Launching Pantheon: turning long-form videos into a month of clips

Pantheon is an open AI-agent harness that I explain in long-form video. Long videos don't travel on Shorts, Reels, TikTok, LinkedIn or X on their own.

The goal: engagement across LinkedIn and other platforms, and to get me moving on content I'd always wanted to make but where there felt like too much of a barrier to entry. Even with the raw videos, cutting, clipping and cross-posting was a lot for something that wasn't producing an outcome. This has been a great first step; we'll keep improving and iterating, and I look forward to true KPIs as we go.

1. 1
   API push from your CMS

   Clips go into the flayr library through the API

   The short-form cuts came from two places: Opus Clip (6 of the 12 published shorts) and Claude cutting from the transcripts (5); one short and both long-forms are raw takes. My video pipeline uploaded every clip into the Dostal Tech library through the flayr API, filed straight into the right brand.

2. 2
   Brand isolation

   One brand, with my accounts shared into it

   Every Pantheon post sits in the Dostal Tech brand. That brand posts through accounts shared into it once: the @dammitdostal handles, my LinkedIn and the Dostal Technology Facebook Page. One persona, linked once, and nothing went to the Firefly Events accounts.

3. 3
   Posting queues

   Posting queues booked the calendar in one pass

   Shorts go out at 9:00 CT on weekdays to 8 platforms (YouTube, Instagram, TikTok, X, Threads, Bluesky, Facebook and LinkedIn) from two queues: Mon/Wed/Fri and Tue/Thu. Long-form goes out Wednesdays at 10:00 CT to YouTube, LinkedIn and Facebook. An agent created the queues in one pass on Sep 26 and booked the posts through the flayr API.

4. 4
   Referral footer

   A referral footer on the captions, a watermark on the video

   At publish, flayr adds "Posted with flayr.social" and my referral link to the caption footer. The "Posted by flayr.social" watermark and logo were burned into the videos on hive, my own render machine, before upload.

5. 5
   Labels & experiments

   Labels make Opus, Claude and raw cuts comparable

   Every published post carries a cut: label (cut:opus, cut:claude or cut:raw) and a parent: label for the video it came from, plus an experiment label, so the cuts can be compared per platform.

• Fully verify the accounts on Facebook and the like first. That stuff is hidden and a pain to chase down.
• Beyond that, I'd use all of flayr's newest features: every time something could be better, I changed flayr to help. flayr now ties in with Opus Clip and does its own lighter clipping (Opus if you have the budget, flayr's to start lighter). We do voice cloning now, with a training studio to improve it; a video on that is coming. And I'd tag clips and set KPIs from day one, to line up with analytics. Next I'm digging into ads.
• What's next: more Pantheon through October and November, the Free Fall release that puts all the tools out, and per-clip tagging so the next write-up compares Opus cuts, flayr's own lighter clipping and agent cuts on real numbers. After that, ads.

Pantheon, videos and narration

Mathew Dostal, co-founder of flayr and founder of Firefly Events

Short-form cuts

Opus Clip, plus Claude cuts from the transcripts