The Archive

In the first 24 hours of the assault on Iran, the US military struck more than 1,000 targets, nearly double the scale of the "shock and awe" attack on Iraq over two decades ago. This acceleration was made possible by AI systems that speed up the targeting process. Chief among them is the Maven Smart System. In her new book, Project Maven: A Marine Colonel, His Team, and the Dawn of AI Warfare, journalist Katrina Manson investigates the development of Maven from its inception in 2017 as an experiment in applying computer vision to drone footage. The project spurred employee protests at Google,...

Stratechery weekly digest covering Tim Cook's Apple departure, Cursor IDE, SpaceX developments, and geopolitical competition.

The AI-driven memory shortage is hitting Samsung's bottom line.

OpenAI and Anthropic job listings show unexpectedly high go-to-market hiring relative to engineering and research roles.

Abstract Chain-of-Thought uses reserved token vocabulary for latent reasoning, matching verbal CoT performance with shorter generation.

Apple’s sold-out Mac mini is spawning marked-up eBay listings as demand surges for the compact desktop, now favored for running local AI models and tools.

Reddit user visualizes Claude adoption normalized by state population, excluding DC as statistical outlier.

Reddit user reports ChatGPT provided superior character design feedback, raising questions about creative tool utility.

Reddit discussion asking about TurboQuant KV cache optimization implementation in llama.cpp.

[https://www.anthropic.com/transparency](https://www.anthropic.com/transparency)

Parametric decomposition method for analyzing sub-breath respiratory airflow signals in medical contexts.

CRAFT: clustering-based training data selection for efficient sequence-to-sequence model fine-tuning.

Reddit user reports GPT Image 2 generates realistic candid photography with compositional nuance; anecdotal testing feedback.

Cohere preparing new MoE model with vLLM optimization support, signaling open-weights or community-accessible release.

Supply chain fragmentation in AI hiring systems obscures bias accountability across vendors, model developers, and deployers.

BERAG: Bayesian ensemble RAG for visual QA mitigating lost-in-the-middle effect and document attribution failures.

WG-SRC: white-box signal-subspace probe for interpretable graph neural network diagnosis and node classification.

Per Bloomberg: > Google will invest $10 billion in Anthropic PBC, with another $30 billion potentially to follow, strengthening the relationship between two companies that are at once partners and rivals in the race to build artificial intelligence. > > Anthropic said that Google is committing to invest $10 billion now in cash at a $350 billion valuation, the same amount it was valued at in a funding round in February, not including the recent money raised. The Alphabet Inc.-owned company will invest another $30 billion if Anthropic hits performance targets, the startup said Friday...

Reddit user reports unexpected token consumption increase on Anthropic's Claude Opus 4.7 MAX plan; possible pricing/rate change speculation.

GP-MLMPC: iterative Gaussian Process model learning for nonlinear control of batch processes.

Chinese semiconductor researcher Danhao Wang died at University of Michigan; Chinese government alleges hostile US law enforcement questioning preceded death.

Human-centered audit of eight Shapley value variants in high-stakes risk workflows reveals misalignment with practical utility.

Query Performance Prediction for selective execution of query variants in RAG pipelines reduces computational overhead.

Associativity-peakiness metric for clustering algorithm contingency table evaluation.

User reports Claude effectiveness for sysadmin tasks: log analysis, error detection, and automated remediation script generation.

DeepSeek v4 Flash outperforms Claude Haiku on tool-calling benchmarks in production chat system evals.

Reddit post comparing Claude API pricing to commodity wheat prices; anecdotal cost commentary without substantive technical or business insight.

Probabilistic framework for quality-driven selective mutation in deep learning testing combining realism and test guidance.

Adversarial malware generator for Linux ELF binaries achieves 67.74% evasion rate via semantic-preserving transformations.