hrbrmstr's Daily Drop

  • Drop #528 (2024-09-12): Happy ThursdAI

    Another ThursdAI Drop is upon us! In it, we cover Reader-LM, a small language model by Jina AI, for HTML-to-Markdown conversion; Janelle Shane’s exploration of AI-generated paint colors; Jon Evans’ frustrations with large language models (LLMs); and Space-Man, a showcase game using OpenAI’s new technologies.

    hrbrmstr

  • Drop #523 (2024-08-29): Happy ThursdAI!

    Today’s Drop covers creating a custom Ollama model to generate concise names for CVE vulnerabilities and discusses concerns about Google’s new Pixel 9 “Reimagine” feature. Additionally, it introduces “The Inference,” an editorial project by Danny Palmer exploring the impact of AI on cybersecurity and society

    hrbrmstr

  • Drop #519 (2024-08-22): Cache Me If You Can

    Today’s Drop discusses creating a lightweight internet research setup by integrating SearXNG, Jina, and a small Go program to provide a caching layer for search results. SearXNG is a self-hostable metasearch engine, while Jina’s Reader endpoint converts URLs to Markdown. Through a combined CLI, we can preview search results in the terminal and access cached…

    hrbrmstr

  • Bonus Drop #57 (2024-08-03): Vector Similarly Search And RAG-Lite

    The weekend Bonus Drop concludes a miniseries on vector search operations with VSS in R. But wait! There’s more! We also work with Ollama and three models to build a “lite” retrieval augmented generation application over the CISA KEV corpus. #RStats

    hrbrmstr

  • Drop #514 (2024-08-09): DuckDB Vector Search

    Today’s Drop showcases DuckDB’s nascent vector search superpowers by walking through an example of how to generate embeddings for CISA KEV vulnerabilities, performing similarity search, and having Ollama guess good titles for clustered groups of KEV vulns. Tis def a code block-heavy post. #DuckDB #RStats

    hrbrmstr

  • Drop #513 (2024-08-08): TUI Thursday

    Today’s Drop covers three terminal user interface (TUI) tools! “oatmeal” is a clean interface for chatting with large language models, “discovery-rs” monitors multicast DNS services, and “otree” navigates JSON, TOML, and YAML files.

    hrbrmstr

  • Drop #511 (2024-08-06): Typography Tuesday

    Today’s typography-centric edition of the Drop talks about the FontJoy Embedding Projector, a visualization tool for font data exploration. We also 👀 into the lost and rediscovered Pouchée alphabets from the 1820s. And, it features LCT. Sinar, a contemporary free font with a reverse contrast variable style.

    hrbrmstr

  • Drop #510 (2024-08-02): SQLite Vector Search

    Today’s Drop covers the introduction of two SQLite extensions, sqlite-vec and sqlite-rembed, designed to facilitate vector search operations. These extensions provide fast and efficient vector search functionality, sponsored by the Mozilla Builders project, and enable easy integration of text embeddings from “remote” APIs (which included ollama/llamafile). With further development potential, these extensions aim to democratize…

    hrbrmstr

  • Drop #508 (2024-07-31): Wild, Wild, Wednesday

    The Drop is caught up to Wed! Trail of Bits conducted a security audit of Homebrew, uncovering some concerns including executable code vulnerabilities and CI/CD issues. We look at the highlights in the Drop and shunt readers to the completel report for more details. I expanded my Mastodon-thread of the initial features of Apple Intelligence…

    hrbrmstr

  • Drop #503 (2024-07-23): Typography Tuesday

    Today’s typography-centric edition of the Drop covers Eco fonts, which are designed to reduce ink consumption and environmental impact in printing. It also takes a look at AI-driven font recommendation systems thar=t are emerging to simplify font selection (include links to a spiffy open model, too).

    hrbrmstr