The Archive

Federated learning framework uses CRF-based inference to weight heterogeneous client contributions for improved convergence.

PCHS applies physically-constrained signal processing to heart/respiratory rate estimation from wrist PPG under motion artifacts.

Framework disentangles grammatical gender from semantic bias in contextual embeddings for gendered languages like Spanish.

FacePlex generates speech and synchronized facial motion jointly in real-time for conversational avatars via full-duplex tokens.

DNA Language Models benchmark compares transformers (DNABERT2) and CNNs (ConvNova) on genomic tasks, questioning pretraining costs.

Warrant mechanism enforces that attention-retrieved values only contribute to predictions if warranted, addressing relevance-permission gap.

Extends strategic classification to handle decision-dependent manipulation costs, modeling how algorithm decisions affect the economics of gaming behaviors.

Proposes query-aware spreading activation for knowledge graph RAG to improve multi-hop retrieval by maintaining query signals throughout traversal.

Introduces hyper-network neural functional maps for 3D shape matching under partiality and topological noise via unsupervised non-linear alignment.

Empirical study across 25 LLMs showing chain-of-thought gains derive from semantic content, not token length, using in-distribution paired comparisons.

Extends gauge/gravity duality via PINNs to reconstruct holographic bulk potentials under large hierarchies and false vacua in quantum field theories.

Identifies three open problems in Constitutional AI preference reconstruction: principle composition, decision rule specification, and interpretable constraint encoding.

Proposes SA-VLA, a state-aware tokenizer for Vision-Language-Action models that conditions discrete action codes on robot proprioceptive state for improved manipulation.

Introduces AgentCanvas, a framework automating embodied agent architecture search via typed-graph runtime, extending prior text-domain AAS to perceptual agents.

Proposes Physics-Anchored Certification (PHACT), a propose-certify loop where deterministic engines validate LLM-generated designs across five scientific domains.

Framework for neural network verification under imprecise probabilistic information using interval belief structures and copulas to model marginal and dependence uncertainty.

Stratechery newsletter on summer break June 29; editorial schedule notice.

OpenAI report maps EU labor market exposure to AI automation across occupations, identifying jobs at risk or poised for growth.

Meta AI develops Brain2Qwerty, a non-invasive brain-computer interface converting EEG signals to text without surgical implants.

Jon Udell argues for 'agent-assisted' framing in software development: agents join human workflows rather than replacing human authority in autonomous loops.

China's Zhipu AI (Z.ai) released its open-weight GLM-5.2, and some researchers have claimed that it matches Mythos in certain bug-finding and cybersecurity scenarios. While GLM lags behind models from Anthropic and OpenAI in other, more general tasks, it seems that China has dramatically reduced the gap in the capabilities between its models and those of the US. This level of advancement is particularly concerning to the US government, which has worked to restrict China's access to powerful models like Anthropic's Mythos and Fable, as well as the hardware necessary to train and run them. The ...

Suno has ambitions to be more than just a toy to churn out AI slop, it also wants to be a streaming destination and to break new artists. Spark is their new incubator program for independent artists that provides grants, mentorship, and marketing support. To apply, artists need to be an unsigned singer, songwriter, or producer releasing music under their own name. They also need to agree to some terms and conditions that have raised some eyebrows over on the Suno subreddit. For one, you need to agree to make your songs available on Suno for remixing. That's not necessarily super concerning, b...

Hack Your Summer is a 4-week production sprint for students launched in response to 2024 internship shortage, offering mentorship and real project work.

"Mistakenly we thought that by just introducing artificial intelligence ... that would produce a high-quality product.”

HP Inc. expands OpenAI partnership to integrate AI across customer experiences, software development, and enterprise operations.

Austria proposes hosting Anthropic operations in Europe in response to US export restrictions on AI model access.

Eager to find more public AI-related companies that may do as well as Nvidia, Wall Street investors think they've found a winner with Micron.

Jonathan Rinderknecht was facing arson charges for setting a fire on New Year's Day in 2025, which became one of the deadliest wildfires in LA history. To make their case, prosecutors turned to location data from his iPhone, security camera footage, and witness testimony. But they also turned to his ChatGPT logs. Prosecutors said that Rinderknecht had ChatGPT generate images of fire, asked the chatbot, "Why am I so angry all the time?", and ranted to it about how the wealthy were destroying the world. They also pointed to a screen recording in which Rinderknecht asked ChatGPT whether someone ...

Brief dismissive commentary on Anthropic-related concerns; lacks substantive detail or newsworthy content.

Anecdotal report that Claude now delegates tasks via instructions rather than executing code, attributed to token-cost optimization.