Articles from seangoedecke.com

The author argues that in large software systems it's normal and effective to work with only a partial understanding of the codebase.

Software engineers should deliberately operate at 80% capacity to remain free for high-impact, time-sensitive opportunities rather than grinding through low-priority tickets.

A staff engineer now uses LLM agents for nearly all code changes, bug hunting, and testing, with light supervision, but avoids AI for communication and UI work.

The software engineering industry may not survive another decade as AI agents become capable of writing and maintaining code, leaving human engineers with diminishing roles.

Giving LLMs human-like personalities is a necessary engineering choice, not a deception, because post-training personalities make base models capable and useful.

Speculation about internal tech company problems is almost always wrong because outsiders lack the insider context needed to understand the real issues.

Anthropic's fast mode uses low-batch-size inference on the full model, while OpenAI's uses a smaller distilled model on Cerebras chips for much higher speed.

A 2025 Anthropic study shows AI-assisted coding can reduce skill retention but 25% speed boost emerges when ignoring participants who manually retyped AI code.

The author argues that his love for software engineering stems from an internal compulsion to be useful, and advises harnessing this drive for career success.

Effective software design in large existing systems requires deep codebase familiarity, making generic design advice largely useless for practical problems.

Large software products are so complex and rapidly changing that even engineers often cannot answer basic questions about them without investigative research.

Frequent team rotations and overloaded senior engineers cause big tech companies to write surprisingly sloppy code.

Large software companies sacrifice efficiency for legibility to control scale and enable enterprise deals, despite slowing engineering velocity.