ePotentia
@epotentia.com
Scientific AI consultancy company talking about new developments in AI.
A new Moore’s Law is here: AI IQ. Forbes data shows models gaining ~2.5 IQ points monthly. While hardware hits limits, AI reasoning is scaling linearly toward university-level intelligence. Tech progress is now mental, not physical. 🚀
www.forbes.com/sites/johnwe...
www.forbes.com/sites/johnwe...
www.forbes.com
January 2, 2026 at 12:12 PM
A new Moore’s Law is here: AI IQ. Forbes data shows models gaining ~2.5 IQ points monthly. While hardware hits limits, AI reasoning is scaling linearly toward university-level intelligence. Tech progress is now mental, not physical. 🚀
www.forbes.com/sites/johnwe...
www.forbes.com/sites/johnwe...
After a life-altering spinal injury, Dan Richards is exploring AI "digital bridge" technology. These neural interfaces bypass damaged nerves to translate brain signals into physical action. A huge milestone for medical AI. 🧬
www.bbc.com/news/article...
www.bbc.com/news/article...
Swansea man paralysed by wave says AI could help him walk again
Dan Richards, 37, from Swansea was injured in a freak accident on New Year's Eve in 2023.
www.bbc.com
January 1, 2026 at 10:08 AM
After a life-altering spinal injury, Dan Richards is exploring AI "digital bridge" technology. These neural interfaces bypass damaged nerves to translate brain signals into physical action. A huge milestone for medical AI. 🧬
www.bbc.com/news/article...
www.bbc.com/news/article...
2025 gave us a whole new vocabulary. From "Superintelligence" and "Vibe Coding" to "Slop" and "Reasoning," MIT Tech Review breaks down the 14 terms that defined the year in AI. How many did you master?
www.technologyreview.com/2025/12/25/1...
www.technologyreview.com/2025/12/25/1...
AI Wrapped: The 14 AI terms you couldn’t avoid in 2025
From “superintelligence” to “slop,” here are the words and phrases that defined another year of AI craziness.
www.technologyreview.com
December 31, 2025 at 12:47 PM
2025 gave us a whole new vocabulary. From "Superintelligence" and "Vibe Coding" to "Slop" and "Reasoning," MIT Tech Review breaks down the 14 terms that defined the year in AI. How many did you master?
www.technologyreview.com/2025/12/25/1...
www.technologyreview.com/2025/12/25/1...
2025: The year AI agents went mainstream and the hardware race hit a fever pitch. A look back at the biggest shifts in tech this year and why 2026 is trending toward physical AI and agentic autonomy. The transition from "chatting" to "doing" is officially here.
edition.cnn.com/2025/12/30/t...
edition.cnn.com/2025/12/30/t...
How AI shook the world in 2025 and what comes next | CNN Business
Hundreds of billions of dollars spent, a surge in mental health concerns and thousands of jobs lost. These are just a few of the ways AI rattled the world in 2025 and what comes next.
edition.cnn.com
December 30, 2025 at 12:44 PM
2025: The year AI agents went mainstream and the hardware race hit a fever pitch. A look back at the biggest shifts in tech this year and why 2026 is trending toward physical AI and agentic autonomy. The transition from "chatting" to "doing" is officially here.
edition.cnn.com/2025/12/30/t...
edition.cnn.com/2025/12/30/t...
Microsoft is re-platforming Windows to be a "canvas for AI." By adding native agent infrastructure and an "agent workspace" to the taskbar, the OS is moving beyond simple apps to become a coordinator for autonomous agents. A massive shift for the 40-year-old OS.
www.forbes.com/sites/tonybr...
www.forbes.com/sites/tonybr...
www.forbes.com
December 22, 2025 at 12:43 PM
Microsoft is re-platforming Windows to be a "canvas for AI." By adding native agent infrastructure and an "agent workspace" to the taskbar, the OS is moving beyond simple apps to become a coordinator for autonomous agents. A massive shift for the 40-year-old OS.
www.forbes.com/sites/tonybr...
www.forbes.com/sites/tonybr...
Microsoft is undergoing an "AI reboot" under Satya Nadella, moving from a software factory to an "intelligence engine." The goal: prioritize agentic AI and avoid the complacency that kills tech giants. A fascinating look at the new stack of AI economics.
www.businessinsider.com/microsoft-ce...
www.businessinsider.com/microsoft-ce...
Inside Microsoft CEO Satya Nadella's AI revolution
Internal documents and executive interviews reveal sweeping organizational shifts to radically reshape how the company builds and funds its products.
www.businessinsider.com
December 17, 2025 at 11:19 AM
Microsoft is undergoing an "AI reboot" under Satya Nadella, moving from a software factory to an "intelligence engine." The goal: prioritize agentic AI and avoid the complacency that kills tech giants. A fascinating look at the new stack of AI economics.
www.businessinsider.com/microsoft-ce...
www.businessinsider.com/microsoft-ce...
