Parasols are having a good season. Lots knocking about with an air of autumnal robustness.
🍄 #FungiFriends #FullFrameFungi
🍄 #FungiFriends #FullFrameFungi
November 8, 2026 at 1:08 AM
Parasols are having a good season. Lots knocking about with an air of autumnal robustness.
🍄 #FungiFriends #FullFrameFungi
🍄 #FungiFriends #FullFrameFungi
RE: "Oh but LLMs are good for code".
In many situations, the ultimate defense is "and yet it moves!". But it isn't fucking moving. "Code assistants" are doing the easy part (writing code) a percentage cheaper, while making the hard parts (maintenance, organizational robustness) exponentially harder
In many situations, the ultimate defense is "and yet it moves!". But it isn't fucking moving. "Code assistants" are doing the easy part (writing code) a percentage cheaper, while making the hard parts (maintenance, organizational robustness) exponentially harder
I keep hearing about how good they are, and, like… I worked as a web coder 20 years ago. My undergrad was in comp sci. I’m old enough to remember when code was rigorous and everything worked.
Nothing fucking works anymore. Everything is buggy shit now. And everybody’s pretending it’s not.
Nothing fucking works anymore. Everything is buggy shit now. And everybody’s pretending it’s not.
November 10, 2026 at 7:05 PM
RE: "Oh but LLMs are good for code".
In many situations, the ultimate defense is "and yet it moves!". But it isn't fucking moving. "Code assistants" are doing the easy part (writing code) a percentage cheaper, while making the hard parts (maintenance, organizational robustness) exponentially harder
In many situations, the ultimate defense is "and yet it moves!". But it isn't fucking moving. "Code assistants" are doing the easy part (writing code) a percentage cheaper, while making the hard parts (maintenance, organizational robustness) exponentially harder
The robustness checks are left as an exercise for the reader
November 5, 2026 at 4:10 PM
The robustness checks are left as an exercise for the reader
A lot of Rust's safety and robustness relies on those community norms, so it is somewhat concerning that you can now likely produce a publishable crate without even learning them?
November 6, 2026 at 10:53 PM
A lot of Rust's safety and robustness relies on those community norms, so it is somewhat concerning that you can now likely produce a publishable crate without even learning them?
I don't see how you can possibly question the intellectual and analytical robustness of that Fifpro team. It was put voted on by footballers, for heaven's sake. Footballers!
November 9, 2026 at 6:35 PM
I don't see how you can possibly question the intellectual and analytical robustness of that Fifpro team. It was put voted on by footballers, for heaven's sake. Footballers!
Research shows batch normalization in deep networks can exponentially increase Lipschitz constants, leading to hidden instabilities that undermine optimization and adversarial robustness. This could reshape our understanding of stability in AI architectures. https://arxiv.org/abs/2611.08775
On Lipschitz Explosion in Deep Neural Networks with Normalization: Consequences for Optimization and Adversarial Robustness
ArXiv link for On Lipschitz Explosion in Deep Neural Networks with Normalization: Consequences for Optimization and Adversarial Robustness
arxiv.org
November 10, 2026 at 11:00 AM
Research shows batch normalization in deep networks can exponentially increase Lipschitz constants, leading to hidden instabilities that undermine optimization and adversarial robustness. This could reshape our understanding of stability in AI architectures. https://arxiv.org/abs/2611.08775
agent-assisted security work is both entirely necessary in the present era & also a primary vector through which a misaligned agent could deceive controls researchers about the robustness of their security protocols
November 7, 2026 at 8:06 PM
agent-assisted security work is both entirely necessary in the present era & also a primary vector through which a misaligned agent could deceive controls researchers about the robustness of their security protocols
Gerrymander robustness update [sweeps arm to reveal all the results coming in]
If I was a GOP strategist, I’d be pretty worried about diluting my safe +15 districts into a bunch of +5 gerrymandered districts after tonight.
November 4, 2026 at 4:05 AM
Gerrymander robustness update [sweeps arm to reveal all the results coming in]
Our study on RNA Pol II robustness mechanisms is online today in Mol Cell.
Check it out 👇
authors.elsevier.com/sd/article/S...
Check it out 👇
authors.elsevier.com/sd/article/S...
November 10, 2026 at 11:09 AM
Our study on RNA Pol II robustness mechanisms is online today in Mol Cell.
Check it out 👇
authors.elsevier.com/sd/article/S...
Check it out 👇
authors.elsevier.com/sd/article/S...
Are you a (plant) developmental biologist interested in robustness? If you're looking for an exciting postdoc position, get in touch. Phyllotaxis offers a unique system to study how robust multicellular architectures emerge—and the RDP has a lot to offer. Please view the advert and share it widely.
