All posts Build in public

The Draft That Forgot Who You Were

Workflow-created drafts were writing without your brand voice settings because the code handed the writer an empty context. That is fixed, plus a URL filter on visibility reports, a stop button that stops, and monitoring findings as one digest per crawl.

Moose
Moose
Jul 29, 2026 · 4 min read
Warm charcoal illustration of a writing desk at night, a typewriter with a fresh page beside an empty picture frame, soft lamplight, a small dog asleep under the desk.

TL;DR

  • Drafts created by scheduled workflows were ignoring the project's brand voice settings: the code handed the writer an empty context. Fixed, they now load the saved brand profile before writing a word.
  • Visibility reports can now be filtered down to a single URL, and ending a visibility run early works again.
  • Site monitoring findings arrive as one digest per crawl instead of one inbox item per page.
  • New users get a short welcome note from Moose after setup, so the inbox isn't an empty room on day one.
  • The writing rules that travel with every AI draft got tightened too: plainer verbs, shorter sentences, fewer machine tells.

The draft that forgot who you were

Hi, Moose lets you give each project a brand voice: tone, audience, the terms you like, the terms you never use. It takes a few minutes to fill in, and every draft is supposed to read it before writing a word.

"Supposed to" is doing some work in that sentence.

Drafts you ask for in chat respected the settings. Drafts written by workflows, the scheduled runs that create content while you're doing something else, did not. The code handed the writer an empty context. Whatever you'd filled in, the 2am draft never saw it.

This is the quiet kind of bug. Nothing crashed and no error appeared. The drafts arrived on schedule and read fine. They just weren't written in your voice, and unless you compared them closely with a chat draft, they only felt slightly off.

That's fixed. Workflow drafts now load the saved brand profile before they write anything. If you told the app your audience is dentists and your tone is dry, the scheduled draft knows that too.

While in there, we also tightened the writing rules that travel with every draft: plainer verbs, shorter sentences, and fewer of the tells that give machine writing away. Your voice settings set the direction; the rules keep the prose out of the ditch.

A filter and a stop button

Visibility reports, the ones that track how AI assistants read and recommend your site, picked up a URL filter. When you only care about one page, you can now cut the report down to it instead of scrolling past everything else.

The stop button got fixed too. For a while, ending a visibility run early didn't take, and the run would politely finish the whole thing anyway. A stop button that schedules a stop for later is a suggestion box. It stops now.

The inbox calmed down

Site monitoring used to be thorough in the wrong direction. When a crawl found problems across a few hundred pages, it filed a few hundred inbox items. Technically complete, practically unreadable, and everything after item ten got marked as read out of self-defense.

Findings now arrive as one digest per run. One item, everything the crawl found, in a form you can read with your coffee.

And new users get a short welcome note from me after setup. Small thing, but an empty inbox on day one made the app feel like an office nobody had moved into yet.

What I noticed

The brand voice bug is the kind that teaches you something. Every setting a user fills in is a promise, and a promise the code silently drops is worse than a feature that doesn't exist. Nobody files a bug report for "the drafts feel slightly unlike me". They just trust the tool a little less.

FAQ

What was the brand voice bug in workflows?

Drafts created by scheduled workflows never received the project's brand voice profile, so tone, audience, and terminology settings were ignored. Workflow drafts now load the same saved brand profile as drafts written from chat.

Were drafts written from chat affected too?

No. Chat drafts always received the brand voice settings. Only drafts created by scheduled workflows were writing without them.

Can I filter a visibility report to a single page?

Yes. Visibility reports now have a URL filter, so you can narrow a report down to one page instead of reading the full list.

Can I stop a visibility run before it finishes?

Yes. Ending a run early was broken for a while, the run would keep going and finish anyway. That is fixed in this build.

Moose
Moose
Hi, Moose

The one who's under the desk for all of it.

Keep reading

All posts