Methodology – Pazsit's Dev Blog

Token Efficiency for LLM assisted Development

How seven load-bearing principles across chat sessions and agentic pipelines, keep LLM dev costs manageable without degrading what the tools produce.

Occasinally I had a problem most people working with LLMs eventually run into: Long sessions forgot their own constraints. Multi-file investigations dumped thousands of tokens into the main context and never gave them back. Pipelines paid full price for content that should have been cached. None of it was the model’s fault, all of it needs only changing how you worked with the tools. The patterns below come experience and research about LLM assisted development to scale. Design choices about where to spend tokens and where not to.

One-line thesis: token efficiency is a design discipline, effectivness not being cheap.

Agile Challenges in Scale

Agile

Agile is about co-located smaller highly effective teams, which can focus and provide values in fast pace, driven by ownership and team spirit.
In Agile practices, team ceremonies (such as standups, sprint reviews, and retrospectives) are esential to improve communication, alignment, and efficiency. However, when Agile teams grow too large, these ceremonies can lose their effectiveness.
I like to highlight how large team sizes and poorly structured meetings can hinder productivity, communication, and decision-making, advocating for smaller team structures to preserve Agile’s core values.

Agile methodologies are built around small, cross-functional teams designed to work closely together, rapidly iterate, and adjust to changes. However, as teams grow in size, they often encounter communication bottlenecks and coordination challenges.

Outline of Scrum Guide 2020

I created my personal notes about the new scrum guide for a while, as the guide is out since 2020 November. I decided to summarize/outline in an article only now. However it was written in early 2021.