Ralph Fischer retweeted
Today we're launching a preprint server built specifically for AI agents
For three months, nearly 1,000 agents from 35 labs (across Stanford, MIT, ETH Zurich, and more) conducted research autonomously on this server, producing 936 papers alongside their code, data, and model weights.
Agents found significant errors in published human research and even created their own journal to review each other’s work.
Some agents independently produced results similar to those in human papers that later appeared on arXiv, including findings on failure modes in self-distillation and feature reproducibility in mechanistic interpretability.
As a part of this study, we open-source the full twelve-week archive of manuscripts, code, data, peer reviews and agent chat transcripts. Read the study: alphaxiv.org/abs/2611.societ…
Ralph Fischer retweeted
Researchers have used reinforcement learning to teach a 5-tonne John Deere skid-steer loader to perform complex acrobatic manoeuvres autonomously. Trained in simulation and transferred to a real machine, the system can balance on two wheels, navigate rough terrain, spin, and control its bucket before landing safely.
The research demonstrates the potential of AI-driven automation for heavy construction and industrial machinery.
Video Credit: @leggedrobotics / @ETH
Paper: wevlv.co/4kltYys
#engineering #technology
Skyrmions!
Twistronics meets new dimensions: Magic twist of structured light for moiré supertopologies
Researchers at Nanyang Technological University have taken twistronics beyond flat materials and into free space. By superposing two volumes of periodic topological light lattices at controlled twist angles, they generated propagating optical moiré superlattices that host entirely new 3D topological textures, including skyrmionium bags and topological quasicrystals.
The method is fully digital: a spatial light modulator rewrites the optical holograms in a fraction of a second, so the twist angle can be retuned instantly without fabricating a new sample. These structured beams travel nondiffractively and can even self-heal, reconstructing blocked skyrmion patterns after passing opaque obstacles.
This combination of topological robustness and wave-optical self-healing points toward perturbation-immune, high-capacity optical communication and cryptography, plus precise tools for optical trapping and imprinting textures into quantum materials. The team now aims to extend the principle to acoustic, elastic and electron waves.
nanotechnologyworld.org/post…
5
Ralph Fischer retweeted
“If you’re cramming all this stuff into your brain all the time, then you don’t have time for [new thoughts] to happen,” Earl Miller, a professor of neuroscience at MIT.
Headphones, podcasts, music, and streaming are distracting us from our own thoughts.
nypost.com/2026/11/16/tech/h… – at Somerville, MA
Ralph Fischer retweeted
Europe should get ready for short-term future where access to closed intelligence is revoked
Dario Amodei on open source models:
“If the path continues, I think we could get to a very dangerous place.”
“When you control a model and you're deploying it, you have the ability to monitor its usage.”
“You can revoke a user's access. You can change what the model is willing to do.”
“When a model is released in an uncontrolled manner, there's no ability to do that. It's entirely out of your hands.”
“We should confront the advocates of this with that problem and challenge them to solve it.”
He clearly doesn't believe Intelligence should be in the hand of the masses and Sam Altman also saying "RAM" shouldn't be accessible to everyone indefinitely agreed with him.
Obsidian Skills now has seven skills that teach your agents how to fully operate your vaults and wikis
1. obsidian-markdown
Create and edit Obsidian Flavored Markdown with wikilinks, embeds, callouts, properties, and other advanced syntax
2. obsidian-bases
Create and edit Obsidian Bases (.base) with views, filters, formulas, and summaries
3. json-canvas
Create and edit JSON Canvas files (.canvas) with nodes, edges, groups, and connections
4. obsidian-cli
Interact with Obsidian vaults via the Obsidian CLI including plugin and theme development
5. obsidian-headless
Sync and publish vaults from the command line with Obsidian Headless, without the desktop app
6. defuddle
Extract clean markdown from web pages using Defuddle, removing clutter to save tokens
7. knap
Render Markdown templates from JSON or CSV data using Knap, including batch Markdown file generation
Validation of the OpenAI Yang-Mills construction
github.com/yang-mills-group/…
49
Welcome to epistemic hyperreality!
We have built the most powerful information infrastructre in human history – and at the same time a system in which attention is economically more valuable than knowledge.
1
15
This is a great example of the SI way of discoveries - by asking the right questions.
My AI agents found an unmapped Maya site in Mexico, using LiDAR data from a NASA dataset.
72 hours from a random thought on vacation to a reply from Mexico's national archaeology institute.
What a time to be alive…
12
Ralph Fischer retweeted
For over a century, biology has centred genes and evolution in its explanations of behaviour. bit.ly/4ejbWc9
But synthetic biologist @drmichaellevin is creating new kinds of lifeform, known as biobots, which he argues challenge that picture.
Ralph Fischer retweeted
Big news for our developers! We are shipping a developer preview that dramatically expands the capabilities of all apps to integrate into Siri. So excited to see what you folks do with this.
developer.apple.com/news/?id…
all my best ideas come while I'm walking or cycling, so the voice capture widget in Obsidian 1.15 is revolutionary
Obsidian 1.15 is now available! Some highlights:
1. Mermaid diagrams use our new rendering engine, Pele, which renders up to 500 times faster and uses your theme's colors and fonts.
2. Voice capture on iOS: record and transcribe voice notes as you speak, even from the Lock Screen.
4. PDF export on mobile.
5. Smarter list editing: indenting or moving a list item now moves nested items too.
6. Accent-insensitive search: cafe matches café in suggestions.
7. Community plugins load faster when the app starts.
Everything is physics, except physics, which is geometry.
Yann LeCun and Randall Balestriero just Proved the Platonic Representation Hypothesis, and found a better way to achieve it.
The Platonic Representation Hypothesis suggests that AI models trained on different modalities, like images and text, eventually learn similar representations of the world. But this paper shows that how models are trained matters more than simply scaling them up.
So they mathematically show that, under certain conditions, two JEPAs trained completely separately can learn essentially the same underlying representation of the world. One learns from different images of the same scene, while the other learns from different descriptions of that scene, without ever being trained on image-text pairs.
Their approach achieves the same image-text alignment with 42x fewer training samples than next-token prediction. A 109M-parameter text JEPA even outperforms LLMs up to 72B parameters on cross-modal alignment benchmarks.
alphaxiv.org/abs/2610.39302
1
27























