What do LLMs mean for the future of software engineering? Will vibe-coded AI slop be the norm? Will software engineers simply be less in-demand? Rain and David join Bryan and Adam to discuss how rigorous use of LLMs can make for much more robust systems.
Previously, on Oxide and Friends:
- OxF s03e08 – Does a GPT future need software engineers
- OxF s04e04 – Helios
- OxF s05e28 – Systems Software in the Large
- OxF s04e20 – Pragmatic LLM Usage with Nicholas Carlini
Some of the topics we hit on, in the order that we hit them:
- The issue Bryan was fixing
- iddqd: the crate Rain built
- Ghostty
- Rain's nextest bug: SIGTTOU when test spawns interactive shell
- Oxide RFD 619: Managing types across Dropshot API versions
- drift: the crate Adam built
If we got something wrong or missed something, please file a PR! Our next show will likely be on Monday at 5p Pacific Time on our Discord server; stay tuned to our Mastodon feeds for details, or subscribe to this calendar. We'd love to have you join us, as we always love to hear from new speakers!