Reading up on Graphite
11 deep · digging since nov 28, 25
- Git Is Not Fine
Git's immutable commit model makes stacked PRs, async workflows, and partial-state editing genuinely painful despite its success as a distributed source store.
- Twenty Years of Stacking Commits — jd:/dev/blog
Stacked commits, not monolithic PRs, are the correct unit of code review—a truth Gerrit knew in 2008 that AI's speed and tooling like Mergify Stack are now forcing GitHub to finally admit with its native stacked PRs feature.
- When LLMs Get Personal - by Joshua Budman
LLM answers to the same query vary across users but share a stable semantic core, with personalization concentrated in examples, framing, and local detail rather than arbitrary divergence.
- GitHub Stacked PRs
GitHub now natively supports stacked pull requests within monorepos, enabling developers to break large changes into smaller, reviewable PRs.
- In praise of the stacked pull request
Stacked pull requests, managed with tools like Graphite, keep code review fast and team velocity high by breaking large work into small, independent, mergeable units.
- AddyOsmani.com - AI writes code faster. Your job is still to prove it works.
AI-generated code shifts the bottleneck from writing to verification; human code review remains essential for security, context, and accountability.
- This week on How I AI: How Webflow's CPO built an AI chief of staff
Webflow's CPO built an AI chief of staff to run her week, using markdown files and disposable widgets for calendar delegation and meeting prep.
- Graphite Gets Bought By Cursor: Three Reflections from Six Years of Work [PMF didn’t happen right away; working IRL mattered; three founders, three distinct roles]
Graphite's acquisition by Cursor reflects lessons on finding product-market fit through iteration, the value of in-person work, and three co-founders staying united over six years.
- More Articles Are Now Created by AI Than Humans
In November 2024, the quantity of AI-generated articles published online surpassed human-written articles, though the proportion has plateaued since May 2024.
Takes
I have a draft blog post swirling around this exact topic (but not refined enough to publish yet). I think the key thing is I (personally) don't want a NEW GitHub. I want GitHub to be better. For example: - GitHub issues should be as beautiful and good as Linear - GitHub PRs should be as good as Graphite - GitHub Git infra should be as fast/minimal as Pierre - GitHub wikis should be more like Notion - GitHub discussions & shouldn't exist (multiple "better issue" providers including Linear show why) - etc. I'm not saying to clone those full companies outright, but their core product, arguably the core features, aren't even 2% as good as those external products. Maybe aim for 10% to start. There's the "oh no there's so much tech debt" argument. And I'm sure GH is on an absolutely mountain of tech debt. That's why in my prior twoots I've argued to just make them separate products to start only for agility reasons, unapologetically do not integrate with "old github." Net net startups beat encumbants all the time for reasons. That's just a product/technical POV though. GitHub also has a huge PR/marketing problem. They talk through corp speak, their marketing pages (e.g. the dot com) speaks to multiple personas confusingly, they have no singular visionary to look up or trust, they have nobody who makes the outward community feel seen. There's so much more here... I think for the human side, GitHub already has what it needs to be really, really, really good. It really feels like they just like fearless vision, and the courage/power to say "fuck you" to a whole lot of things that are distracting them.
@mitchellh
When the cost of code goes to zero, marketing is your only advantage.Introducing Flint. It builds you a unique page for every ad, keyword, and customer.We’re already doubling conversions for @Cognition and @Graphite. Sign up at @tryflint. pic.twitter.com/QU3cW8zW5o
@michlimlim