Mark Zuckerberg was right 2025 brought us mid life engineer level work via AI coding tools.
Following up on my 2025 workflows, this post explores the shift from manual prompting to agentic feedback loops. I break down why model choice now outweighs prompt engineering, how Claude Code has become my primary driver alongside Cursor, and why the "SWE II" level of AI requires a new focus on automated eval loops and organizational standards.
Following up on my 2025 workflows, this post explores the shift from manual prompting to agentic feedback loops. I break down why model choice now outweighs prompt engineering, how Claude Code has become my primary driver alongside Cursor, and why the "SWE II" level of AI requires a new focus on automated eval loops and organizational standards.
Pros and cons lists break down quickly in complex engineering decisions. Decision matrices create alignment, make tradeoffs explicit, and lead to clearer system designs.
As a staff engineer at an early-stage startup, I rely on LLM-powered coding assistants like Cursor, Claude, and ChatGPT to move faster while writing less code by hand. This post shares practical workflows, prompting strategies, and principles for using AI tools effectively in real-world software engineering.