Lore: A Deterministic System of Record for Coding Agents
Lore, a new tool for coding agents, aims to solve the problem of agents re-doing things that teams have already ruled out. By keeping a team’s recorded knowledge, including requirements, decisions, designs, roadmaps, and prompts, in a deterministic system of record, Lore enables agents to cite team decisions instead of violating them. This approach mirrors the shift towards “infrastructure as code” and “policy as code” in the DevOps and cloud native communities, where configuration and rules are managed as version-controlled code.
Lore’s architecture is built on top of Requirements as Code (RAC), an open-source engine that provides a deterministic system of record for team knowledge. RAC uses Markdown files in a Git repository to store knowledge, which is then served read-only to coding agents over the MCP protocol. This approach ensures that retrieval is deterministic and reproducible, eliminating the need for fuzzy retrieval mechanisms like RAG or embeddings.
By using Lore, teams can enforce consistency between their knowledge and the decisions made by their coding agents. This is particularly important in industries where regulatory compliance or safety requirements demand strict control over the decision-making process. Lore’s deterministic approach ensures that agents can be held accountable for their decisions, reducing the risk of errors or non-compliance.
RAC: The Open-Source Engine Underneath Lore
RAC, the open-source engine underneath Lore, provides a library for managing team knowledge as code. RAC writes to the same carrier format as Google’s Open Knowledge Format (OKF), a Git tree of Markdown with YAML front matter, but adds write-time enforcement in CI. This ensures that malformed artifacts, broken links, and references to superseded decisions are rejected deterministically before the knowledge lands.
RAC’s public surface is exposed through the `rac.__all__` library, which provides a range of functions for working with team knowledge. The `rac validate` and `rac relationships –validate` functions reject malformed artifacts and broken links, ensuring that the knowledge stored in RAC is accurate and up-to-date.
RAC’s architecture is designed to be extensible, with a range of connectors and tools available for integrating with other systems. The `rac-import` agent skill, for example, turns one existing document into one valid artifact, with a human-review step before anything is written.
Winners and Losers in the Shift to Deterministic Coding Agents
The shift to deterministic coding agents, enabled by Lore and RAC, will have significant implications for teams and organizations. Those who adopt this approach early will benefit from improved consistency and accountability in their decision-making processes. However, teams who fail to adapt may find themselves struggling to keep up with the demands of regulatory compliance and safety requirements.
Adjacent markets, such as the market for coding agent platforms, will also be affected by this shift. Platforms that fail to integrate with deterministic systems of record like Lore may find themselves at a competitive disadvantage. Supply chain actors, such as those involved in the development of coding agents, will need to adapt to the new requirements and constraints imposed by deterministic systems of record.
Job categories, such as those involved in knowledge management and decision-making, will also be impacted by this shift. Professionals who are able to work effectively with deterministic systems of record will be in high demand, while those who are not may find themselves struggling to adapt.
The Skeptical Case: Will Deterministic Coding Agents Replace Human Judgment?
One potential criticism of Lore and RAC is that they may be too rigid, too inflexible, and too reliant on deterministic systems of record. Will these systems replace human judgment and creativity, or will they simply augment and support it? The answer to this question depends on how these systems are designed and implemented.
Historically, attempts to automate decision-making have often failed due to the complexity and nuance of human judgment. However, Lore and RAC take a different approach, using deterministic systems of record to augment and support human decision-making rather than replace it. By providing a clear and transparent record of team knowledge and decisions, these systems can help to reduce errors and improve consistency, while still allowing for human creativity and judgment.
The Signal to Watch Next: Adoption Rates and Regulatory Response
The next signal to watch in the development of Lore and RAC will be adoption rates and regulatory response. Will teams and organizations adopt these systems quickly, or will they be slow to adapt? How will regulatory bodies respond to the shift towards deterministic coding agents, and what new requirements and constraints will they impose?
One key indicator to watch will be the adoption rate of Lore and RAC among early adopters, such as those in the fintech and healthcare industries. If these systems prove successful in these industries, it is likely that they will be adopted more widely, leading to a shift towards deterministic coding agents across the board.
Pick one tactic from this post and apply it today. Which one will you start with?
By Daniel Cross, Digital Growth Strategist at TrendFlashy
Ready to launch your own asset?
Check out our guide on Building a Profitable Online Business.
