Week in review - Sunday, Dec 15th
This week’s lineup spans everything from how Graphite’s own engineering team accelerates code delivery, to the future of AI-driven developer workflows, to rethinking the fundamentals of version control. If you missed any posts, here’s a quick recap and direct links to each.
AI in dev tools, late 2024: where we are and where we might be headed
We’ve evolved from basic code generation tools into more nuanced AI-driven analysis and insights. The next frontier is integrating AI deeper into build, test, and deployment pipelines—places that demand high reliability. Expect AI’s role to expand from code suggestions to holistic optimization of the entire software lifecycle.
Git is bad. Here's how I'd make it better
Git’s complexity intimidates even experienced developers. This post imagines a simpler interface focused on “diffs” instead of branches, with straightforward commands and easy undo steps. The goal: a version control workflow that’s intuitive, less error-prone, and accessible to newcomers.
Pricing Intelligence: Is ChatGPT Pro too expensive for developers?
With some AI coding assistants priced at $200/month, are developers willing to pay? Reflecting on the history of tool pricing—from expensive proprietary IDEs to the free open-source era—we’re now facing a new normal. As AI tools become indispensable, higher price tags may feel increasingly justifiable for the productivity gains they deliver.
Amazon Q and GitLab Duo: Still Missing a Killer Differentiator
Amazon and GitLab each launched AI coding assistants integrated into their ecosystems, but neither product stands out against established players. To truly differentiate, these platform-native tools need to leverage their unique vantage points—like cost optimization or pipeline intelligence—rather than just matching generic AI code suggestions.
Stacking and Agentic PRs Go Hand in Hand
Stacking small, incremental PRs is already boosting developer velocity and clarity. Pairing this approach with AI-generated (“agentic”) PRs could be transformative. As AI tools continuously generate code changes, stacking provides a natural structure—letting both humans and AI systems iterate rapidly without compromising quality.
How Graphite’s Eng Team Ships Code Remarkably Fast
Our engineering team’s secret sauce is a combination of stacked pull requests, parallelized workflows, and thoughtful iteration—no all-nighters required. We share real data showing how we far exceed industry benchmarks for merged PRs per engineer, all while keeping changes reviewable and quality high.