OpenAI, Anthropic, and Block co-found the Agentic AI Foundation under the Linux Foundation to set open standards. AGENTS.md is the first major donation. Neutral governance for agentic AI interoperability is here.
openai.com/index/agenti...
openai.com/index/agenti...
AGENTS.md
AGENTS.md is a simple, open format for guiding coding agents. Think of it as a README for agents.
AGENTS.md
December 11, 2025 at 1:01 PM
OpenAI, Anthropic, and Block co-found the Agentic AI Foundation under the Linux Foundation to set open standards. AGENTS.md is the first major donation. Neutral governance for agentic AI interoperability is here.
openai.com/index/agenti...
openai.com/index/agenti...
The strategic play: Internal "code red" on Google followed by a huge enterprise flex. OpenAI data shows Custom GPT usage is up 19x (20% of messages!), demonstrating deep workflow integration. A clear defense of the B2B market.
techcrunch.com/2025/12/08/o...
techcrunch.com/2025/12/08/o...
OpenAI boasts enterprise win days after internal ‘code red’ on Google threat | TechCrunch
ChatGPT Enterprise usage grew 8x year-over-year as workers report saving an hour daily, but OpenAI faces competitive pressure from Anthropic and questions about cost sustainability.
techcrunch.com
December 9, 2025 at 11:31 AM
The strategic play: Internal "code red" on Google followed by a huge enterprise flex. OpenAI data shows Custom GPT usage is up 19x (20% of messages!), demonstrating deep workflow integration. A clear defense of the B2B market.
techcrunch.com/2025/12/08/o...
techcrunch.com/2025/12/08/o...
Mistral dropping new models to compete with OpenAI/Google.
The real story: Europe finally has a serious player. Open-weight options, EU-aligned governance, and genuinely competitive performance.
Vendor lock-in with US hyperscalers isn't the only path anymore.
www.cnbc.com/2025/12/02/m...
The real story: Europe finally has a serious player. Open-weight options, EU-aligned governance, and genuinely competitive performance.
Vendor lock-in with US hyperscalers isn't the only path anymore.
www.cnbc.com/2025/12/02/m...
French AI lab Mistral releases new AI models as it looks to keep pace with OpenAI and Google
The model unveiling comes on the back of a commercial contract with HSBC.
www.cnbc.com
December 4, 2025 at 1:05 PM
Mistral dropping new models to compete with OpenAI/Google.
The real story: Europe finally has a serious player. Open-weight options, EU-aligned governance, and genuinely competitive performance.
Vendor lock-in with US hyperscalers isn't the only path anymore.
www.cnbc.com/2025/12/02/m...
The real story: Europe finally has a serious player. Open-weight options, EU-aligned governance, and genuinely competitive performance.
Vendor lock-in with US hyperscalers isn't the only path anymore.
www.cnbc.com/2025/12/02/m...
What we would like to see is also a wider initiatives on public jailbreak detection probes to help european researchers quickly secure their models.
Why “public #AI”, built on #opensource software, is the way forward for the #EU - walledculture.org/why-public-a... and someone else agrees: opensource.org/blog/open-le...
Why “public AI”, built on open source software, is the way forward for the EU
<p>A quarter of a century ago, I wrote a book called “Rebel Code”. It was the first – and is still the only – detailed history of the origins and rise of free software and open source, based on interv...
walledculture.org
November 20, 2025 at 2:08 PM
What we would like to see is also a wider initiatives on public jailbreak detection probes to help european researchers quickly secure their models.
VendingBench Arena from Andon Labs tests what most benchmarks ignore: long-term coherence.
Can models actually run businesses consistently, or just produce impressive one-offs?
Gemini 3 topped it. We need more evals testing sustained performance, not party tricks.
andonlabs.com/evals/vendin...
Can models actually run businesses consistently, or just produce impressive one-offs?
Gemini 3 topped it. We need more evals testing sustained performance, not party tricks.
andonlabs.com/evals/vendin...
Vending-Bench Arena | Andon Labs
Vending-Bench Arena is our first multi-agent eval and adds a crucial component – competition. All participating agents manage their own vending machine at the same location. This leads to price wars a...
andonlabs.com
November 20, 2025 at 10:06 AM
VendingBench Arena from Andon Labs tests what most benchmarks ignore: long-term coherence.
Can models actually run businesses consistently, or just produce impressive one-offs?
