An overview of interesting AI News for the Yate and Sodbury U3A's monthly AI News meeting held on April 9th 2026 (meetings are second Thursday of month, 1400-1500, St John's Church Centre, Wickwar Road, Chipping Sodbury, UK - all welcome). Note that this is a summary to act as background as the topics are discussed in detail at the meeting itself. We sometimes have extended meetings where we demonstrate other AI products and services.
The 'Introduction to AI' course we run in Chipping Sodbury now uses these News sessions to replace the original classes 5 and 6 as the topics in them (humanoid robots, self driving cars, ethics, futures) are now 'news' not 'distant futures'.
More to follow:
- RSI (Recursive Self Improvement) the loop driving rapid improvement
- Claude Code Horse Racing Experiment
From an OpenAI Forum discussion early April 2026 link: Sam Altman on Building the Future of AI .
📁 Sam Altman, CEO of OpenAI, says this moment feels like right before something big.
— Jon Hernandez (@JonhernandezIA) April 8, 2026
Like the early days of COVID, when few saw it coming.
The shift has already happened.
But society has not caught up yet. pic.twitter.com/Q0XQ92aK7a
After initial uncertainty the last year brought AI 'Reasoning' models then coding was solved, followed by AI Agents leading to self-improvement.
Marc Andreessen explains how AI actually works in real world (This time it's actually different)
— Tony Jacob | FindaClip.com (@tonjkb) April 3, 2026
"Now it's working. LLMs from ChatGPT through spring '25, skeptics said it's just pattern completion, hallucinations too high, great for creative writing but not coding, medicine,… pic.twitter.com/cTO0tvCs2b
Mythos is said to be as good as a very good professional at finding 'exploits' (bugs) in complex software. So much so that it cannot yet be released. Project Glasswing is Anthropic's initiative to protect the software industry.
So this is Anthropic’s case for why Mythos is staying off the public shelf, out of fear of what damage it could cause 🤯
— Rohan Paul (@rohanpaul_ai) April 8, 2026
Massive leap in capabilities, especially in cybersecurity. It's being used internally at Anthropic and shared only with a small group of vetted partners… https://t.co/wfoKwx1S2l pic.twitter.com/ivUuf4U0ND
Two years worth of research coming to fruition.
🥔 Spud is coming.@gdb: "Spud is a new pretrain and the culmination of two years of research."
— Dan McAteer (@daniel_mac8) April 5, 2026
Spud is a BIG DEAL.
Expect something special.
Greg said it will:
> Solve harder problems
> Have a more nuanced understanding of context
> Conform to you rather than you to it
The… pic.twitter.com/x4SSBbbkec
Apparently it was losing money and using scarce 'compute'. Below is an example of their finest work. There are many competing alternatives of course, for example Google with Gemini Veo and X with X Imagine Video. Elon says that their video tool is already profitable and very important as they build world models for robotics.
My most popular Sora video was “an Elaborate regency romance where everyone is wearing a live duck for a hat (each duck is also wearing a hat), a llama plays a flute, prestige drama”
— Ethan Mollick (@emollick) March 25, 2026
I am not sure why OpenAI has decided their compute has more valuable uses. Really a mystery. pic.twitter.com/bXCGXUiFSm
Why huge deals like the one for Sora with Disney, made just a few months ago, were sacrificed.
Sam Altman explains why OpenAI decided to shut down Sora
— Tony Jacob | FindaClip.com (@tonjkb) April 3, 2026
"So apparently you're not doing Sora anymore. You guys have made the decision to shut it down. So take me to the room, Sam. It like three months ago openai signed a deal with Disney This is like a landmark deal in so many… pic.twitter.com/fjo0TZuuM0
The many competing designs are showing greater and greater functionality.
Today, three such companies appear to have, almost by mutual agreement, released updates on their progress:
— CyberRobo (@CyberRobooo) March 25, 2026
Google DeepMind continues to expand its humanoid robotics ecosystem (Apptronik Agility …)through a collaboration with Agile Robotics. Gemini Robotics is being deployed to… pic.twitter.com/ZvCXPlt6Mw
Many suppliers have been showing robots with superb manual dexterity such as this.
The Chinese industrial robotics company ROKAE shared this footage of its humanoid robot repeatedly threading a needle.
— Mike Kalil (@mikekalilmfg) March 27, 2026
The Beijing-based firm said its dual AR Series arms achieve around ±1 mm and micro-level precision.
Threading a needle once is considered a major an… pic.twitter.com/8lJxiwmJJh
The big unveil for the Tesla humanoid robot was meant to be by the end of the first quarter of 2026. Elon Musk says it just needs a few finishing touches.
long line of people waiting in line for popcorn from Optimus pic.twitter.com/56mnBDT2ZD
— Whole Mars Catalog (@wholemars) March 31, 2026
The video is said to be about recruiting people to work on the Optimus flowline, but in reality it is a little early publicity for what is coming.
Optimus
— Elon Musk (@elonmusk) March 25, 2026
pic.twitter.com/d6AU3p4xBn
There are many discussions on how people will gain income after the Singularity. Brian Roemmele is a futurist with a great track record. He suggest that the government will provide only a Universal Basic Income. One idea - he recommends buying a robot that can then make another robot and so on.
