviors.
I work in XR, not AI, yet I complain a lot about AI, specifically its LLM and GenAI hype, not public research on AI as a field.
That said it does not mean commercial XR is perfect.
What is important is NOT how a technology is being sold but rather how it's actually used www.bbc.com/news/article...
That said it does not mean commercial XR is perfect.
What is important is NOT how a technology is being sold but rather how it's actually used www.bbc.com/news/article...
Smart glasses: 'I was secretly filmed and trolled online'
Footage of Oonagh was posted on social media where it attracted hundreds of sexual and derogatory comments.
www.bbc.com
January 13, 2026 at 9:25 AM
I work in XR, not AI, yet I complain a lot about AI, specifically its LLM and GenAI hype, not public research on AI as a field.
That said it does not mean commercial XR is perfect.
What is important is NOT how a technology is being sold but rather how it's actually used www.bbc.com/news/article...
That said it does not mean commercial XR is perfect.
What is important is NOT how a technology is being sold but rather how it's actually used www.bbc.com/news/article...
@redskyforge.bsky.social hi there, any chance to get notifications on replies? I commented on techposts.eu/post/159 but have to manually go there. I wish there was an easier way to keep track of on-going conversations.
Tech Posts Europe
Tech News, Tech Jobs and Startup Jobs from across Europe
techposts.eu
January 13, 2026 at 8:55 AM
@redskyforge.bsky.social hi there, any chance to get notifications on replies? I commented on techposts.eu/post/159 but have to manually go there. I wish there was an easier way to keep track of on-going conversations.
Yet another "agentic" project on HackerNews news.ycombinator.com/item?id=4658...
Honestly it feels like most LLM for code or agents related projects are reverse castles of cards.
They are VERY complex, quite brittle ... and arguably often pointless.
Sure they exist, they run... but why?!
Honestly it feels like most LLM for code or agents related projects are reverse castles of cards.
They are VERY complex, quite brittle ... and arguably often pointless.
Sure they exist, they run... but why?!
a man wearing glasses is building a house out of playing cards including a king
ALT: a man wearing glasses is building a house out of playing cards including a king
media.tenor.com
January 12, 2026 at 1:41 PM
Yet another "agentic" project on HackerNews news.ycombinator.com/item?id=4658...
Honestly it feels like most LLM for code or agents related projects are reverse castles of cards.
They are VERY complex, quite brittle ... and arguably often pointless.
Sure they exist, they run... but why?!
Honestly it feels like most LLM for code or agents related projects are reverse castles of cards.
They are VERY complex, quite brittle ... and arguably often pointless.
Sure they exist, they run... but why?!
Damn scientists or historians with their so called "facts"!
What we need is a true leader! Somebody who will ... use their... powerful... voice and ... do stuff! Ah! Now that is leadership!
/s obviously.
What we need is a true leader! Somebody who will ... use their... powerful... voice and ... do stuff! Ah! Now that is leadership!
/s obviously.
What can history teach us about what happens when a populist strongman with an idiosyncratic taste for low interest rates undermines central bank independence?
January 12, 2026 at 10:01 AM
Damn scientists or historians with their so called "facts"!
What we need is a true leader! Somebody who will ... use their... powerful... voice and ... do stuff! Ah! Now that is leadership!
/s obviously.
What we need is a true leader! Somebody who will ... use their... powerful... voice and ... do stuff! Ah! Now that is leadership!
/s obviously.
The fact that top AI researchers kept on working in a non-profiy turned for-profit shows no matter how tech smart you can be, you can still get manipulated by a conman.
January 10, 2026 at 8:29 AM
The fact that top AI researchers kept on working in a non-profiy turned for-profit shows no matter how tech smart you can be, you can still get manipulated by a conman.
Honestly tired of the "kill Nazis" posts.
The point of having a history with actual Nazis is to NOT give them power in the first place!
If there is ONE lesson that millions of lives lost could give the rest of Humanity forever is precisely that : do not kill Nazis, just don't offer them power.
The point of having a history with actual Nazis is to NOT give them power in the first place!
If there is ONE lesson that millions of lives lost could give the rest of Humanity forever is precisely that : do not kill Nazis, just don't offer them power.
January 10, 2026 at 8:03 AM
Honestly tired of the "kill Nazis" posts.
The point of having a history with actual Nazis is to NOT give them power in the first place!
If there is ONE lesson that millions of lives lost could give the rest of Humanity forever is precisely that : do not kill Nazis, just don't offer them power.
