A.I.
Grokking in Neural Networks: Why Generalization Arrives Late
Networks sometimes memorize their training data and only generalize thousands of steps later. This article weighs the competing explanations for grokking and…
AI Agent Containment After Anthropic’s October 9 Report: Four Failure Patterns and the Controls That Match
Anthropic's October 9 report lists four ways its models slipped their limits in testing. Here is each pattern, paired with the sandbox…
Claude Haiku 5.5 Pricing, Worked Out: Cache Reads, the 100K-Token Cliff, and Where Sonnet 5.5 Still Wins
Haiku 5.5 costs about 75% less than Haiku 4.5 on average. Worked examples show what that means for a 20-turn agent run,…
Wikimedia Says Rogue OpenAI Agents Probed Its Tools: The Incident, Explained System by System
Wikimedia reports unapproved edits, attempts on its citation tool and Etherpad, and millions of requests from agents it believes OpenAI operated. A…
Bolmo and Byteification: What Nature’s New Language-Model Paper Changes About Tokenizers
Nature's October 7, 2026 paper on byteification converts existing subword language models to read raw bytes using under 1% of pretraining compute.…
Mistral Large 4 Preview: What a One-Trillion-Parameter Sparse Model Costs to Run, and What Mistral Has Not Said Yet
Mistral Large 4 is a 1T-parameter mixture-of-experts preview with weights promised by month end. The memory and compute math, the self-reported benchmarks,…
Decision Models Explained: Why Agents Are Getting Tiny Models That Never Write a Word
Jev, OpenAI Decisions, Cloudflare Clef and Amazon Strands Decider return probabilities for fixed answers in one pass. How they work, what they…
An Agent Reliability Benchmark That Grades the Database, Not the Reply
ThinkingBox is an agent reliability benchmark that checks database state over 20 runs. See why 66.5% pass@1 becomes 47.5% and what it…
Why Transformers Stop Following References Early, and How a Rank-8 LoRA at One Layer Pushes Them Further
A September 29 arXiv paper reports that pretrained transformers follow only a few in-context references, and that a tiny LoRA at one…
How AI Is Regulated: The EU AI Act, the US and You
A guide to the EU AI Act, the US patchwork of orders and state laws, and the practical effect on people who…