The Branchpost blog
Posts on developer publishing, AI-assisted content, the GitHub-as-CMS workflow, and what we're learning while building Branchpost.
Why we make every blog draft a pull request
Review tools already exist for code. They're called pull requests. Here's what happens when you point that same machinery at content — and why it changes the tone of AI-assisted writing.
Why blog posts belong in your monorepo, not a SaaS dashboard
Content is code-adjacent. It ships with releases, needs review, and benefits from version history. Here's why the dashboard is the wrong surface for AI writing tools, and what changes when drafts arrive as pull requests instead.
GitHub App permissions: designing for least privilege when AI writes to your repo
The exact scopes Branchpost requests from GitHub, why each one is justified, and the scopes we deliberately refused even when they would have made the product easier to build.
Frontmatter as a contract between AI and your CMS
Why the YAML block at the top of every MDX file is the most important interface in an AI content pipeline, and what breaks when you treat it as decoration.
How to review an AI-generated draft before you merge it
A code-review mental model, ported to prose. What to read for, what to ignore, and where AI drafts most often fail in ways line-by-line review misses.
The cost shape of an AI content pipeline
Where tokens actually go in a Branchpost generation run, why retries dominate the bill more than people expect, and how prompt caching changes the math.
Writing prompts that produce code-aware blog drafts
How Branchpost turns repo signals into prompt inputs so drafts reference real files, real commits, and real decisions instead of generic AI filler.
Try another keyword or clear the search to return to the full archive.
One post per week. Zero AI slop.
New posts on developer publishing, AI-assisted content, and the GitHub-as-CMS workflow. Unsubscribe with a single click.