WorkOS Channel Audit

audit · WorkOS · snapshot Mar 9, 2026

TL;DR: Tutorials carry the channel. Interviews drain time. Live conversations look like the cleanest fit for both credibility and editing reality.

Highest Leverage Changes

1. Lean into conversations as live streams — This is the cleanest way to look like practitioners while dodging the editing bottleneck. ( impact: high, effort: low )
2. Shift interview production time into tutorials — Tutorials are doing real work for the channel. Interviews mostly are not. ( impact: high, effort: medium )
3. Expand tutorial coverage for the AI-native stack — MCP, agent auth, and fine-grained auth are where audience interest and product relevance are lining up right now. ( impact: high, effort: medium )
4. Repackage interviews around topics instead of guest labels — If interviews stay, they need a topic-first wrapper instead of a guest-first wrapper. ( impact: medium, effort: low )

Context

• Goal: Brand and thought leadership. Be seen as practitioners who walk the walk.
• Audience: Builders of AI-native apps and services.
• Team: Solo producer (chantastic).
• Bottleneck: Post-production and editing.
• North star: Prospects reference the channel in customer discovery calls.

Snapshot

• Date: 2026-03-09
• Subject: WorkOS (@WorkOS)
• Videos analyzed: 176 total, including 172 uploads plus 4 live stream archives not present in the flat-playlist output
• Total views: 370,551

This channel is not trying to be a creator channel.

It is trying to build trust. That matters when reading the numbers.

Format Breakdown

Short average is heavily skewed by “Introducing WorkOS AuthKit” at 138K views. Median Short views: 239.

Format Definitions

• Tutorial: Integration walkthroughs showing how to build something with WorkOS.
• Feature Walkthrough: Product-focused demos showing what a WorkOS feature does.
• Stage Presentation: On-stage talks, keynotes, fireside chats, and historical release events.
• MCP Night Talk: Speaker talks from the WorkOS MCP Night meetup series.
• Conversation: Peer-to-peer conversations and live streams between practitioners.
• Conference Talk: Enterprise Ready Conf session recordings.
• Product Launch: Feature announcements, launch recaps, and dashboard updates.
• Interview (Booth): One-directional conference interviews recorded against a branded backdrop.
• Short: Videos under 60 seconds.
• Event Recording: Raw event recordings not broken into segments.

Top 10 Performers

Findings

1. Tutorials are the workhorse

Tutorials average 3,417 views. Nothing else on the channel does the same steady job of earning attention and proving usefulness.

They also fit the actual production reality. One solo producer can make these repeatedly without inventing a new workflow every time.

What is working:

• Specific, actionable titles with framework names like Next.js, Laravel, and Remix
• Time promises like “in 7 minutes” or “in 3 minutes”
• Strong thumbnail identity: chantastic to camera, bold text, partner or framework logo, dark background
• A repeatable structure: hook, setup, implement, test, close

2. Feature walkthroughs are solid, but a little loose

Feature walkthroughs average 2,466 views, though that number gets a bump from the 36K “Radix UI Overview” outlier. Even backing that out, the format still clears the channel median.

The bigger issue is consistency.

The title patterns drift. The thumbnail system drifts. The series framing drifts.

There is something here. It just is not packaged as tightly as it could be.

3. Interviews are eating time and giving almost nothing back

There are 36 booth interviews. That is 20% of the catalog.

They produce 1.6% of total views and land at a median of 94 views.

That is a bad trade.

Why they underperform:

• Thumbnails are visually identical
• Titles are generic and guest-first instead of insight-first
• Multiple interviews were batch-published, competing with each other
• Guests have little incentive to share because the packaging is not built for discovery

The one exception, “Interview: Dexter Horthy (HumanLayer) on Research-Plan-Implement,” proves the point.

The clearer the topic maps to audience pain, the better it does.

4. MCP Night talks and stage content work better than they look

Event-sourced talks do better when the topic is hot and the title is sharp.

