Rubber Duck Thursdays - Let's cowork! Cassidy's back! Hey!

Key Concepts

AI in Software Development: The increasing integration of AI tools, particularly AI-driven coding assistants like GitHub Copilot, into the software development lifecycle.
Developer Productivity: The impact of AI tools on accelerating development speed, iteration, and the overall efficiency of developers.
Agentic AI: AI systems designed to act autonomously to achieve specific goals, such as writing documentation or executing tasks.
GitHub Universe: A recent conference where GitHub announced new features and discussed the future of developer tools.
Octoverse Report: GitHub's annual report detailing trends and statistics in the developer community.
MCP (Modular Code Patterns/Modular Component Patterns): An emerging standard or framework for structuring code to improve interoperability with AI tools.
Agility and Adaptability: The necessity for developers to be agile and adaptable in the face of rapid technological changes driven by AI.
Open Standards: The importance of open standards like MCP in enabling AI tools to work effectively across different frameworks and technologies.
Human Craftsmanship vs. AI Assistance: The evolving role of human developers, emphasizing critical thinking, problem-solving, and strategic oversight rather than rote coding.
Transparency and Attribution: The ongoing discussion about how to acknowledge and attribute the use of AI in software development.
Late Adopters: The role and benefits of developers who adopt new technologies after the initial hype has subsided.
Copyright and Intellectual Property: The complex legal and ethical considerations surrounding AI-generated content and code.
Rubber Ducking: A debugging technique where developers explain their code or problems to an inanimate object to gain clarity.

Introduction and Return to Streaming

The speaker, Cass, returns to "Rubber Duck Thursdays" after maternity leave. She acknowledges the presence of her baby crying in the background as a reality of working from home. She expresses excitement about being back and notes the significant progress made by her team during her absence.

GitHub Universe and Recent Announcements

Cass highlights GitHub Universe as a recent event that featured discussions on "maintainer wins" and "paper cut works" aimed at improving the developer experience on GitHub. She also mentions "agent HQ mission control" for running Copilot agent things. She encourages viewers to check out the GitHub YouTube channel for more details.

Community Engagement and Developer Projects

Cass actively engages with the chat, asking about viewers' projects and pain points. Several viewers share their current work:

One user is offboarding a project and exploring how to leverage Copilot to backfill documentation. Cass finds this a "super handy" and practical use case, especially for reducing mental and emotional effort when handing over projects.
Another user is building MCP tooling.
A user is developing a decentralized application that uses a maps API to claim plots of land for a daily yield percentage.
Someone is building a portfolio in Python and PyTorch for image processing.
A significant challenge mentioned is applying agentic AI in enterprise, requiring new design patterns for agents to understand the domain. Cass agrees, emphasizing the role of documentation and the need for codebases to be architected for easy integration of different AI tools without extensive fine-tuning.

The Pace of AI Development and Developer Agility

A key theme is the overwhelming pace of AI development and the "overload" of new tools. Cass frames this as a "test of agility" for developers. She suggests that focusing on adaptability and being open to new tools is crucial, as the "best" tool today might be different tomorrow. She notes that her own blog posts can become outdated within weeks due to this rapid evolution.

Cass predicts a future where standards will be refined rather than entirely new ones being created, citing MCP as an example of a standard that allows for iteration and expansion. She emphasizes that while specific models might excel at different tasks, the focus should be on a flexible tool stack that allows for high productivity with a low learning curve.

The Octoverse Report and Developer Trends

Cass shares insights from the Octoverse report (octoverse.github.com), highlighting staggering statistics:

Nearly a billion commits in 2025.
230 new repositories created every minute. She acknowledges that a significant portion of this activity is AI-generated but stresses that developers are still actively involved in shaping and working with this code. The report prompts a discussion on how developers can find their place and iterate on their roles without generating "slop."

AI as a Tool: Productivity Gains and Late Adopters

Cass addresses the polarized views on AI in coding. She argues that while some applications of AI might be "silly" or unnecessary, its productivity gains are undeniable. She compares AI-driven coding tools to the advent of autocomplete, which was initially met with skepticism but is now a standard part of the developer toolkit.

She believes there is a space for late adopters, who can benefit from the refinement of tools that have emerged from the initial hype. However, she cautions against being overly resistant, stating that AI is becoming a reality in the developer's toolkit.

Collaboration, Innovation, and Openness

Cass shares an "underrated lesson" about the power of openness. She believes that giving away knowledge, code (open source), or kindness improves product quality, productivity, and user experience. She also references the mentor's advice to "lift as you climb," emphasizing the importance of elevating others while advancing one's own career.

