The Recap
Nora Tech Plan was introduced as a way for engineers to generate AI-driven technical plans that break work into phases, outline architecture changes, and guide implementation. It allowed engineers to move from a high-level idea to a structured plan that could feed directly into downstream tools like Nora PR.
While the core capability worked well, consistent feedback emerged from engineers who used it in practice. Plans were often very long, entirely text-based, and time-consuming to review, especially during the first pass. Important TDD elements like cost estimation and customer impact were missing, and open questions required manual follow-up. On top of that, because Nora Tech Plan ran purely in the CLI, there was no visibility into how often it was being used or where teams were getting stuck.
This second edition focuses on turning Nora Tech Plan from a powerful but heavy tool into one that is easier to review, easier to refine, and easier to scale across the organization.
The New Capabilities
The biggest change in Nora Tech Plan v2 is the introduction of a two-level structure. Each plan now begins with a concise, high-level section that acts as an executive summary. This first layer is designed to help reviewers quickly understand whether the proposed approach makes sense before diving into implementation details.
That high-level section also includes AI-generated diagrams, such as UML or state diagrams, produced directly by Nora. These visuals make it much easier to evaluate architectural decisions at a glance, rather than parsing through pages of prose.
Below the summary, the plan expands into a detailed implementation guide. This section contains the full technical depth (phases, code snippets, and contextual details, etc.) that engineers or AI agents can use directly when moving into execution with Nora PR.
Another major improvement is the integration of interactive questioning. Instead of listing open-ended questions at the end of the plan and requiring the user to manually track and answer them later, Nora Plan now uses Claude Code’s Ask User Question tool to prompt engineers inline. This allows the plan to be refined dynamically during generation, reducing back-and-forth and improving accuracy.
Finally, Nora Plan v2 introduces observability. Usage is now tracked through Nora’s monitoring infrastructure, making it possible to understand adoption trends, identify friction points, and continuously improve the tool based on real usage data.
The Impact
These changes significantly reduce the cognitive load of reviewing AI-generated plans. Engineers can quickly assess whether a plan is on the right track using the high-level summary and diagrams, then drill down only if needed.
By adding missing sections like cost estimation and customer impact, Nora Tech Plan aligns more closely with traditional TDD expectations while still remaining AI-native. The interactive questioning removes a common source of friction, and observability gives the team real insight into how the tool is being used across the company.
The result is higher trust in the output, faster reviews, and increased willingness to adopt Nora Plan as a default part of the engineering workflow.
What’s Next
With the structural improvements in place, the next focus areas include continuing to refine diagram quality, expanding cost and capacity modeling, and using observability data to automatically tune plan verbosity based on user behavior.
Longer term, Nora Tech Plan is moving toward becoming the standard entry point for AI-initiated development at Abnormal, where plans are not only generated and reviewed, but seamlessly handed off to execution tools like Nora PR with minimal friction.
The second edition brings Nora Tech Plan closer to that vision by making AI-generated planning not just powerful, but genuinely usable at scale.
