About:

Hello there! I’m Ricardo, and here are some facts about me and things I’ve done/do: Medical doctor, Self-taught software engineer, Working on data engineering and machine learning projects.

Website:

Specializations:

Outgoing Links:

Subscribe to RSS:
The author expresses frustration with Codex's command permissions, which lack configurability for allowed or denied commands. After several failed attempts to create a custom solution, the author discovers the DEBUG trap in bash, ...
Utilize Preact without a build step to simplify development for small projects, featuring routing and state management with signals.
LLM agents, likened to 'Oompa-Loompas', are transforming programming by allowing developers to focus on high-level thinking and creativity rather than routine coding tasks.
Mastering LLM coding tools demands active practice and personalized experimentation rather than simply copying existing prompts or techniques.
A two-step pipeline for structured outputs improves efficiency and reduces errors in data extraction from language models by separating reasoning from formatting.
The post discusses the drawbacks of appending data to files, such as issues with multiple threads, race conditions, and difficulties in data management. It advocates for creating new files for each event or log entry, highlighting...
A custom Python hook for Claude Code enhances file search functionality, addressing issues with git worktrees and improving search accuracy and flexibility.
The post explains how to run Ansible playbook tasks in parallel using a simple strategy. It requires defining a list of tags to run in parallel and using a ThreadPoolExecutor to create a thread for each tag. The main thread waits ...
The author explains the benefits of using executable scripts instead of shell aliases, including not needing to source .bashrc, faster iteration, availability in more places, easy organization, and simple deployment.
The author discusses how to use the Cursor IDE to make better references to official documentation and write better code by downloading official docs directly to the machine and using them as a knowledge hub.
The author discusses the need to bundle artefacts from a project into a single file for easier management, versioning, and distribution. They experiment with ZIP files and provide code and explanations from their experiments. The ...
The author discusses the htl library, which uses template literals to return a regular DOM Node object, making it easier to manipulate the object before appending it to the current document. The library is a single index.js file, ...
The text discusses handling keyboard shortcuts in JavaScript, including a function that returns another function to check if a keyboard event matches specified key combinations. It also provides examples and a self-contained HTML ...
The text explains how to query a list of Python dictionaries using DuckDB without converting to another format. It provides a longer script, including timings, and different approaches.
The text explains how to run cron-like jobs inside a FastAPI app, using recurring background tasks as part of the app. It also provides a caution about running multiple workers or multiple container replicas at once.
The text is about the author's exploration of web components, mainly lit, tonic, and facet. The author built a simple component inspired by tonic and lit-html. The text includes code, explanations, and example usage.
DuckDB version 0.10.0 has been released, bringing new array functions that can be used to turn DuckDB into a vector database. The text explains how to insert data from numpy arrays and query the data to get the top 10 most similar...
The post discusses how to build HTML components from Python functions, similar to React, using lxml. It covers generating HTML from Python objects, creating views, adding utilities, components, and state, and compares performance ...
The text provides Python snippets for timing and profiling functions using the standard library. It includes context managers and decorators for profiling and timing.
The post discusses the idea of replacing FFI with a CLI, using a subprocess instead of a dependency, and the advantages and drawbacks of this approach. It also mentions the use of serialization and deserialization for interoperabi...
The text explains a pattern to improve parameter handling in shell scripts using parameter expansion. It also shows how to fake named function parameters and use the pattern in Makefiles.
The text provides a starter template for Python projects using Makefiles, including basic configuration, using the virtual environment, setting PWD and the repo root, default goal and help message, injecting paths into PYTHONPATH,...
The text explains how to use IPython for timing and profiling Python functions. It also provides utilities to compare different implementations and their relative performance improvements.