The post discusses the challenges of generating structured JSON outputs using language models (LLMs) and introduces the concept of 'Structured Output' parsing. It explains how LLMs tokenize input strings and generate predictions b...
About:
A developer's blog featuring thoughts, ideas, and all things engineering.
Website:
Outgoing Links:
Subscribe to RSS:
Microsoft's phi-4 reasoning model, a 14B parameter model, utilizes supervised fine-tuning and synthetic dataset curation to outperform larger models like DeepSeek-R1-Distill-Llama-70B. The model is trained on high-quality datasets...
The author shares their experience building a PC aimed at machine learning, focusing on the Antec Flux Pro case with excellent airflow. They detail the process of reverse engineering the temperature display software, which is only...
The post explores the behavior of various language models, particularly in relation to the seahorse emoji phenomenon. It discusses tests conducted on models like microsoft/Phi-4-mini-Instruct and HuggingFaceTB/SmolLM2-135M-Instruc...
The blog post provides a detailed review of the GMKTec Evo X2 hardware, focusing on its performance for local model inference. The author shares their experience with installing Ubuntu Server, testing various AI models using the O...
The DORA State of AI Assisted Software Development report analyzes AI adoption in the software industry based on a study of 5,000 participants. It highlights trends in AI tool use, productivity perceptions, and trust in AI-generat...
The blog post discusses the Model Context Protocol (MCP) and its implementation in Claude's desktop app, detailing how to configure and troubleshoot server setups using a JSON configuration file. The author shares a personal exper...
The author describes a situation where a routine maintenance task on their Ubuntu server interrupted a database migration for GoToSocial, leading to an error. They detail the recovery process using VPS snapshot backups and share a...
The blog post discusses OpenAI's recent announcement of the AppSDK for ChatGPT, positioning it as a platform similar to app stores. It details how developers can create custom UI components that integrate with ChatGPT through MCP ...
The article discusses the impact of AI on the software industry, contrasting the roles of professional engineers and 'script kiddies' who lack deep understanding. It emphasizes the importance of expertise in problem-solving and de...
The blog post explains the concept of vector norms, which describe the magnitude of a vector. It covers several types of norms including the Euclidean norm (L2), Manhattan norm (L1), generalized norm (L-p), and max norm (L-infinit...
The blog post explains how to add Gaussian noise to a signal using Python, specifically with the numpy library. It details the process of generating noise that follows a Gaussian distribution and demonstrates this through code exa...
4.8Notes on the DuckDB UI
2025-03-14 • notebook duckdb ui framework data visualization sql
DuckDB has introduced a new local UI package that allows users to visualize databases and run SQL queries through embedded notebooks. The UI can be started via the terminal, and it operates on a familiar code execution model simil...
The post explains vector projection, detailing how to project one vector onto another using mathematical formulas and Python code. It includes the definition of vector projection, the formula for calculating the projection, and pr...
The article explains vector normalization, a mathematical process that scales a vector to have a magnitude of 1 while maintaining its direction. It provides the formula for normalization, examples of calculating the magnitude of a...
Microsoft recently released the phi-4 reasoning model and its technical report, which explores using supervised finetuning and synthetic dataset curation to train a 14-billion parameter model. The model is trained on high-quality ...
The author encountered an issue with a server, interrupting a database migration process, and had to recover an older snapshot of the database. They then created a backup solution using systemd timers and Rust, and packaged it as ...
The text explains how to add Gaussian noise to a signal in Python using numpy. It provides code examples and visualizations to demonstrate the process.
The author built a new PC aimed at machine learning and wanted a case with great airflow. They found a case with a display that shows CPU and GPU temperatures, but the software to run the display is only available on Windows. The ...
0Notes on the DuckDB UI
2025-03-14 • mysql user interface design notebook duckdb sql
DuckDB recently released a new local UI package that allows users to visualize DuckDB databases and run SQL queries through embedded notebooks. The UI can be started by running duckdb -ui in the terminal, which starts an http serv...
0Vector Projection
2025-03-02 • mathematics resources map projections
Vector projection is the process of projecting one vector onto another, and it is the adjacent side of a right triangle with the original vector as the hypotenuse. The formula for the projection of vector u onto vector v is given ...
0Vector Normalization
2025-03-02 • programming mathematics resources
Vector normalization is the process of scaling a vector to have a magnitude of 1 while keeping the direction the same. The formula for vector normalization is provided along with an example. The process can be implemented in code ...
The impact of AI on the software industry is a topic of debate. The author discusses the value of expertise in the industry, comparing professionals to script kiddies. The author believes that generative AI has the potential to cr...
The text provides notes on setting up Claude Desktop MCP servers. It explains how Claude's desktop app works and how to launch servers declared in claude_desktop_config.json on startup. The author encountered an issue with the ser...