I switched to Claude a year ago. Writing, building, client work, strategy, code.
I enjoy the experience more. I prefer the outputs. Better model for how I work. And I respect the company building it.
I still use ChatGPT for quick things. But Claude is where I work. This is why.
the company
I align with how Anthropic thinks about AI. Dario Amodei wrote a 20,000 word essay called The Adolescence of Technology about what happens when humanity gets power it is not mature enough to use. That is not a product announcement. That is a CEO thinking about what he is building. Anthropic refused to let the Department of Defense use Claude for mass surveillance or autonomous weapons. Their Constitutional AI operates on a priority stack: safety, then ethics, then compliance, then helpfulness.
Claude pushes back and asks questions. ChatGPT agrees with everything.
the 4Ds
Anthropic's framework for working with AI is the 4Ds.
Delegation. Deciding what to hand off to AI and what stays human. The skill is knowing the line.
Description. Communicating intent precisely. A good prompt is a good brief. Context, examples, constraints, tone. Claude rewards clear configuration.
Discernment. Evaluating output critically. Testing, comparing, knowing when the reasoning broke down. Describe, evaluate, refine.
Diligence. Transparency and ethics. Documenting AI contributions.
configuration over prompting
Most people who say AI does not work for them have a configuration problem, not a tool problem. Both let you provide context. The difference is what happens when you do.
I spend hours in Claude every day. The UI is clean and stays out of the way. Projects keep client work separated. Custom Instructions load into every conversation. No character limit on project instructions. ChatGPT caps at 1,500 characters.
the outputs
Same brief, same context. Claude writes in full sentences, takes a position, handles nuance. ChatGPT defaults to bullet points and filler. Most visible in proposals, emails, strategy docs.
skills vs custom GPTs
A Custom GPT is a standalone app: prompt, files, API actions. A Claude skill is a markdown file that works alongside your Custom Instructions, Projects, and Memory. Not a separate app. A layer in your operating context. Skills work across 30+ platforms. GPTs only work inside ChatGPT.
claude code
Claude Code changed how I build. Terminal agent that reads my codebase, works across files, manages git. I build my site, client tools, and automations with it. OpenAI has Codex. Claude Code is the one I trust.
where things stand
what I use for the gaps
ChatGPT has image generation and more mature voice mode. Claude cannot generate images. For that I use Gemini or Nano Banana. For deep research I use Perplexity. For voice input I use Whispr Flow. Claude's web search handles most of what I need day to day.
Cowork burns tokens. About 10% inconsistency on complex parallel tasks.
None of that touches what I need: an AI that reads my context, follows my instructions, writes in my voice, and produces work I can use.
I am not trying to convince anyone to switch. This is what I chose and why.
A year in, I have not looked back.
comments
No comments yet.