March 2026 Summaries

In this reflection from the SageOx team, CTO Ryan Snodgrass discusses the evolving nature of workflows and the importance of constantly refining tools and techniques to stay efficient. He highlights the use of tools like Beads, which allows for breaking down tasks into manageable epics to efficiently coordinate multiple agents on a project, and Conductor, which facilitates managing work trees and tabs, making complex projects more navigable. Ryan emphasizes the value of creating expert and sub-agents tailored to specific skills and tasks, often utilizing both ChatGPT and Claude for optimal results. He likens the customization of these agents to hiring specialized personnel, underscoring the importance of contextually relevant expertise. Additionally, Ryan discusses the significance of regularly updating skills and techniques to avoid falling behind, using recent projects as a source of insights for developing new skills and enhancing existing ones, thereby continuously improving workflow efficiency.

A shift in developer tools is underway as coding agents, rather than human users, increasingly engage with command-line interfaces (CLIs) by generating plausible commands without relying on documentation, leading to a new understanding of user experience. This concept, known as "Agent UX," is inspired by the idea of "desire paths," where agents create unplanned command patterns that reveal potential features. Instead of viewing agent errors as failures, innovative teams use them as data for improvement, exemplified by the development of FrictionAX, a library that analyzes failed commands to refine software via suggestion systems, structured responses, and telemetry collection. This approach has led to rapid feature development at companies like SageOx by capitalizing on patterns surfaced through friction data, highlighting a shift where human judgment becomes more valuable even as intelligence becomes cheaper. Complementary to FrictionAX, AgentX enables CLIs to recognize and adapt to specific agents, maintaining a clean user interface while supporting agent learning. This paradigm, coupled with the emerging "Agent Experience Platforms," positions companies to thrive by addressing the needs of agents, potentially transforming the industry much like PostHog did for web analytics.

SageOx's development journey emphasizes the integration of agent UX concepts and human judgment to rapidly iterate and deliver features. The friction dashboard, which logs agent interactions, identifies user needs and informs development priorities by capturing agents' mental models of tool functionality. This approach led to the release of Ox CLI v0.3, which included team context search using a combination of BM25 and vector search, and v0.4, which added codebase search capabilities. The development process, inspired by Steve Yegge's Desire Paths methodology, relies on agents generating code while domain experts guide decision-making, emphasizing the value of human judgment over mere intelligence. This strategy aims to transform traditional monolithic codebases into modular, queryable units that are more manageable and comprehensible for both agents and humans, with Ox CLI serving as a critical tool for navigating and understanding these modules.

In the rapidly evolving tech landscape of 2026, SageOx exemplifies a paradigm shift in software development, achieving what previously took seven person-years at XetHub in just three person-months, thanks to an AI-native workflow during the Claude Opus 4.5 era. The team's journey from a 10x to a 40x speedup involves a strategic, phased approach: initially embracing accelerated experimentation on customer features and new business models, followed by a transformative bootcamp encouraging team members to innovate boldly and restructure into small, highly autonomous units. A critical aspect of maintaining velocity is the comprehensive documentation and open sharing of all project-related information, allowing both human and AI team members to access and leverage institutional knowledge effectively. By institutionalizing a culture of relentless experimentation and adaptability, the SageOx framework ensures that teams remain agile in the face of constant technological advancements, with a focus on leveraging the latest models to revisit and successfully tackle previously challenging projects.