Generative foundation models can remove visual artifacts on X-rays through realistic image inpainting, but they can obscure subtle, clinically relevant features and introduce latent bias. @FelipeMatsuoka @eduardomjfarina @UNIFESP_Rad @unifesp @ddiunifesp arxiv.org/abs/2511.23066

This dataset contains 3597 MR Images (MRI) of the head acquired in one of 3 possible planes (axial, sagittal, coronal) and one of 7 sequence types (T1, T1Gd, T2, FLAIR, b1000, ADC map, SWI). kaggle.com/datasets/felipeki…

Someone built a transparent Mario game that runs OVER IDE so can play while waiting for Copilot to write code.

O 4º Desafio de #IA da #SPR traz uma nova proposta: usar #NLP para interpretar laudos de #mamografia e prever categorias #BIRADS. Uma iniciativa que conecta tecnologia, ciência e prática clínica na #Radiologia. Saiba mais e participe!⬇️ noticias.spr.org.br/jpr-2026…

Judging by my tl there is a growing gap in understanding of AI capability. The first issue I think is around recency and tier of use. I think a lot of people tried the free tier of ChatGPT somewhere last year and allowed it to inform their views on AI a little too much. This is a group of reactions laughing at various quirks of the models, hallucinations, etc. Yes I also saw the viral videos of OpenAI's Advanced Voice mode fumbling simple queries like "should I drive or walk to the carwash". The thing is that these free and old/deprecated models don't reflect the capability in the latest round of state of the art agentic models of this year, especially OpenAI Codex and Claude Code. But that brings me to the second issue. Even if people paid $200/month to use the state of the art models, a lot of the capabilities are relatively "peaky" in highly technical areas. Typical queries around search, writing, advice, etc. are *not* the domain that has made the most noticeable and dramatic strides in capability. Partly, this is due to the technical details of reinforcement learning and its use of verifiable rewards. But partly, it's also because these use cases are not sufficiently prioritized by the companies in their hillclimbing because they don't lead to as much $$$ value. The goldmines are elsewhere, and the focus comes along. So that brings me to the second group of people, who *both* 1) pay for and use the state of the art frontier agentic models (OpenAI Codex / Claude Code) and 2) do so professionally in technical domains like programming, math and research. This group of people is subject to the highest amount of "AI Psychosis" because the recent improvements in these domains as of this year have been nothing short of staggering. When you hand a computer terminal to one of these models, you can now watch them melt programming problems that you'd normally expect to take days/weeks of work. It's this second group of people that assigns a much greater gravity to the capabilities, their slope, and various cyber-related repercussions. TLDR the people in these two groups are speaking past each other. It really is simultaneously the case that OpenAI's free and I think slightly orphaned (?) "Advanced Voice Mode" will fumble the dumbest questions in your Instagram's reels and *at the same time*, OpenAI's highest-tier and paid Codex model will go off for 1 hour to coherently restructure an entire code base, or find and exploit vulnerabilities in computer systems. This part really works and has made dramatic strides because 2 properties: 1) these domains offer explicit reward functions that are verifiable meaning they are easily amenable to reinforcement learning training (e.g. unit tests passed yes or no, in contrast to writing, which is much harder to explicitly judge), but also 2) they are a lot more valuable in b2b settings, meaning that the biggest fraction of the team is focused on improving them. So here we are.

SOMEONE MADE A DIGITAL WHIP TO MAKE CLAUDE WORK FASTER 💀

Tesla self-driving saves a lot of lives – the statistics are unequivocal. That doesn’t mean it’s perfect, of course. Even when we improve safety 10X, saving 90% of the million lives lost in auto accidents every year, Tesla will still get sued for the 10% who did die. The 90% who are still alive mostly won’t even know that Tesla saved them. Nonetheless, it is the right thing to do.

If you think AI video can't be funny then explain this. 😹

“A friend left a party at 9:30 on a Saturday. Not tired. Not sick. He wanted to get back to his agents. Nobody questions it anymore. Half the room is thinking the same thing. The other half are probably checking the progress of their agents. At a party” writing.nikunjk.com/p/token-…

A Note of Thanks to our wonderful reviewers for their time and expertise! doi.org/10.1148/ryai.260176 #reviewers #ML #MachineLearning

Supporting communication in Radiology AI with ROADMAP, a new ontology doi.org/10.1148/ryai.260069 @cekahn @abhisuri97 @_ragonzales #Radiology #ontology #MachineLearning

Supporting communication in Radiology AI with ROADMAP, a new ontology doi.org/10.1148/ryai.260069 @cekahn @abhisuri97 @_ragonzales #Radiology #ontology #ML

The Checklist for AI in Medical Imaging addresses #AI applications that include classification, segmentation, and reconstruction of medical images rsna.org/claim #MachineLearning #Radiomics #ArtificialIntelligence

New data resource with 4-view 2D mammograms and corresponding BI-RADS assessment, breast density, machine ID, and proof of benign or malignant outcome doi.org/10.1148/ryai.250375 @FelipeKitamura @haritrivedimd @emoryradiology #breast #BreastRad #cancer

A new narrative review describes the evolution of AI applications for radiology from LLMs to autonomous agents doi.org/10.1148/ryai.250651 @k_bressem @ShahriarFaghani @Khosravi_Bardia #agent #agents #AgenticAI

Karpathy's llama2.c showed you could train a real transformer in pure C with no frameworks. A solo researcher (and Claude Code) just took that same model, Stories 110M, Llama2 architecture, trained on real text and ran it on Apple's M4 Neural Engine (ANE) for less than a watt. He reverse-engineered the undocumented private APIs, bypassed CoreML, and found Apple's abstraction layer was hiding 2-4x of the chip's real throughput. The ANE delivers 6.6 TFLOPS per watt, roughly 80x more efficient than an Nvidia A100. The real implication here is inference: there are hundreds of millions of Apple devices with one of the most efficient AI accelerators ever shipped in consumer hardware, and Apple's own software stack is the thing standing between developers and its actual performance. h/t @maderix

" "Trying to predict the future is a discouraging and hazardous occupation... The only thing we can be sure of about the future is that it will be absolutely FANTASTIC " -Arthur C. Clarke - Starlink 🛰

Establishing normative, age-based, pediatric kidney length and volume percentiles from a large US dataset with a natural language processing model doi.org/10.1148/ryai.250056 @ChildrensPhila #USRad #AI #MachineLearning

New data resource with 4-view 2D mammograms and corresponding BI-RADS assessment, breast density, machine ID, and proof of benign or malignant outcome doi.org/10.1148/ryai.250375 @FelipeKitamura @haritrivedimd @emoryradiology #mammogram #breastcancer #MachineLearning

Even in the age of digital articles, conciseness matters radiologyai.substack.com/p/w… #sciencewriting #technicalwriting #MachineLearning