Unix Pipes Are Better Than Agents
Abstract: The industry is racing to give LLMs more tools, APIs, and agent frameworks. Reserve, a Go CLI for exploring FRED macroeconomic data, takes the opposite approach: enabling the tool to explain itself. Every command can expose context, intent, examples, and pipeline patterns, allowing models to discover functionality, compose deterministic workflows, and explain exactly how a result was produced. Rather than letting an agent reason over commands, flags, and raw APIs, Reserve pushes collection, transformation, and analysis into deterministic, reproducible Unix pipelines, preserving precious tokens for insight rather than orchestration. The outcome is simpler, more transparent, and surprisingly powerful. The industry keeps asking how much intelligence can be embedded in upstream orchestration, while the Unix philosophy has spent the last fifty years moving orchestration into composition.
Presentation
A PDF version of the presentation can be downloaded here. The actual talk will be posted on YouTube by the Gopher Academy.
Implementation
RESERVE is an open-source project published under an MIT License. Feedback, pull requests, or even a star are welcom at : https://github.com/derickschaefer/reserve
Architectural Proposal
The agentic discovery onboarding concepts presented at Gopher Con 2025 and its implementation in RESERVE are published in the proposal paper Repository Architecture for Agentic Systems . Feedback and contributions are welcome.
CLI Book
The creator of RESERVE is also the author of >CLI: A Practical Guide to Creating Modern Command-Line Interfaces . It is available on Amazon and at Barnes & Noble. All examples are written in Go!