Fragments: April 9

Source: Martin Fowler | Tags: ai, documentation, podcast, programming, security | Published: 2026-04-09 | Novelty: 38%

The article discusses Simon Willison's podcast with Lenny Rachitsky providing insights into the evolution of programming, and Gergely Orosz's interview with Thuan Pham on Uber’s microservices. Additionally, it highlights a post-mortem from Axios detailing a sophisticated supply chain attack that tailored its process to an individual by impersonating their company, including fake Slack channels and meetings. The article also introduces Diátaxis, a framework for organizing technical documentation, which categorizes documentation into tutorials, how-to guides, reference, and explanations. Finally, it shares Lalit Maganti’s experience using AI to develop tools for working with SQLite, noting the importance of objective tests in the development process.

Multimodal Embedding & Reranker Models with Sentence Transformers

Source: Hugging Face - Blog | Tags: multimodal, sentence-transformers, v5-4 | Published: 2026-04-09 | Novelty: 38%

The article introduces v5.4 of Sentence Transformers with several new features including support for the Qwen3-VL embedding model, which can handle multimodal inputs like text and images together. Notable code includes using processor_kwargs to set image resolution bounds and model_kwargs for model precision settings when loading the SentenceTransformer model.

Improving the academic workflow: Introducing two AI agents for better figures and peer review

Source: The latest research from Google | Tags: ai, peer-review, research, visualization | Published: 2026-04-08 | Novelty: 30%

The article introduces PaperVizAgent and ScholarPeer, two AI agents designed to enhance academic research. PaperVizAgent generates high-quality figures from textual input with a multi-agent system approach, significantly outperforming existing tools in experiments. ScholarPeer automates peer review by dynamically constructing domain narratives and verifying technical claims through active web search, providing highly critical and literature-grounded reviews. Both systems are experimental prototypes intended for research purposes only.