Build Git-Native AI Workflows That Remember, Learn, and Improve with Every Run
Coding agents can help you solve a difficult problem, understand a codebase, and make good engineering decisions. Then the session ends, and much of that context disappears. The next session starts rebuilding knowledge, revisiting decisions, and asking questions your team has already answered.
In this free 1-hour technical session, Rick Hightower will show you how to turn those isolated agent sessions into a shared engineering system that retains knowledge and improves over time. You’ll see a practical, Git-native approach using Harness Engineering, WikiTicket SDD, project second brains, ContextPacks, and Graph Engineering to make agent work visible, preserve institutional knowledge, and feed what each run learns into the next.
What You’ll Learn
- Why agent quality can deteriorate as the volume of AI-generated work increases
- How Harness Engineering adds orchestration, control, and evaluation around coding agents
- Where to use LLM judgment and where deterministic code should take over
- How procedural, working, and long-term memory play different roles in an agentic system
- How to manage context without continuously filling the agent's context window
- How WikiTicket SDD creates a Git-native, shared source of truth for agent and human work
- How to build a project second brain that preserves decisions, experiments, assumptions, system knowledge, and ownership
- How ContextPacks provide agents with bounded, relevant information for the task at hand
- How Graph Engineering turns persistent shared state into a foundation for self-improving workflows
- How each agent run can leave behind better knowledge and context for the next run
Who Is This For?
This session is designed for:
- Engineering leads exploring how coding agents can work reliably across teams
- Staff and senior engineers already incorporating coding agents into development workflows
- Developers using tools such as Claude Code, Codex, Cursor, or other coding agents
- Platform and developer productivity engineers building AI-assisted engineering workflows
- Engineers interested in agent memory, context management, Harness Engineering, and evaluation
- Teams looking for a practical way to preserve engineering knowledge beyond individual AI sessions
Lineup
Rick Hightower
Good to know
Highlights
- 1 hour 30 minutes
- Online
Location
Online event
Agenda
1. Harness Engineering, Context & Memory
Start with the foundations of a reliable agentic system. Rick will break down what an agent harness actually does, how orchestration, control, evaluation, and hooks fit together, and why teams need clear boundaries between LLM judgment and deterministic code. You’ll also explore the three types of agent memory and how progressive disclosure helps provide useful context without overwhelming the model.
2. Making Agent Work Visible with WikiTicket SDD
Agents and humans can be working on the same project without having a reliable view of each other's state. Rick will demonstrate WikiTicket SDD, using an append-only Git-native event log and deterministic state generation to make work visible across agents, developers, and worktrees without introducing another SaaS system as the agent's source of truth.
3. Building the Project Second Brain
See how engineering knowledge can move out of temporary chat sessions and individual team members' heads into a portable knowledge graph under Git. Rick will introduce OKF, PKC, SAC, and DEKC, showing how teams can capture decisions, experiments, assumptions, services, ownership, and other institutional knowledge. You’ll also see how ContextPacks give agents bounded, task-relevant access to this information.