Video Analysis

The day before yesterday, I uploaded four long-form videos to YouTube.

They received essentially zero views and generated no recommendations.

Only one video—which compared Supabase and PocketBase—has managed to garner a few views so far.

While browsing Reddit, I came across an advertisement for Supabase, which allowed me to confirm two key points:

First:

Videos regarding Supabase are likely to attract some traffic, as it is currently a particularly popular open-source backend project.

Second:

I understand that PocketBase is a solo open-source project. Although I personally like it and use it myself, it does not offer a cloud-hosted service, nor does it possess the funding required to run advertisements.

In contrast, Supabase offers users the flexibility to either self-host a free local version via Docker or utilize its managed cloud service; consequently, it relies on advertising and promotional efforts to acquire customers.

Third:

Indeed, many independent developers tend to prioritize Supabase when making their choices; naturally, they also tend to share technical insights and address issues related to Supabase.

Yesterday, I uploaded two more long-form videos, neither of which received any significant views or recommendations.

This is because both videos documented the challenges I encountered while using Astro to build a blog website.

I searched for “Astro” on YouTube and discovered that the topic generates very little traffic; many of the existing videos are several years old. Therefore, I would advise against sharing content related to Astro on YouTube, as there simply isn’t much of an audience for it.

Today, I uploaded two additional long-form videos, marking the beginning of my content curation efforts in the field of AI.

One of these videos—specifically the one covering Qwen 3.6 35B-A3B—has already garnered a few views and recommendations.

This is because Qwen 3.6 35B is a large-scale, open-source AI model designed for local deployment that was released just this past week. Much like Gemma 4, it is currently generating a massive amount of buzz; people everywhere are paying attention to it, experimenting with it, and sharing their experiences.

Consequently, even though my content may lack entertainment value, offer limited practical utility, and feature no elaborate editing, the sheer power of the video’s title alone was sufficient to drive a certain amount of recommended traffic to it.

Video Optimization

Always remember: YouTube belongs to Google. Therefore, you should pay special attention to Google’s products.

For instance, I plan to experiment with the smaller Gemma 4 model tomorrow. I’ve heard it offers excellent support on mobile devices; while I can’t run the larger 26B and 31B versions, the smaller one is definitely worth a try.

I also intend to explore other tools, such as Firebase.

Additionally, I plan to watch more YouTube videos related to AI.

Alternatively, I might head over to Reddit to browse forum threads posted by new YouTubers.

I want to observe how others operate and manage their YouTube channels.

Comparison

A Table-Based Approach Below, I present a direct, side-by-side comparison of Gemma 4 vs. Qwen 3.6 in a single table (based on practical usage dimensions—no fluff):

Dimension	Gemma 4	Qwen 3.6
🧠 Positioning	Dialogue-oriented / General Assistant	Engineering-oriented / Agent / Task Execution
💬 Conversational Naturalness	⭐⭐⭐⭐☆ (More human-like speech)	⭐⭐⭐☆☆ (More tool-like responses)
💻 Programming Capability	⭐⭐⭐⭐☆	⭐⭐⭐⭐⭐ (Stronger)
🐛 Debugging Capability	⭐⭐⭐⭐☆	⭐⭐⭐⭐⭐ (Significantly stronger)
🧩 Multi-step Tasks / Agents	⭐⭐⭐☆☆	⭐⭐⭐⭐⭐ (Clear advantage)
🔧 Tool Calling / Function Calling	⭐⭐⭐☆	⭐⭐⭐⭐⭐
🖼️ Multimodal Capability	⭐⭐⭐⭐⭐ (More comprehensive)	⭐⭐⭐⭐☆ (Focused on practical analysis)
⚡ Inference Stability	⭐⭐⭐⭐☆	⭐⭐⭐⭐☆ – ⭐⭐⭐⭐⭐
🖥️ Local Deployment Friendliness	⭐⭐⭐⭐⭐ (More resource-efficient)	⭐⭐⭐⭐☆ (Slightly heavier)
🧾 Long-form Writing	⭐⭐⭐⭐⭐	⭐⭐⭐⭐☆
🎯 Ideal Use Cases	Chatting, Writing, Lightweight Applications	Programming, Automation, Agent Systems