November 2025 Summaries

In the evolving landscape of software engineering, traditional DORA metrics are no longer sufficient due to the impact of AI, leading to the introduction of AI3 metrics to better assess AI's effectiveness within teams. These metrics include AI Output Percentage, which measures how much of the code is AI-generated, Cost, which evaluates the financial expenditure on AI tools per engineer, and Turnover, which tracks the rate at which code is rewritten or deleted within 30 days. AI3 metrics provide a comprehensive understanding of AI's role, allowing teams to make informed decisions about AI's contribution to productivity and cost-effectiveness. By analyzing these metrics, teams can discern patterns in AI usage, connect financial expenditure to output, and evaluate the quality of AI-generated code, ultimately aiming to optimize AI's integration into their workflows.

Counting Lines of Code (LOC) is a traditional yet flawed metric for measuring developer productivity, akin to judging a chef by their number of pots. While LOC can quantify the size of a codebase, it does not reflect the quality or efficiency of the work done, as different programming languages vary in verbosity and effective solutions often involve less code. Using LOC as a productivity measure can incentivize undesirable practices, such as writing overly complex code or avoiding necessary refactoring, due to Goodhart's Law, which suggests that when a measure becomes a target, it ceases to be useful. Instead of focusing on the volume of code, productivity should be assessed based on outputs like problem-solving, system robustness, and codebase health, employing advanced tools like Weave that leverage machine learning to gauge the complexity and time required for tasks. As the industry evolves, moving beyond outdated metrics like LOC to prioritize quality, maintainability, and real-world impact is essential for true engineering productivity.

Many teams mistakenly rely on a single AI tool for code review, expecting it to catch all issues, but Weave found that using multiple specialized tools is more effective. At Weave, every pull request is reviewed by four different AI tools—Cursor Bugbot for bug detection, Greptile and Cubic for general code review, and Wispbit for context-aware standards—before reaching a human reviewer. This multi-tool approach covers different aspects of code quality, catching around 60% of bugs early and allowing human reviewers to focus on complex, high-level decisions that AI cannot handle. The strategy emphasizes that each tool should bring unique value, and the cost of running multiple tools should be outweighed by the value of the issues they catch. The process is automated in Weave's CI/CD pipeline to ensure consistency, demonstrating that the best code review setup involves using a combination of imperfect tools to achieve comprehensive coverage.

Weave is an engineering analytics platform designed to help teams measure AI adoption and optimize engineering productivity by connecting development tools like GitHub, Cursor, and Claude to provide benchmarks, track AI return on investment, and assess the impact of AI on engineering velocity. Tailored for CTOs and engineering leaders, Weave offers insights into AI tool usage, productivity improvements, and code quality, enabling data-driven decisions about tool investments and team performance. Unlike traditional platforms, Weave focuses on AI-native metrics, such as AI-generated code quality and tool effectiveness, and integrates directly with AI coding tools for unique insights. It supports enterprise-level requirements, offers rapid implementation, and ensures data security through best practices, allowing organizations to track multiple AI tools simultaneously and make informed decisions about AI tool adoption and investment.

As of 2025, the role of an engineering manager has become increasingly complex due to the heightened emphasis on efficiency, developer productivity, and the challenges of managing hybrid teams. To navigate this evolving landscape, engineering managers are encouraged to adopt a modern toolkit comprising seven essential types of tools. These include Engineering Intelligence Platforms (EIPs) for real-time insights into development processes, project and task management tools for coordinating complex projects, and centralized communication hubs for managing distributed teams. Additionally, documentation and knowledge management systems are crucial for onboarding and scaling team knowledge, while CI/CD and DevOps platforms automate the software delivery pipeline. Code quality and security scanners help uphold code standards and security, and AI-powered developer tools like GitHub Copilot serve as productivity multipliers. By leveraging these tools, engineering managers can move beyond task management to lead high-performing, engaged teams with clear, data-driven insights for strategic decision-making.

Artificial Intelligence is transforming engineering analytics by moving away from traditional, superficial metrics to provide data-driven, predictive, and personalized insights that enhance engineering performance. AI tools are revolutionizing areas such as design and code quality by offering real-time, automated analysis that improves efficiency and reduces errors. They enable predictive analytics by forecasting timelines and identifying bottlenecks, thus allowing proactive problem-solving and risk management. Furthermore, AI delivers personalized feedback to engineers, helping them identify skill gaps and foster professional growth, as exemplified by platforms like Weave, which act as personal feedback engines. This paradigm shift demands that engineers develop data literacy and AI skills as core competencies, with educational resources now available to support this transition. As AI becomes integral to software development, it is empowering engineers to focus on higher-level problem-solving while automating routine tasks.

Standups often fail because they prioritize updates over meaningful discussions that can unblock work, such as conversations about blockers, dependencies, and priorities. Effective standups compress updates to brief essentials, allowing more time for important conversations that require collective input. Teams should quickly share what they did yesterday, what they're doing today, and any blockers, using tools like Weave's Standup Page to streamline and visualize progress. Crucially, follow-up discussions should be captured and addressed immediately after the standup in smaller groups, which allows for detailed problem-solving without holding everyone hostage. Standups should adapt to the rhythm of the workweek, sometimes requiring only a quick check-in and other times needing more in-depth recalibration. The ultimate purpose of a standup is not to monitor team members but to align efforts, surface blockers, initiate important conversations, and maintain momentum throughout the day.

Wooly AI, also known as the Deep Research Agent (DRA), is an advanced AI tool designed to transform fragmented engineering data into actionable insights for engineering leaders and team leads. Unlike traditional analytics that are reactive and limited to pre-configured metrics, Wooly AI is exploratory and synthesizes data from multiple sources, such as project management tools and version control systems, to provide contextual answers to complex questions. It excels in anomaly detection, performance analysis, and multi-dimensional correlation analysis, enabling teams to identify bottlenecks, understand system dynamics, and make data-driven decisions about productivity, quality, and team health. The agent operates on a unified data layer, allowing it to analyze comprehensive performance reports and detect patterns that are otherwise impossible to discern when data is siloed. Wooly AI's architecture is sophisticated, employing parallel research agents for deep analytical queries, and is currently focused on research and reporting, with future plans to automate project management and engineering operations.