“The Browser MCP is all you need” is a good example. Specific. Opinionated. Searchable.

The newer AI- and MCP-shaped talks are also clearly stronger than the older GTM and release-event era.

5. Conversations are the most interesting underused format

There are only three conversation videos, so the sample is tiny.

Still, this is the format with the best structural fit for the channel’s real goal: show that WorkOS people actually do this stuff.

It also dodges the biggest bottleneck. Live streams do not ask for a heavy edit pass.

That matters a lot.

6. Shorts are promising, but the average lies

The 21,535 average is fake-nice because one AuthKit launch Short does most of the lifting.

Still, the pattern underneath it is useful:

• Shorts that announce something exciting work
• Shorts that quietly demo a feature mostly do not

7. The pocket is 3-8 minutes

The 67 videos in the 3-8 minute range average 2,497 views.

That is comfortably above the channel average, and it also matches the editing reality of a solo producer.

Duration Analysis

Short average is skewed by a single outlier.

Thumbnail Analysis

Tier 1: Strong
Tutorials and recent feature content use chantastic’s face, bold product or framework text, and dark high-contrast backgrounds. This is the clearest visual system on the channel.

Tier 2: Neutral
MCP Night talks and stage presentations are functional but not especially grabby. They look like event recordings because they are.

Tier 3: Invisible
Interviews use nearly identical booth-backdrop thumbnails with no hook, logo, or short text system. They blur together in a feed.

Key Findings

1. Tutorials are the strongest format, but they are only 13% of the catalog.
2. Interviews are the weakest format, but they are 20% of the catalog.
3. The 3-8 minute range is the best fit for both performance and production.
4. Conversations may be the best fit for the channel’s credibility goal.
5. MCP is the hottest topic in the current mix and should shape what comes next.

Recommendations

1. Lean into conversations as live streams

This is the cleanest move on the board.

It removes the editing bottleneck and leans straight into the trust goal. Two practitioners, one live stream, one real topic. That is the whole pitch.

Suggested cadence: weekly or biweekly, with a very clear topic in the title.

2. Shift interview production time to tutorials

Tutorials average about 21x the views of interviews.

That is not a small gap.

The format is proven. It is repeatable. It also does a better job of showing usefulness.

This does not mean “never interview anyone again.” It does mean the bar should be much higher.

3. Increase tutorial coverage for the AI-native stack

This is where audience interest and product relevance are meeting right now.

High-opportunity areas:

• MCP server auth across more frameworks and languages
• Agent-to-agent auth patterns
• Fine-Grained Authorization implementation walkthroughs
• AuthKit paired with AI frameworks such as LangChain, CrewAI, and the OpenAI Agents SDK

The audience has already said yes to quick, practical MCP content.

4. Repackage interviews as topical content

If interviews stay, the wrapper has to change:

• Lead with the insight, not the word “Interview”
• Add company logos and short hooks to thumbnails
• Stop batch publishing
• Clip one or two Shorts from each conversation so the guest has something shareable

5. Extract Shorts from existing long-form content

This is probably the cheapest way to widen the channel surface area.

MCP Night talks, conversations, and stage presentations should all be able to cough up 30-60 second clips worth shipping.

Methodology

• Catalog source: yt-dlp --flat-playlist for the main catalog, plus a separate pull from the /streams tab for live stream archives not captured in that output
• Data gaps: Flat-playlist output does not include upload_date, like_count, or comment_count
• Categorization: Formats were initially auto-categorized by title patterns and then refined through direct discussion with the channel owner
• Thumbnails: Visually reviewed for the top 30 videos by views and the bottom 5 interviews

Limitations

• View counts are point-in-time snapshots from 2026-03-09
• Upload-date gaps make time-normalized comparisons harder
• Like, comment, retention, and traffic-source data are not included here
• Some observations about talks and thumbnails are qualitative, not purely numeric