Another quick update from HeyNews (the AI tool we're building to help newsletter operators draft issues in their voice).

This week was mostly about shortening the path between connecting your newsletter and actually having a usable AI Writer, plus a couple of resilience fixes in Content Desk and Sources. Sharing here because the onboarding work in particular is the kind of thing most products underbuild for too long.

What changed:

Connect once, start writing sooner. When you connect beehiiv, Kit, or a newsletter archive during setup, onboarding now finishes from that connection, and Voice Training kicks off right away. If training succeeds, your AI Writer is ready to use without waiting for manual approval, and you'll receive a notification when it's ready. The use case: trimming the "I signed up, now what?" gap that kept showing up in onboarding sessions, without sacrificing the safety check when training fails or produces weak output.
A wider 12-month story window. Usable accounts now share the same 12-month story access window. Older-but-recent stories no longer disappear behind plan-specific upgrade prompts as long as they're inside the 12-month storage horizon. Less special-casing, fewer "wait, why can't I see this story from March?" reactions.
Content Desk keeps your edits. Adding a URL story used to be brittle: if a page had thin metadata or extraction came back partial, the title, details, tags, and content you'd already entered could get overwritten. Now they stick. You can also edit story titles and details after saving, and those edits stay scoped to the current newsletter, so shared story records aren't rewritten for your other newsletters.
Sources are easier to maintain. Fetch now is back for RSS feeds, social profiles, and shared sources. Source names, descriptions, notes, and tags are saved as source-level metadata rather than overwriting the article or video titles attached to stories. A cleaner source library no longer means messier story headlines.

Plus the usual round of fixes. ActiveCampaign public archives now import the real issue URLs instead of treating archive index pages as issues. Fresh Content Desk stories stay visible in Compose even before story scoring finishes. Removing a source no longer leaves stale source groups in the current newsletter, and it doesn't hide the same story from your other newsletters. Voice Training handles larger past-issue libraries more reliably during source discovery. Cancel and save-offer choices are guarded against duplicate clicks and race conditions.

If you're building anything with a "connect external service, kick a long-running training job, surface readiness without a manual approval bottleneck" loop, happy to compare notes in the thread. That was the part of the week that took the most rethinking.

Cagri Sarigoz

Activity

Products

Recently upvoted

Posts