The point of having a history with actual Nazis is to NOT give them power in the first place!
If there is ONE lesson that millions of lives lost could give the rest of Humanity forever is precisely that : do not kill Nazis, just don't offer them power.
Reposted by Utopiah
TL;DR: "model creators realized they had a powerful source of labelled training data: the behavior of the users themselves. If an assistant offered up suggested code, the code ran successfully, and the user accepted the code, that was a positive signal"
Reverse self-improvement.
Reverse self-improvement.
Newer AI Coding Assistants Are Failing in Insidious Ways
Newer AI Coding Assistants Are Failing in Insidious Ways
One AI coding assistant power user says the tools are hitting a plateau, and some are even declining. What's causing this unexpected twist in tech?
spectrum.ieee.org
January 8, 2026 at 2:26 PM
TL;DR: "model creators realized they had a powerful source of labelled training data: the behavior of the users themselves. If an assistant offered up suggested code, the code ran successfully, and the user accepted the code, that was a positive signal"
Reverse self-improvement.
Reverse self-improvement.
"all models struggled on tasks involving false beliefs reported in the first person"
That's precisely when my expertise was valued. It wasn't when I told a colleague "Damn... what a good idea" despite thinking from own expertise that it was wrong, rather it was when I warned them it looked off.
That's precisely when my expertise was valued. It wasn't when I told a colleague "Damn... what a good idea" despite thinking from own expertise that it was wrong, rather it was when I warned them it looked off.
#AIML AI’s Wrong Answers Are Bad. Its Wrong Reasoning Is Worse - examples from Stanford KaBLE James Zou arXiv:2410.21195 and medical agents Yinghao Zhu arXiv:2510.10185 (Edd Gent 2025-12-02): spectrum.ieee.org/ai-reasoning...
Researchers Are Uncovering Fundamental Flaws in How AI Reasons
As AI takes on agent roles in critical fields, reasoning failures raise risks.
spectrum.ieee.org
January 8, 2026 at 2:11 PM
"all models struggled on tasks involving false beliefs reported in the first person"
That's precisely when my expertise was valued. It wasn't when I told a colleague "Damn... what a good idea" despite thinking from own expertise that it was wrong, rather it was when I warned them it looked off.
That's precisely when my expertise was valued. It wasn't when I told a colleague "Damn... what a good idea" despite thinking from own expertise that it was wrong, rather it was when I warned them it looked off.
The biggest risk is repeating terms from the industry.
The models are NOT "thinking", nor "reasoning". We conflate grand claims because we are too lazy to use technical terms, but doing so helps to sell lies.
There is no reasoning.
A broken clock is right twice a day, it doesn't "give time" then.
The models are NOT "thinking", nor "reasoning". We conflate grand claims because we are too lazy to use technical terms, but doing so helps to sell lies.
There is no reasoning.
A broken clock is right twice a day, it doesn't "give time" then.
#AIML AI’s Wrong Answers Are Bad. Its Wrong Reasoning Is Worse - examples from Stanford KaBLE James Zou arXiv:2410.21195 and medical agents Yinghao Zhu arXiv:2510.10185 (Edd Gent 2025-12-02): spectrum.ieee.org/ai-reasoning...
Researchers Are Uncovering Fundamental Flaws in How AI Reasons
As AI takes on agent roles in critical fields, reasoning failures raise risks.
spectrum.ieee.org
January 8, 2026 at 1:37 PM
The biggest risk is repeating terms from the industry.
The models are NOT "thinking", nor "reasoning". We conflate grand claims because we are too lazy to use technical terms, but doing so helps to sell lies.
There is no reasoning.
A broken clock is right twice a day, it doesn't "give time" then.
The models are NOT "thinking", nor "reasoning". We conflate grand claims because we are too lazy to use technical terms, but doing so helps to sell lies.
There is no reasoning.
A broken clock is right twice a day, it doesn't "give time" then.
More wind and snow today, just as amazing video.benetou.fr/w/p/tk8vGN1A...
Ice skating in Tervuren, Belgium
Over https://www.openstreetmap.org/way/30079253
video.benetou.fr
January 7, 2026 at 12:12 PM
More wind and snow today, just as amazing video.benetou.fr/w/p/tk8vGN1A...
That's way to kind, they're just snake oil salesmen, con artists even. They are not labelled as such simply because of the scale at which it's done : too many people are in on the con.
Prophet revises time of apocalypse after originally projected date passes uneventfully.