The Math of Abundance: UBI becomes UHI+
— Brian Roemmele (@BrianRoemmele) April 1, 2026
Core Takeaway:
Universal High Income (UHI)
isn't a bigger government transfer.
It is the market value of your personal robotic production meeting near-zero living costs. https://t.co/mtjNxlrZwy pic.twitter.com/Vgwf7RM9NC
Specialist AI models are being built across all of the Sciences. As an example, Heaviside is designed to speed up the evaluation and design of large classes of electronic circuits.
Today, we're announcing Heaviside, our foundation model for electromagnetism.
— Arya Hezarkhani (@_i_am_arya) March 31, 2026
Trained on tens of millions of designs and over 20 years of proprietary simulation data, Heaviside predicts electromagnetic behavior from geometry in 13ms, which is 800,000x faster than a commercial… pic.twitter.com/hZCpq3mBzZ
We don't know if these accounts are true or correct, but there are more and more reports of serious problems.
AI psychosis is getting worse pic.twitter.com/ZjXeJM65Cs
— Mo Bitar (@atmoio) April 3, 2026
Here the only source material was the pdf of a proposed 'Free Speech' law.All else is produced by NotebookLM.
I asked Google's NotebookLM for a video explainer of the proposed UK Free Speech Bill from @ASI @prestonjbyrne to use in our next AI class pic.twitter.com/5oZXb1ZiDO
— The Silver AI Project - free AI training materials (@SilverAIProject) April 2, 2026
Gemma 4 is Gemini's little sister though still powerful by all accounts.
Friendly reminder that Google has an official app to run Gemma 4 on your phone.
— Paul Couvert (@itsPaulAi) April 4, 2026
- 100% open source
- Fully offline and private
- Multimodal with text/audio/image
- Works with Gemma E4B and E2B
And the app is available on both iOS and Android.
Steps and download below pic.twitter.com/Ii0Mh4XpHH
Soon the RoboTaxi factory flowline comes on stream. Will they all be able to go into service? Meanwhile in China, a system problem caused Baidu Apollo Go self-driving taxis in Wuhan to stop operating. Some were on major highways at the time. Teslas are self-contained in that all AI used for driving is on board.
Unsupervised Tesla robotaxi drives me from LA Fitness to Whole Foods with no issues. Nobody in the car except me and no driver pic.twitter.com/lWNI3OcpcI
— Abhimanyu Yadav (@WorldlyReviewer) April 3, 2026
It seems as if Claude can 'feel' similar emotions to humans. Or at the least it uses similar 'circuits' internally when it is experiencing situations that would lead humans to feel those emotions.
New Anthropic research: Emotion concepts and their function in a large language model.
— Anthropic (@AnthropicAI) April 2, 2026
All LLMs sometimes act like they have emotions. But why? We found internal representations of emotion concepts that can drive Claude’s behavior, sometimes in surprising ways. pic.twitter.com/LxFl7573F9
There are massive redundancies across many large tech firms, many are directly due to extensive use of AI.
Wow… I know this is happening but it’s still so jarring to listen to
— Midnight Capital LLC (@Midnight_Captl) April 2, 2026
This is Owen Jennings, the Business Lead at Block (the company that just laid off 40% of it’s entire company) talking about the new workflow for the company post reorg
>Dev teams of 14 turned to 3
>Claude… pic.twitter.com/TZWKrI6X9K
Anthropic have resisted Claude becoming part of third party tools, preferring instead to build their own complete solution.
Introducing Claude Managed Agents: everything you need to build and deploy agents at scale.
— Claude (@claudeai) April 8, 2026
It pairs an agent harness tuned for performance with production infrastructure, so you can go from prototype to launch in days.
Now in public beta on the Claude Platform. pic.twitter.com/vHYfiC1G56
The mission - to build vast volumes of ultra-high-performance chips for AI Servers and more.
Intel is proud to join the Terafab project with @SpaceX, @xAI, and @Tesla to help refactor silicon fab technology.
— Intel (@intel) April 7, 2026
Our ability to design, fabricate, and package ultra-high-performance chips at scale will help accelerate Terafab’s aim to produce 1 TW/year of compute to power… pic.twitter.com/2vUmXn0YhH
Tesla and SpaceX/xAl are working together on this massive project, making it more likely that they will one day merge to become an enormous company leading the way into expansion in Space.
Tesla and SpaceX have introduced an updated Terafab website along with a new rendering:
— Sawyer Merritt (@SawyerMerritt) April 7, 2026
"Tesla, SpaceX, and xAl are launching the most epic chip-building effort ever - combining logic, memory and advanced packaging under one roof. We aspire to be a galactic civilization."… pic.twitter.com/AV2cW44NtB
Slow moving delivery robots make for an easy target for a new generation of 'Luddites'. This incident looks as if it could be staged, but the feelings expressed have been seen many times elsewhere.
The robot revolution is meeting resistance
— Mike Kalil (@mikekalilmfg) April 2, 2026
California-based Serve Robotics is busy dealing with public backlash.
In recent viral clips, a man in Los Angeles delivers a profane rant to one its delivery robots named "Mingo" that was stuck at a crosswalk. He accused the service… pic.twitter.com/sBXVbGCeWW