Vì Sao Mình Bỏ ChatGPT Sang Claude

Summary

The video argues that the creator has made Claude—especially Claude Code—his primary AI tool after years of using ChatGPT. His comparison is based on personal use, including a month of testing paid plans from both providers. He presents his conclusions as personal experience, not as a definitive benchmark.

Why he prefers Claude

• Focus and business direction: He argues that Anthropic has concentrated on making Claude useful for professional and enterprise work, particularly coding. In his view, OpenAI is spreading its efforts across more products, which can divert resources from its core tools.
• Product features and pace: He says Anthropic has often introduced features he finds useful for work—such as MCP integrations, Skills, Artifacts, and Claude Cowork—before comparable features appeared in OpenAI products.
• Model experience: He considers the leading models from both companies broadly competitive in intelligence. Personally, he finds Claude quicker to understand what he wants and better at matching his intended style, especially for design and visual tasks.
• The harness matters as well as the model: He explains that an AI’s “harness”—its system instructions, tools, integrations, workspace, memory, and ability to delegate work—strongly affects results. He thinks Claude’s work-oriented tools make it more effective for multi-step tasks.

Practical comparison: A YouTube research app

To compare the tools, he gave a top model from each company the same prompt: build an app to analyze successful videos from selected YouTube channels. The app was meant to examine video performance, titles, thumbnails, topics, and opening hooks, then recommend ideas for future videos.

• ChatGPT’s result was clean and usable, but comparatively shallow in its explanations and recommendations.
• Claude’s result provided more detail in both its implementation plan and finished app. It explained data sources and processing, identified difficult parts, analyzed individual videos and thumbnails, and connected titles, viewer emotions, and the first 30 seconds of a video.

The creator’s takeaway is that Claude more often provides useful detail without requiring repeated prompting. He also acknowledges that ChatGPT’s interface looked cleaner in parts of the comparison.

Cost and when he would still use ChatGPT

The creator says that, despite similar-looking subscription prices, Claude let him complete more work in his own testing. He also cites a SemiAnalysis comparison of higher-priced plans that, according to his account, found a substantial difference in the amount of usage provided. These are claims and comparisons presented in the video, not independently verified here.

He gives three reasons someone might continue using ChatGPT:

1. Their company already uses it, and they have little choice.
2. They need image generation. He says OpenAI’s image-generation tools are strong, while Claude can understand images but does not generate them.
3. They want to use a third-party agent or harness and prefer OpenAI’s subscription access for that setup. He describes Anthropic’s policies for third-party tools as having changed and being less predictable.

Guide included

The video includes a short ChatGPT-to-Claude memory transfer tutorial:

1. Prompt ChatGPT to summarize what it remembers about the user, organized into five categories.
2. Paste the result into Claude Code along with a second prompt.
3. Start a fresh Claude Code session and ask what it knows to check that the transfer worked.

Main speaker and sources

• Main speaker: Nhat Anh, sharing his personal experience as an AI tools user and startup operator.
• Sources and references mentioned: Artificial Analysis model rankings, WebDev Arena user comparisons, and a SemiAnalysis report comparing Claude and OpenAI subscription usage.