Software should work, and talking about it needn’t be boring
This opinion piece from The Register argues that streamed tech conferences are usually too uneven to watch in full, but praises Software Should Work 2026, held in July, as a rare exception that rewards binge-watching from start to finish. The conference focuses on software reliability in its broadest sense, covering not just working code but abstraction, formal methods, complexity management, tooling, team culture and individual craft, making it feel like a genuine symposium rather than a marketing showcase.
The event comprised 13 talks, watchable in a day, featuring speakers with deep hands-on experience who chose their own subjects rather than following a fixed theme. Highlights include Richard Feldman's comparison of web stacks and dependency growth across 1996, 2006 and 2026 (spanning languages such as TypeScript, Python, Rust and Go), and SQLite creator Richard Hipp's personal account of how his codebase evolved alongside changing test methods. Other talks touch on formal analysis, the "superhero engineer" myth, AI-related pitfalls, and debates over tools like Nix, with the author framing reliability as a culture-wide challenge rather than a purely technical one.
- The Register praises Software Should Work 2026 as a binge-worthy tech conference
- Its 13 talks explore software reliability beyond just working code
- Speakers include Richard Feldman and SQLite creator Richard Hipp
New here? Start with this
Software Should Work 2026 is a technology conference held in July, aimed at software developers and engineers rather than the general public. Rather than promoting particular products, it brought together experienced practitioners to discuss why software so often fails to work reliably, and what can be done about it, covering topics from coding practices to team culture. Recordings of the talks were later made available to watch online.
The article being discussed is an opinion piece from The Register, a technology news publication, reviewing the conference. It highlights individual talks, including one from Richard Feldman comparing how web development tools have changed over three decades, and another from Richard Hipp, the creator of the widely used SQLite database software, reflecting on his own experience testing code over many years.
The wider context is an ongoing conversation within the software industry about reliability: as software underpins more of daily life, from banking to transport, developers and commentators increasingly debate how to prevent bugs, manage complexity, and build systems that work as intended, and whether this is mainly a technical problem or one rooted in how teams and individuals work.