Parse notes
In 2025, PostgreSQL continued its market dominance, shown by acquisitions from major tech companies and new distributed projects. The year also saw the widespread adoption of Anthropic's Model Context Protocol (MCP) for LLM interaction with databases, and a new intense battleground for new open-source columnar file formats challenging Parquet.
SQL parsers convert SQL text into tree structures that computers can analyze. There are many different parsers because every database has its own dialect quirks, so no single parser handles everything perfectly. If you just need to analyze queries, extract lineage, or convert between dialects, then a SQL parser is good for that, while a query engine is needed for actually running queries.
Functors, Applicatives, and Monads are common patterns developers already use. A Functor allows applying a function to a wrapped value (like `Maybe.map`). An Applicative combines multiple wrapped values with a multi-argument function (like `Maybe.map2`). A Monad handles chaining operations where each step returns a wrapped value, preventing nested containers (like `Maybe.andThen`).
This dev built an imaginary programming language, Beep, over two weeks in close cooperation with Claude Code and Opus 4.5. Claude was a great programming partner, providing efficient solutions to complex design challenges like lexical shadowing and dynamic scope mutation, and easily handling mechanical refactors and obscure parser combinator tasks. While Claude occasionally struggled with issues requiring changes to external libraries or specific, non-obvious incantations, its failure rate was low.
AI has increased code generation speed but has made code verification more critical, with over 30% of senior developers now shipping mostly AI-generated code that contains 75% more logic errors and 45% security flaws compared to human-written code. Solo developers can move at "inference speed" by relying heavily on automated testing (>70% coverage) as safety nets, while teams still require human code review for context, security, and knowledge transfer.
Advancing beyond a senior software engineer role to staff or principal levels requires expertise, visibility, and intentionality. Deliberate planning and strategic action are necessary for career progression, going beyond simply accumulating skills. Engineers should use AI where possible to increase the speed of accumulating expertise and visibility.
Sponsored
Modern B2B organizations operate across multiple systems, APIs, and data models. Without a clear approach to integration and data extensibility [10], even well-designed architectures become hard to scale and expensive to maintain. Learn how to build a future-ready B2B stack on the Elastic Path blog [10].
Repogrep is an ultra-fast AI coding agent designed for codebase search. It allows users to search across any public GitHub repository. Users can input a GitHub repo URL or search by keyword to explore codebases.
Tailsnitch is a security auditor for Tailscale configurations designed to scan tailnets for over 50 misconfigurations, overly permissive access controls, and security best practice violations. It helps users identify critical and high-severity issues across various categories like access, authentication, and network security.
Claudish is a CLI tool that lets developers run Claude Code with any AI model by proxying requests through a local Anthropic API-compatible server. It works with multiple providers as well as local models. Claudish can run agents in parallel, with each instance getting an isolated proxy.
"Anti-frameworkism" is the practice of starting web development projects with native browser APIs rather than defaulting to JavaScript frameworks like React or Vue. Modern browsers now handle most problems that frameworks were originally designed to solve, including component encapsulation (Web Components), styling (modern CSS), and network requests (Fetch API), making frameworks often unnecessary.
In late 2024, AI leaders like Sam Altman and Kevin Weil predicted that 2025 would be the "Year of AI Agents," where these systems would join the workforce and perform multi-step tasks like human employees. However, this hype didn't translate into reality, as released AI agent products weren't as effective as expected and fell significantly short of their promised capabilities. Large language model technology isn't yet robust enough for full human replacement.
To earn a promotion, consistently demonstrate you can handle the responsibilities of the next-level role by proactively addressing broader team problems with solutions over a sustained period of time.
This dev switched from VSCode to Zed, which he found to be a fast, stable, and familiar IDE, despite a minor initial configuration challenge for Python language server settings.
The Concise TypeScript Book is a free, open-source resource that provides an overview of TypeScript's capabilities, from its type system to advanced features, for both beginners and experienced devs.
AI is transforming the traditional to-do list into a "done list" by automating the execution of tasks, shifting human focus to defining clear intent and evaluating outcomes.