Leading AI expert delays timeline for its possible destruction of humanity
Former OpenAI employee Daniel Kokotajlo says progress to AGI is ‘somewhat slower’ than first predicted
www.theguardian.com
January 6, 2026 at 3:55 PM
That's way to kind, they're just snake oil salesmen, con artists even. They are not labelled as such simply because of the scale at which it's done : too many people are in on the con.
Ice skating while snow is still falling video.benetou.fr/w/72CG96XHUz... so quiet, so good. Can't wait to go back!
Tervuren VID_20260106_094102 long session (~2min) under falling snow
2nd morning on https://www.openstreetmap.org/way/30079253
video.benetou.fr
January 6, 2026 at 3:44 PM
Ice skating while snow is still falling video.benetou.fr/w/72CG96XHUz... so quiet, so good. Can't wait to go back!
IMHO demonstrations like media.ccc.de/v/39c3-hacki... and media.ccc.de/v/39c3-bluet... are great reminders : don't assume your home is safe from hackers. It's not. You're just not interesting enough, so far! #39c3
Hacking washing machines
Almost everyone has a household appliance at home, whether it's a washing machine, dishwasher, or dryer. Despite their ubiquity, little i...
media.ccc.de
January 5, 2026 at 1:54 PM
IMHO demonstrations like media.ccc.de/v/39c3-hacki... and media.ccc.de/v/39c3-bluet... are great reminders : don't assume your home is safe from hackers. It's not. You're just not interesting enough, so far! #39c3
Data Centre Drought www.arte.tv/en/videos/12... in English on @artefr.bsky.social a 2025 documentary by Pierre Morel
"Data centres have been built on the outskirts of the capital Santiago for google and other tech giants..." or the very clear cost of BigTech.
"Data centres have been built on the outskirts of the capital Santiago for google and other tech giants..." or the very clear cost of BigTech.
ARTE Reportage - Data Centre Drought - Watch the full documentary | ARTE in English
In Chile water is disappearing, and not just because of drought. Data centres have been built on the outskirts of the capital Santiago for google and other tech giants, encouraged by the socialist gov...
www.arte.tv
January 5, 2026 at 12:55 PM
Data Centre Drought www.arte.tv/en/videos/12... in English on @artefr.bsky.social a 2025 documentary by Pierre Morel
"Data centres have been built on the outskirts of the capital Santiago for google and other tech giants..." or the very clear cost of BigTech.
"Data centres have been built on the outskirts of the capital Santiago for google and other tech giants..." or the very clear cost of BigTech.
There is something special about leaving the first trace video.benetou.fr/w/p/tk8vGN1A... ❄️ way to start 2026! :D
Ice skating in Tervuren, Belgium
Over https://www.openstreetmap.org/way/30079253
video.benetou.fr
January 5, 2026 at 11:55 AM
There is something special about leaving the first trace video.benetou.fr/w/p/tk8vGN1A... ❄️ way to start 2026! :D
OK ok we know that cars pollute, we have all witness rush hour traffic with 1 person per car... but have you considered 0 (ZERO!) person per car?!
What the fuck.
Honestly, as a species we are pretty smart but damn our dominant economical system based on greed is NOT working for our ecosystem!
What the fuck.
Honestly, as a species we are pretty smart but damn our dominant economical system based on greed is NOT working for our ecosystem!
January 4, 2026 at 6:58 PM
OK ok we know that cars pollute, we have all witness rush hour traffic with 1 person per car... but have you considered 0 (ZERO!) person per car?!
What the fuck.
Honestly, as a species we are pretty smart but damn our dominant economical system based on greed is NOT working for our ecosystem!
What the fuck.
Honestly, as a species we are pretty smart but damn our dominant economical system based on greed is NOT working for our ecosystem!
Reposted by Utopiah
Elon Musk, Feb 2025: "America should mind its own business rather than push for regime change all over the place"
Elon Musk, Jan 2026:
"Congratulations, President Trump!
This is a win for the world and a clear message to evil dictators everywhere."
Elon Musk, Jan 2026:
"Congratulations, President Trump!
This is a win for the world and a clear message to evil dictators everywhere."
January 4, 2026 at 5:00 PM
Elon Musk, Feb 2025: "America should mind its own business rather than push for regime change all over the place"
Elon Musk, Jan 2026:
"Congratulations, President Trump!
This is a win for the world and a clear message to evil dictators everywhere."
Elon Musk, Jan 2026:
"Congratulations, President Trump!
This is a win for the world and a clear message to evil dictators everywhere."