Copilot Interfaces and Tools

When asked about preferred Copilot interfaces, Cass admits to not having extensively used the Copilot CLI due to her leave. She prefers the VS Code interface for its ability to easily accept or reject suggestions and provide feedback, citing a personal preference for control.

She also mentions Puppeteer as an underrated and powerful tool for web scraping and screenshotting, which can be integrated with AI for further analysis and modification.

Applying Agentic AI in Enterprise and Standards

The challenge of applying agentic AI in enterprise is discussed. Cass believes that documentation will be key, with AI assisting in initial documentation but requiring human guidance. She also highlights the need for codebases to be architected so that any model or tool can be integrated without extensive fine-tuning. This leads to questions about custom instructions, prompts, and new linting methods for LLMs.

The discussion touches upon how tools can support cutting-edge technologies like .NET 10. Cass points to Convex.dev as an example of a framework that has integrated AI code generation tools and documentation, allowing for up-to-date support. She advocates for open standards like MCP to facilitate this integration across different tools and frameworks.

Overcoming Fear and Making Learning Enjoyable

To combat the "panic" surrounding AI, Cass advises developers to "build what you like." While acknowledging job security concerns, she suggests that pursuing personal projects can be a more enjoyable and effective way to learn. She notes that the job market is active, but finding the right roles can be challenging.

Validation and Testing in the AI Era

The Octoverse report indicates a significant increase in testing and automation. Cass infers that as AI generates more code, there will be a corresponding need for increased automation in testing and validation. She believes that while humans should review, testing should facilitate rapid iterations rather than becoming a bottleneck.

Personal Anecdote: Copilot Agent on Vacation

Cass shares a personal story about using Copilot agent mode while on vacation with limited access to her laptop. She was able to update a website by providing documentation to the agent, which successfully made the necessary changes, allowing her to enjoy her vacation. This illustrates AI's ability to accelerate tasks and get developers "part of the way there."

The Future of Development Roles

The question of whether a developer who started before the AI era would "win" against one who only uses AI tools is met with the response that "everyone can win." Winning is defined differently for each individual.

Open Standards and Innovation (MCP and RSS)

Cass draws a parallel between MCP and RSS (Really Simple Syndication). She explains how RSS, as an open standard, has empowered podcasting and led to significant improvements in related tooling. She anticipates similar growth and innovation for MCP as it becomes more widely adopted.

AI and Copyright Concerns

Regarding Disney's lawsuits against LLMs, Cass admits to not knowing the technical implications of rolling back AI knowledge. She acknowledges the importance of copyright but also notes that people can exploit the system. She expresses caution about discussing legal matters she is not fully informed about.

Flutter Development and AI Integration

Cass addresses the question of whether full-stack Flutter development is still worthwhile. She states that "all development is AI development right now," meaning that AI is becoming integrated into most development practices. She advises Flutter developers to continue building what they are comfortable with and to be open to using AI tools within their existing workflows.

AI as a Crutch vs. a Tool

When asked about controlling AI usage during learning, Cass suggests turning it off occasionally to allow for independent problem-solving. She emphasizes that AI can be a powerful learning tool but should not become a crutch that prevents developers from understanding why things break.

AI and Job Replacement

Cass reiterates that AI is unlikely to fully replace developers but will change how they work. She compares the situation to travel agents, where the internet provided new tools that changed the role but didn't eliminate it entirely. Expertise and the ability to leverage new tools become paramount.

Transparency and Attribution of AI Usage

The discussion on whether to indicate AI usage is ongoing. Cass personally believes in transparency and is upfront about using AI tools. She shares her website (cassadoo.coi) as an example where she clearly states her AI usage policy. However, she acknowledges that not everyone agrees, and established norms are still developing.

The Role of Fundamentals and Frameworks

In response to a question about generalists vs. specialists, Cass suggests the "T-shaped developer" model, advocating for broad knowledge with deep expertise in specific areas. She believes understanding language fundamentals is crucial for grasping framework functionalities, but also acknowledges the productivity of pure framework developers.

PocketCal Project

Cass highlights her project PocketCal (pocketcal.com) as an example of a project she started with AI and then developed further. It's an open-source, basic calendar app where the entire calendar state is saved in the URL.

Rubber Ducking and Future Streams

Cass explains the concept of rubber ducking in software development. She mentions her own "rubber duck" is a ceramic chicken. She announces that "Rubber Duck Thursdays" will continue weekly, with different team members hosting. She also promotes "Open Source Fridays" for interviews with open-source maintainers and mentions upcoming streams in Spanish and Portuguese.

Final Thoughts on AI and Development

Cass concludes by emphasizing that AI is a tool that assists productivity, not a replacement. She believes that while AI will change roles, human expertise and critical thinking will remain essential. She encourages developers to be adaptable and to leverage AI to enhance their work.