About:

Scott Werner explores AI and software, sharing insights and experiments on his blog, Works on My Machine.

Website:

Specializations:

Interests:

AI Software LLM-generated code Promptable Architecture
Subscribe to RSS:
The author reflects on the rapid evolution of technology and the concept of expertise in the context of AI development. They describe their experience building a software called Protocollie in just four days, emphasizing the chaot...
The post reflects on the nostalgia of software tools and the evolution of software development, suggesting that the future will see an abundance of simple, single-purpose tools created by individuals using AI. It discusses the shi...
The post discusses the unexpected versatility of USB-C and the Model Context Protocol (MCP), comparing them to universal systems that allow for diverse functionalities beyond their original purposes. It highlights how MCP can conn...
The author reflects on the challenges of managing code complexity in the age of AI-assisted development. Using a barber's theory about hair accumulation as a metaphor, they discuss how AI can rapidly generate code, leading to an o...
The author reflects on the concept of Cucumber testing and introduces Botrytis BDD, a tool that allows for more flexible and forgiving testing. The tool uses LLMs to understand the meaning behind test steps and provides a more sem...
The text discusses Synonllm, a Ruby module that combines method_missing with an LLM to provide AI-powered synonym matching for method calls. It demonstrates how the module works and explores potential future applications, such as ...
The text discusses ways to make reviewing AI output fun by reorganizing work to lean into randomness, and design workflows to minimize the impact of generation errors. It includes two experiments: Daily Random Art Frame and Histor...
The text discusses the impact of AI on knowledge work, highlighting the imbalance between AI's ability to produce work and humans' ability to make subjective decisions. It raises concerns about job satisfaction, the inadequacy of ...
The text explores the concept of 'Waste Inferences!' by triggering multiple AI attempts automatically from a project management tool and picking the best result. It showcases an automated workflow triggered by assigning an Asana t...
MonkeysPaw is a web framework that generates entire pages from prompt files, using natural language descriptions. It represents a fundamental shift in how we can think about building things in an AI-first world. The framework is o...
The text discusses the use of polymorphism and loose coupling to leverage AI, specifically LLMs, to generate new variations of components for a codebase. It explains how structuring code to focus on replicating well-defined, polym...
The text discusses the concept of vibe coding and the process of using AI app building tools to create a prototype for a Community CRM app. It also explores the idea of representation conversion and the power of AI in generating u...
The text discusses using code generation techniques to implement const_missing with the ability to generate implementations of missing constants. It explores the idea of using descriptive class and variable names to guide an LLM t...
The text discusses an agent programmed to follow the Test && Commit || Revert (TCR) process, which involves making a commit when tests run correctly and reverting the code when they fail. The agent is designed to help with small, ...
The text discusses the use of AI programs to generate code for the Sublayer gem using GitHub Actions. It explains the process of anticipatory code generation and the challenges of keeping up with AI-generated tasks. The post also ...
The text discusses the use of the Fanout/Parallelize/Reduce pattern for lead enhancement using Asana, Exa.ai, and Notion. It explains how two different agents work together to enrich leads and sync them to a CRM. The author also r...
The text discusses AI-assisted refactoring with Asana and GitHub, demonstrating how to decompose refactoring into small tasks in Asana and turn them into PRs in a GitHub repository. It emphasizes the importance of parallelization ...
The text discusses the use of AI to automate the software development process, specifically in generating Cucumber tests and implementation plans using Gemini 1.5 Pro and GitHub Actions. The author demonstrates a demo of automatin...
The text discusses the exponential growth of computing power and the reduction in cost, and how it should be embraced to create more powerful, flexible, and innovative designs. It emphasizes the concept of 'wasting inferences' by ...
The text discusses the tendency of generative AI to hallucinate and the potential of embracing these quirks in software development. It explores the history of software development and the pendulum between formal and informal appr...