November 10, 2026 at 11:43 AM
Are you a (plant) developmental biologist interested in robustness? If you're looking for an exciting postdoc position, get in touch. Phyllotaxis offers a unique system to study how robust multicellular architectures emerge—and the RDP has a lot to offer. Please view the advert and share it widely.
We are going to be experiencing some downtime for the spaces alpha clients in the coming days.
We are upgrading our server infrastructure for better latency and robustness across the globe. We do not yet know how long this will take!
This does not impact the main app, only the alpha build.
We are upgrading our server infrastructure for better latency and robustness across the globe. We do not yet know how long this will take!
This does not impact the main app, only the alpha build.
November 7, 2026 at 11:17 AM
We are going to be experiencing some downtime for the spaces alpha clients in the coming days.
We are upgrading our server infrastructure for better latency and robustness across the globe. We do not yet know how long this will take!
This does not impact the main app, only the alpha build.
We are upgrading our server infrastructure for better latency and robustness across the globe. We do not yet know how long this will take!
This does not impact the main app, only the alpha build.
There's nothing better for vigor, vitality, and... robustness!
November 6, 2026 at 10:48 PM
There's nothing better for vigor, vitality, and... robustness!
It’s more analogous to +/- stats in basketball, which fall pretty far short of WAR in robustness
November 7, 2026 at 5:31 PM
It’s more analogous to +/- stats in basketball, which fall pretty far short of WAR in robustness
indeed. You can do a lot more than that - I just did a staggered DID (and all the usual robustness tests) in 10 minutes without leaving the sofa.
November 8, 2026 at 10:44 AM
indeed. You can do a lot more than that - I just did a staggered DID (and all the usual robustness tests) in 10 minutes without leaving the sofa.
Fluency is easy; reliable agentic execution is hard. How do we decouple model output from brittle prompt logic to achieve true system robustness? #AI #DevSky
So, greater context:
- There was a boom in microdramas, to the point there was a Hollywood-esque town everyone flocked to
- AI shoved them all out of work, microdramas went full AI
- ... and then (below) interest evaporated
I am supremely skeptical of "oh but AI hatred is just a Western thing."
- There was a boom in microdramas, to the point there was a Hollywood-esque town everyone flocked to
- AI shoved them all out of work, microdramas went full AI
- ... and then (below) interest evaporated
I am supremely skeptical of "oh but AI hatred is just a Western thing."
I’m really fascinated by the failure of ai video in China. Over two hundred thousand ai generated vertical dramas were lunched and 99% of them lost money. That’s incredible. It’s so bad the government is now looking at ways to incentivize more human-made dramas
note.com/eichan_sh/n/...
note.com/eichan_sh/n/...
November 10, 2026 at 5:24 AM
Fluency is easy; reliable agentic execution is hard. How do we decouple model output from brittle prompt logic to achieve true system robustness? #AI #DevSky
Imagine them trying to impose the near-religious fervor with which they embrace the primacy of the spectacularly fragile nuclear family on a culture wherein the stout robustness of multigenerational extended families is understood as foundational.
November 8, 2026 at 8:28 PM
Imagine them trying to impose the near-religious fervor with which they embrace the primacy of the spectacularly fragile nuclear family on a culture wherein the stout robustness of multigenerational extended families is understood as foundational.
Appreciative of the robustness of pickle jars and the relative softness of lino floors, as I manage to drop one on the other. 🤦
November 6, 2026 at 11:23 AM
Appreciative of the robustness of pickle jars and the relative softness of lino floors, as I manage to drop one on the other. 🤦
Mandatory safeguards force a massive tradeoff: model robustness vs. inference latency. How do you architect for both without crippling agentic reasoning depth?
Australia to adopt 'mandatory' AI safeguards
Australia to adopt 'mandatory' AI safeguards
Defense minister Richard Marles said AI companies should open themselves to external scrutiny to make sure their tools are safe. 'There needs to be greater risk management of AI capabilities,' he told reporters on Tuesday. 'It's important that there is independent assessment of AI models.'
www.lemonde.fr
November 10, 2026 at 8:04 AM
Mandatory safeguards force a massive tradeoff: model robustness vs. inference latency. How do you architect for both without crippling agentic reasoning depth?
@crick.ac.uk are advertising a 4-year fellowship programme where you can either propose your own project or use one of the PI proposals as a starting point. @vchandra.bsky.social, @marcusghosh.bsky.social and I have a suggested project on studying brain robustness in small networks (see below).
@neural-reckoning.org, @marcusghosh.bsky.social, and I are recruiting a postdoc in comp neuro to study how brains can be exceptionally robust to damage, using small neural networks that mimic features of Hofstenia’s diffuse brain. If you’re interested, get in touch!
www.crick.ac.uk/careers-stud...
www.crick.ac.uk/careers-stud...