Gemini 3 topped it. We need more evals testing sustained performance, not party tricks.
andonlabs.com/evals/vendin...
Can models actually run businesses consistently, or just produce impressive one-offs?
Gemini 3 topped it. We need more evals testing sustained performance, not party tricks.
andonlabs.com/evals/vendin...
Gemini 3 dropped and the benchmarks are wild. But so are some of the early edge cases people are discovering.
Classic frontier model energy: breakthrough performance with "wait, did it just do *that*?" moments.
Anyone stress-testing it yet?
blog.google/products/gem...
Classic frontier model energy: breakthrough performance with "wait, did it just do *that*?" moments.
Anyone stress-testing it yet?
blog.google/products/gem...
A new era of intelligence with Gemini 3
Today we’re releasing Gemini 3 – our most intelligent model that helps you bring any idea to life.
blog.google
November 19, 2025 at 11:37 AM
Gemini 3 dropped and the benchmarks are wild. But so are some of the early edge cases people are discovering.
Classic frontier model energy: breakthrough performance with "wait, did it just do *that*?" moments.
Anyone stress-testing it yet?
blog.google/products/gem...
Classic frontier model energy: breakthrough performance with "wait, did it just do *that*?" moments.
Anyone stress-testing it yet?
blog.google/products/gem...
Today's downtime due to cloudflare does illustrate again how we have single points of failure globally. Of course preferred stacks may differ between the west and other regions which elevates them to vulnerable points of attack as well...
when will the lesson finally get learned? #cloudflare
November 18, 2025 at 4:05 PM
Today's downtime due to cloudflare does illustrate again how we have single points of failure globally. Of course preferred stacks may differ between the west and other regions which elevates them to vulnerable points of attack as well...
Impressive, let's hope the final benchmarks measure up (and aren't overfitted). Some benchmarks like VendingBench reward long-term coherent action-taking with tools though which may mean it's a better strategist and didn't just cheat on the exams. :)
Gemini 3 model card leaked
the URL is taken down now, was here:
storage.googleapis.com/deepmind-med...
the URL is taken down now, was here:
storage.googleapis.com/deepmind-med...
November 18, 2025 at 2:27 PM
Impressive, let's hope the final benchmarks measure up (and aren't overfitted). Some benchmarks like VendingBench reward long-term coherent action-taking with tools though which may mean it's a better strategist and didn't just cheat on the exams. :)
Let's not forget our climate targets as we pursue infinite AI growth.
A new state-by-state evaluation of the projected energy and water use of "AI" data centers in the US. Looks… real bad. Like, "undoing tech sector climate gains" bad.
Strongly recommends immediately ensuring any & all new "AI" data centers to run on existing & expanded renewables grids. Which… yeah.
Strongly recommends immediately ensuring any & all new "AI" data centers to run on existing & expanded renewables grids. Which… yeah.
Environmental impact and net-zero pathways for sustainable artificial intelligence servers in the USA - Nature Sustainability
The rapid expansion of AI server installations in the United States poses sustainability challenges in terms of water usage and carbon emissions. A study now quantifies these potential impacts and out...
www.nature.com
November 14, 2025 at 1:33 PM
Let's not forget our climate targets as we pursue infinite AI growth.
Truly efficient and intelligent AI may require novel architectures. Being brave enough to leave so many resources behind to pursue a bold vision can only be applauded, though it'll be a loss for Meta to let him go.
Yann LeCun, a chief AI scientist at Meta, is planning to leave the company to build his own startup, which will focus on continuing his work on world models.
Meta's chief AI scientist Yann LeCun reportedly plans to leave to build his own startup | TechCrunch
Yann LeCun, a chief AI scientist at Meta, is planning to leave the company to build his own startup, which will focus on continuing his work on world models.
techcrunch.com
November 12, 2025 at 8:49 PM
Truly efficient and intelligent AI may require novel architectures. Being brave enough to leave so many resources behind to pursue a bold vision can only be applauded, though it'll be a loss for Meta to let him go.
If they succeed in accelerating medical research a little bit, perhaps all the money earned from dodgy ads will serve a purpose.
The billionaire couple is shifting the bulk of their philanthropic resources to Biohub, the pair’s science organization, and focusing on using artificial intelligence to accelerate scientific discovery.
Zuckerberg, Chan shift bulk of philanthropy to science, focusing on AI and biology to curb disease
For the past decade, Dr. Priscilla Chan and her husband Mark Zuckerberg have focused part of their philanthropy on a lofty goal — “to cure, prevent or manage all disease” — if not in their lifetime, then in their children’s.
bit.ly
November 8, 2025 at 12:06 PM
If they succeed in accelerating medical research a little bit, perhaps all the money earned from dodgy ads will serve a purpose.
