Release notes from what you shipped
You already did the hard part. Your AI reads your pull requests, asks you only why, and drafts the post from your answer. You read it and approve.
Why they don’t get written
Shipping is the fun part. Writing it up afterwards is a blank page at the end of a long week, so it waits. Then the next thing ships, and the changelog falls further behind.
But most of a release note already exists. Your pull requests and issues say what changed. The only part that lives in your head is why, and you can say that out loud in a minute.
How it works
- Ship it. Merge as usual. Nothing changes in how you work.
- Ask your AI. Claude, ChatGPT or any assistant connected to Tamable reads the pull requests and issues through your own connectors, and proposes one post per story, not one per merge.
- Answer one question: why. It asks only for what isn’t written down. Answer out loud, long and messy; a dictation app turns it into text, and Tamable saves it as an idea in your own words.
- Read the draft. It’s written in your product’s voice, from your project’s writing strategy, and adds nothing you didn’t say. Leave a note on any line you want changed.
- Approve it. It goes out on your own blog at the date you pick. The same story can become an X or LinkedIn post too.
A real one
When LinkedIn came to Tamable, this is what I said about it, exactly as my dictation app typed it, slips and all. Below it is part of the post that went out.
This sounds perfect. We should have done this earlier, actually, because I can see this opening a whole new market. To be honest there are a bunch of people that need to be active in leasing: executors, business owners, and people that are trying to get fired.
I think that being someone relevant on X/Twitter is more of a hipster thing, in a way. For the cool kids on the block, if you will, Twitter is the place, but honestly, proper business, I think, gets done on LinkedIn. Building a strong LinkedIn profile or being active on LinkedIn can probably have more return on investment than X or a blog.
Let’s do this. I think this is going to be huge
We should have done this earlier, actually, because we can see it opening a whole new market.
X has been Tamable’s home since day one, and it’s a great place for builders. But a lot of proper business gets done on LinkedIn.
There are a lot of people who need to be active on LinkedIn: executives, business owners, and people trying to get hired. For many of them, building a strong LinkedIn profile and posting there regularly can probably bring more return than X or a blog.
Read the whole post“Leasing” was LinkedIn, “executors” were executives, and nobody is trying to get fired. The dictation slips are fixed and the repetition is gone, but every opinion is still mine. The rest of the post, what changed and where to find it, came from the work itself.
What it takes from you
One answer and one read per post. The headings, the order, the transitions and the opening are your AI’s job; the reasons and the final say are yours.
Every post in Shipped, and why on this blog is made this way, and so is Mustloop’s: 23 release posts across two products so far, each one read and approved by me before it went out.
The prompt
Paste this into your AI when something ships, with your project’s name in place of [project].
I just shipped something in [project]. Start with get_workspace, then read what changed since my last post in Tamable, through my own connectors, like pull requests on GitHub or issues in Linear. Group the changes into stories: one post per story, not per pull request. For each, say what changed, why (the problem and the reasoning, as recorded) and what it means for readers. Propose the list first, and ask me only for the why that isn’t recorded. Never guess it. Save my answers as ideas in my own words. Draft only the stories I pick, with that project’s strategy, and add them to its release notes series if it has one. Security fixes in general terms only. Submit the drafts for review when they’re ready.
Starting from a backlog of things you never wrote up? The story so far prompt in the Playbook reads your whole history and turns it into a series, each post dated when its change shipped.