Building the machine that builds the machine (Interview) - The Changelog: Software Development, Open Source Recap
Podcast: The Changelog: Software Development, Open Source
Published: 2026-02-11
Duration: 1 hr 37 min
Summary
In this episode, Paul Dix discusses his experiences with agentic coding and the transformative impact of AI coding agents on software development. While he has embraced these technologies, he also reflects on the necessity of hand-coding in certain situations.
What Happened
Jared welcomes Paul Dix, the co-founder of InfluxDB, to discuss his journey in adapting to an agentic world where AI coding agents assist in software development. Paul shares his mixed experiences with these agents, acknowledging their potential while also noting the challenges of relying on them without supervision. He describes a recent phase where he had to dial back his agentic coding efforts and revert to hand-coding due to some setbacks, illustrating the balance developers must find between leveraging AI and maintaining control over their code.
Throughout the conversation, Paul emphasizes how the landscape of coding has shifted dramatically in the past year, particularly with the advent of improved AI tools like Opus 4.5 and Codex 5.2. He reflects on the initial excitement he felt when first using these tools, describing his early skepticism that quickly turned into enthusiasm as he witnessed their capabilities. Paul believes that the days of hand-writing code could be numbered, urging developers to embrace AI tools to enhance productivity and efficiency in software development.
Key Insights
- AI coding agents can significantly speed up the development process.
- Hand-coding remains relevant, especially when AI tools are not adequately supervised.
- The introduction of advanced AI tools like Opus 4.5 has transformed coding practices.
- Developers should focus on managing AI tools rather than solely writing code.
Key Questions Answered
What are the challenges of agentic coding?
Paul discusses the highs and lows of agentic coding, noting that while it can offer significant benefits, there are also setbacks when agents are left unsupervised. He mentions a recent experience where he had to dial back his use of AI agents due to complications that arose from their unchecked performance.
How has AI changed the coding landscape?
According to Paul, the release of tools like Opus 4.5 and Codex 5.2 has drastically changed the coding landscape. He emphasizes that coding has become easier and cheaper, leading to a shift in how developers approach their work.
What is the role of hand-coding in modern software development?
Despite the rise of AI coding agents, Paul admits that he still finds himself hand-coding at times. He expresses that while it is enjoyable, he often feels slow compared to the pace that AI tools can offer, suggesting that hand-coding still has its place but may be less efficient.
What advice does Paul give to developers regarding AI tools?
Paul encourages developers to fully embrace AI coding tools, suggesting that they should experiment freely without the concern of hitting token limits. He believes that this freedom allows for better exploration and innovation in coding practices.
How did Paul first experience agentic coding?
Paul recounts his initial experiences with AI coding tools after the release of Opus 4, where he began using them for well-defined problems. He quickly recognized their potential and began advocating for their use among his engineering team, highlighting a shift in his perspective on coding.