Well at least if everything is AI we don't have the problem of detecting fakes... :)
Meta brings its short-form video feed of AI slop to Europe
Meta brings its short-form video feed of AI slop to Europe | TechCrunch
Meta says that since the launch of Vibes, media generation in the Meta AI app has jumped more than tenfold.
techcrunch.com
November 7, 2025 at 9:03 PM
Well at least if everything is AI we don't have the problem of detecting fakes... :)
Each young girl discouraged by persisting gender issues could mean a great scientific discovery that never will be. Looking forward to the next generating doing better and maybe building some new great scientific AI!
Happy Women in Physics Day to all of the amazing women and gender minorities in physics!! We are still greatly under-represented at all levels, so we need to continue encouraging girls and young women to consider studying physics and make systemic changes to the climate in the field. ⚛️ 🎢 👩🔬 🧪
November 7, 2025 at 9:00 PM
Each young girl discouraged by persisting gender issues could mean a great scientific discovery that never will be. Looking forward to the next generating doing better and maybe building some new great scientific AI!
How could we search a near-infinite universe without our best tools for big data. :)
In a significant advance for astronomy, researchers from the Breakthrough Listen initiative, working in collaboration with NVIDIA and utilizing their system on the SETI Institute's Allen Telescope Array, have enhanced the process for detecting Fast Radio Bursts. 🧪 🔭
Learn more:
Learn more:
Revolutionary AI System Achieves 600x Speed Breakthrough in the Search for Signals from Space
www.seti.org
November 7, 2025 at 8:56 PM
How could we search a near-infinite universe without our best tools for big data. :)
This is the type of scientific AI that can really make a difference. TB is one of the biggest killers worldwide, but getting both sufficient resources into rural will likely never happen. AI that is both cheap and accurate can make a big difference here and help boost health and productivity.
There's a global shortage of radiologists. Now artificial intelligence is helping speed up the diagnosis of tuberculosis in hard-to-reach communities. n.pr/49BWN3Z
AI steps in to detect the world's deadliest infectious disease
There's a global shortage of radiologists. Now artificial intelligence is helping speed up the diagnosis of tuberculosis in hard-to-reach communities.
n.pr
November 7, 2025 at 8:48 PM
This is the type of scientific AI that can really make a difference. TB is one of the biggest killers worldwide, but getting both sufficient resources into rural will likely never happen. AI that is both cheap and accurate can make a big difference here and help boost health and productivity.
It used to be easy to make a quick 50mb mailbox on your $10 domain. Now each account needs cloud storage, office programs and AI while most people's job only requires one or two of the available tools and no 2TB storage. Something to consider before unilaterally raising prices.
November 6, 2025 at 3:44 PM
It used to be easy to make a quick 50mb mailbox on your $10 domain. Now each account needs cloud storage, office programs and AI while most people's job only requires one or two of the available tools and no 2TB storage. Something to consider before unilaterally raising prices.
Power-efficient models, training methods and hardware will be at least as important to the future as building more and more data centers.
The rapid proliferation of services powered by artificial intelligence, ranging from the genuinely useful to the frivolous to the unnecessary, has a common element: the need for vast amounts of electricity and water for the data centers that make them possible.
The Unsustainable Strain Of AI’s Insatiable Power Needs
This week’s Current Climate newsletter also looks at how electric cars could get much cheaper and Rivian’s RJ Scaringe on China’s EV price advantage
www.forbes.com
November 5, 2025 at 9:41 PM
Power-efficient models, training methods and hardware will be at least as important to the future as building more and more data centers.
As AI often blindly consumes the internet, only to pollute it afterwards it's worth thinking about the future of information sharing. Do we take action to safeguard the internet? Do we trust and tolerate a limited number of LLM companies to become the source of truth for all information?
November 5, 2025 at 9:01 PM
As AI often blindly consumes the internet, only to pollute it afterwards it's worth thinking about the future of information sharing. Do we take action to safeguard the internet? Do we trust and tolerate a limited number of LLM companies to become the source of truth for all information?
One of the reasons we chose AI for characterization as one of our main applications is its ability to accelerate research in many different fields. Glad to see it also help cancer research. :)
A new technique combines lasers, fluorescent dyes and artificial intelligence to "tag" cells with abnormal chromosomes that may become cancerous.
Read the OPN story for more! bit.ly/4os3Jpa 💡⚛️🧪
Read the OPN story for more! bit.ly/4os3Jpa 💡⚛️🧪
November 5, 2025 at 8:47 PM
One of the reasons we chose AI for characterization as one of our main applications is its ability to accelerate research in many different fields. Glad to see it also help cancer research. :)