Three weeks ago someone opened an issue on the formatting engine repo, fixed a bug I'd been ignoring for a month, and closed their laptop without saying another word.
I never got their name on Dev.to. I only know them as a GitHub handle. That one PR did more for PostAll than anything I wrote in the last four release notes combined.
That's the moment I realized I've been treating "build in public" as a one-way broadcast. Post the architecture, post the numbers, post the failures — and just... hope people read it. I haven't actually asked for anything back. So this post is the ask.
If you've read any of the PostAll series — the formatting engine deep-dive, the quality gate breakdown, the scaling post-mortem — you already know where the project stands. 100 beta users, the formatting engine open-sourced, quality gates catching 50% of low-quality output before it ships. This post isn't a recap of that. It's four specific things I need help with, and what you'd actually get in return.
1. Feedback on the feedback loop (yes, really)
The next big piece of PostAll's roadmap is a performance feedback loop — using real engagement data to adjust quality gate scoring weights dynamically instead of the static thresholds I hand-tuned from launch. I flagged this as the highest-complexity, highest-risk item on the roadmap for a reason: I don't have a clean answer for how to avoid the system rewarding whatever the model happens to be good at that week, instead of what actually reads well.
I don't want to build this in a vacuum and find out at 500 users that I baked in the wrong feedback signal. If you've built anything with a scoring system that adapts based on downstream performance — recommendation ranking, spam filters, ad quality scores, anything — I want to hear how you avoided (or didn't avoid) feedback loops that reward gaming the metric over the actual goal.
Specifically useful:
- Have you shipped a dynamic weighting system that went sideways? What broke?
- What's your standard defense against a scoring model overfitting to its own reward signal?
Drop it in the comments or DM me. This is the one where I most want to be told I'm about to make a mistake.
2. Beta testers for brand voice profiles
Per-account brand voice profiles are next in line after the feedback loop — letting each account lock in a tone/style fingerprint so generated content doesn't read like generic AI output every time. Right now it's 64% built.
I need testers who:
- Already have PostAll's beta access, or are willing to sign up specifically for this
- Have an actual brand voice they care about matching (not "make it professional" — something with real texture, an agency's client voice, a personal newsletter tone, whatever)
- Will tell me honestly when the output sounds like their voice and when it very much doesn't
This isn't a polished feature yet. You'll hit rough edges. In exchange, you get early access, direct input on how the profiles get tuned, and credit in the article I write about this once it ships (or doesn't — I'll write that one too if it fails).
3. Contributors on the open-source pieces
The formatting engine — the "parse once, render many" model that generates one universal content object and renders it to blog, social, and email formats — is open source now. It's the piece of PostAll I'm most confident is useful outside of PostAll entirely.
Where I actually need hands:
- CMS integrations — WordPress, Ghost, and Webflow are done. Contentful and Shopify are next on the list and I haven't started either. If you've worked with either API and want to own an integration end to end, this is a real, scoped, mergeable contribution — not a "good first issue" throwaway.
- Renderer edge cases — the universal model handles the common formatting cases well. It does not handle every edge case in every target format, and I guarantee you will find one within an hour of using it for something I didn't anticipate.
I'll be honest about the state of the contributor experience: the docs are thinner than they should be and the test coverage on the newer renderers is not where I want it. If you show up and the first thing you hit is confusion, that's useful information for me too — tell me where you got stuck before you tell me what you fixed.
4. Partnerships — with actual scope, not vague ones
I get "let's partner!" messages that turn out to mean "link to my product in your next article." That's not what this is.
What I'm actually looking for:
- CMS or publishing platforms interested in an official integration (beyond what I can build solo evenings and weekends)
- Agencies running high-volume content operations who'd pilot PostAll against their real workflow and let me publish an honest case study — including the parts where it doesn't save time
- Anyone building complementary tooling in the AI content space where the integration is obvious and mutual, not "we both make AI stuff so let's cross-promote"
If you're at a company where any of that maps onto something you're already trying to solve, reply here or reach out directly. I'd rather have one real conversation than ten "let's hop on a call" messages that go nowhere.
What this isn't
To be straight about it: this isn't a hiring post in disguise, and PostAll isn't in a position to pay contributors right now. What I can offer is real technical collaboration, credit, early access to features before they're public, and an audience — this series has reached more than 100 readers and growing, and I'll happily point that at your work if you point yours at mine.
I also want to be upfront that "autonomous publishing agents" — letting PostAll publish content without a human in the loop — is explicitly off the roadmap for now. If that's the integration you're picturing, save us both the conversation.
What's the one thing you'd want built into a content automation platform that doesn't exist yet? Not a feature request for PostAll specifically — I'm curious what the actual gap in this space looks like from where you're sitting.
Top comments (0)