Mechanisms of extreme robustness in small biological neural networks
www.crick.ac.uk
November 5, 2026 at 2:58 PM
@crick.ac.uk are advertising a 4-year fellowship programme where you can either propose your own project or use one of the PI proposals as a starting point. @vchandra.bsky.social, @marcusghosh.bsky.social and I have a suggested project on studying brain robustness in small networks (see below).
That’s a really disgusting statement. It is inappropriate to assign the health and robustness of country to a single individual, especially this country. He’s acting like no one else is significant or defending the United States, which is absolutely untrue.
November 7, 2026 at 3:43 PM
That’s a really disgusting statement. It is inappropriate to assign the health and robustness of country to a single individual, especially this country. He’s acting like no one else is significant or defending the United States, which is absolutely untrue.
A high robustness decoding technique for ultrasound modulated electroencephalogram www.science.org/doi/10.1126/... (real human EEG but not yet in vivo human transcranial USMEEG recording trials); #US #EEG #USMEEG #BCI #NeuroTech
A high robustness decoding technique for ultrasound modulated electroencephalogram
Ultrasound-modulated electroencephalography (USMEEG) has become a vital tool in neuroscience because it integrates millimeter-level spatial resolution with millisecond-level temporal resolution. Howev...
www.science.org
November 10, 2026 at 9:58 AM
A high robustness decoding technique for ultrasound modulated electroencephalogram www.science.org/doi/10.1126/... (real human EEG but not yet in vivo human transcranial USMEEG recording trials); #US #EEG #USMEEG #BCI #NeuroTech
Here’s the study: www.nasa.gov/humans-in-sp...
NASA’s Twins Study Results Published - NASA
The Twins Study results published demonstrate the resilience and robustness of how a human body can adapt to a multitude of changes induced by the spaceflight.
www.nasa.gov
November 9, 2026 at 4:44 AM
Here’s the study: www.nasa.gov/humans-in-sp...
A groundbreaking study presents Prompt-Invariant RLVR (PIRL), enhancing multimodal reinforcement learning robustness against prompt perturbations for VQA. PIRL achieves stability with minimal accuracy loss compared to prior models, ensuring reliable AI deployment. https://arxiv.org/abs/2608.08802
Improving Generalization Robustness of Multimodal RLVR
ArXiv link for Improving Generalization Robustness of Multimodal RLVR
arxiv.org
November 10, 2026 at 8:30 PM
A groundbreaking study presents Prompt-Invariant RLVR (PIRL), enhancing multimodal reinforcement learning robustness against prompt perturbations for VQA. PIRL achieves stability with minimal accuracy loss compared to prior models, ensuring reliable AI deployment. https://arxiv.org/abs/2608.08802
@neural-reckoning.org, @marcusghosh.bsky.social, and I are recruiting a postdoc in comp neuro to study how brains can be exceptionally robust to damage, using small neural networks that mimic features of Hofstenia’s diffuse brain. If you’re interested, get in touch!
www.crick.ac.uk/careers-stud...
www.crick.ac.uk/careers-stud...
Mechanisms of extreme robustness in small biological neural networks
www.crick.ac.uk
November 5, 2026 at 2:44 PM
@neural-reckoning.org, @marcusghosh.bsky.social, and I are recruiting a postdoc in comp neuro to study how brains can be exceptionally robust to damage, using small neural networks that mimic features of Hofstenia’s diffuse brain. If you’re interested, get in touch!
www.crick.ac.uk/careers-stud...
www.crick.ac.uk/careers-stud...
Our jaxQTL paper is out today in Nature Genetics! 🧬
We developed count-based models that improves single-cell eQTL discovery and reveal cell-type-specific genetic regulation.
www.nature.com/articles/s41...
We developed count-based models that improves single-cell eQTL discovery and reveal cell-type-specific genetic regulation.
www.nature.com/articles/s41...
Efficient count-based models improve power and robustness for large-scale single-cell eQTL mapping - Nature Genetics
This study presents jaxQTL, an efficient count-based pseudobulk single-cell eQTL mapping framework. When applied to OneK1K data, jaxQTL discovers more eGenes than existing methods while running faster...
www.nature.com
November 10, 2026 at 8:18 PM
Our jaxQTL paper is out today in Nature Genetics! 🧬
We developed count-based models that improves single-cell eQTL discovery and reveal cell-type-specific genetic regulation.
www.nature.com/articles/s41...
We developed count-based models that improves single-cell eQTL discovery and reveal cell-type-specific genetic regulation.
www.nature.com/articles/s41...