Show Notes
Almost 25 years ago, in 1997, Daniel Stenburg created cURL, a command line tool for transferring data. The name stands for “client URL,” works on any platform, and is used in billions of installations. Despite maintaining cURL for a quarter of a century, Daniel couldn’t be happier where he is, and wouldn’t want to be doing anything else. We recently sat down with him to hear how he first discovered open source, why he wants to lower the barrier of entry for newcomers, and how he sees cURL evolving in the future, now on The ReadME Podcast.
More from The GitHub Podcast
Cassidy and Andrea are joined by Sameen Karim, Senior Product Manager at GitHub, to dig into stacked pull requests, GitHub's newly launched (and long-requested) way to break large changes into a linked chain of small, reviewable PRs. Sameen pulls back the curtain on just how deep this went: every time a PR updates, GitHub runs a live "ghost merge" simulation behind the scenes to check whether it's mergeable, and stacking meant multiplying that complexity across up to 100 linked PRs at once. The team had to rethink core assumptions about how a PR's base branch relates to its actual merge target, rewrite parts of the merge box that hadn't been touched in over a decade, build an entirely new asynchronous merge path, and close a code owners loophole that could otherwise let people sneak changes past review. Sameen also talks about "agent-driven testing," where GitHub ran swarms of agents around the clock trying to break the feature, how teams like Vercel, Nuxt, and the React team are already adopting stacks, and what's next: support for stacks across forks. The episode closes with three open source picks: Sarah Drasner's visual code explainers, the AGENTS.md standard, and the open source gh-stack CLI Sameen built (and rewrote in Go) to manage stacks locally. Links mentioned in the episode: GitHub Universe Sarah Drasner's code drawings AGENTS.md gh-stack CLI
Cassidy and GPS are joined by Camilla Moraes, a product manager on GitHub's Maintainer Love (aka Tiny Wins) team, to talk about how the team finds and ships small, high-impact fixes for open source maintainers. Camilla walks through her path from engineering to product, and shares how the team's focus shifted this year from quick two-to-three-week "paper cuts" to a bigger, all-hands problem: the rise of low-quality, AI-generated contributions flooding repos. Camilla gives some insight into a new tool the Tiny Wins team has designed to address that issue: PR limits, which lets maintainers cap how many open PRs a user without write access can have at once. The group also talks through how the team prioritizes an endless stream of feedback from Slack, socials, and surveys, and how Camilla built her own repo-as-a-database system with Copilot to track it all. They close out with three open source picks: iOpenPod, Bento, and Camilla's own vibe-coded project for rating the health of a GitHub repo. Links mentioned in the episode: Maintainer Community iOpenPod Bento
Cassidy and GPS dive into a round of AI hot takes to separate the extreme, engagement-bait opinions from the actual nuance. They debate whether developers still need to read AI-generated code, why “jobs won't hire you if you don't use AI” is more complicated than it sounds, and GPS's own hot take against the newest wave of paid coding bootcamps. From there they tackle whether skills have killed MCP, whether RAG is dead, and a borrowed hot take from their boss Martin: if you have to fine-tune a model to your code base, your code is bad. They close out with two open source picks: Pollinations AI, a generative AI platform experimenting with earning credits through code contributions, and Avian Visitors, a beautifully documented build log of a bird-listening e-ink display.
In today’s episode, Cassidy, Marlene, and GPS tackle the ever-growing pile of new AI and agent terminology that's popped up seemingly overnight. They break down loop engineering (designing systems that prompt your agents instead of prompting them yourself), the Ralph loop that helped popularize it, and how squads and fleets let multiple agents split up planning, testing, and review work in parallel. They also dig into harness engineering (the scaffolding around a model that turns raw text generation into something useful), the newly coined “hill climbing,” and why forward deployed engineer might just be a solutions architect with better AI tooling. They wrap with a look at the spectrum between closed, open weight, and open source models, plus each host's open source pick of the week: Astro 7, the RAE API project, and Tau.
Season 2 opener - recap of our favorite moments from season 1!