@masayan-ai-hack Channel Audit: 1,790 Subs, 288 Videos, Growth Read

First thing that jumps out: 288 videos to get to 1,790 subs. That's about 6.2 subs per upload, which is on the low end even for technical Japanese-language content. The lifetime view total of 157,791 spread across nearly 300 videos puts the per-video average at ~548 — so the channel is publishing consistently but most videos are landing somewhere between 100 and a few thousand views rather than producing any clear breakout.

The niche itself is interesting and arguably the most defensible thing here. The description lays out a very specific stack: MCP, Claude Code, AI駆動開発 (AI-driven development), automated testing, design patterns, clean architecture, performance tuning. That's not a generic "AI channel" — that's a working tech lead's actual day-to-day. In the Japanese-language AI engineering corner of YouTube, there genuinely isn't a flood of people going deep on MCP servers or Claude Code workflows. So the positioning is solid. The question is whether the content is reaching the people who'd care.

Upload cadence is where things get a bit hard to read from outside. 288 videos over what appears to be roughly four years of activity suggests something like 5-6 uploads per month — pretty heavy by long-form standards. The most recent 30 uploads are all long-form, zero Shorts. That's a deliberate choice and probably the right one for this audience (engineers don't really discover deep technical content through Shorts), but it does mean every single piece of content is fighting for attention without the algorithmic lottery ticket that Shorts can sometimes provide.

The "0 views" numbers showing up on recent uploads in the scrape are almost certainly a data freshness issue — either these videos were published in the last few hours, or the scrape hit a window where the public counter hadn't refreshed. Worth flagging because it makes the recent performance impossible to evaluate from this snapshot alone. If those numbers are actually real after 48+ hours, that'd be a much bigger signal about what's happening with distribution.

What's working, based on what's observable: the multi-platform presence is real. A Substack, an X account, and a tech blog with 400+ articles is a serious content surface. That's the kind of footprint where the YouTube channel probably isn't the primary funnel — it's one node in a network. Which actually changes how you should evaluate the subscriber count. 1,790 YouTube subs in isolation looks modest. 1,790 YouTube subs plus blog readers plus Substack subscribers plus an X following looks like a developer who's built a meaningful audience and YouTube is just one expression of it.

The gap I'd flag: there's no obvious evidence of a tentpole video — the kind of 20K-50K view explainer that becomes the discovery hook for the whole channel. With 288 uploads and 157,791 total views, the math suggests pretty even distribution rather than a power-law where one or two videos carry the channel. In an AI/MCP niche where search demand is exploding, one well-structured "MCP完全解説" or "Claude Code実践" video that catches the wave could shift the entire trajectory. Tech lead-level expertise + a topic with rising search volume is the right setup for that kind of swing.

One aside that's probably worth saying: publishing 288 videos while also writing 400+ blog posts and running a Substack is genuinely a lot of output. The risk with that kind of volume is that nothing gets the polish treatment. From the outside it looks like the channel is in "keep shipping" mode rather than "engineer one big hit" mode, and those two strategies often pull against each other. If I were giving notes as a peer, I'd say try one quarter where you cut output in half and put the saved time into one really thorough video on whatever's hottest in the Japanese MCP/Claude Code conversation right now. See if that breaks the ~548 view ceiling. If it doesn't, the niche is just harder to grow than it looks. If it does, you've found the lever.