Quick context before the story: I run this blog (bigguyonstuff.com) almost entirely through Claude Code. I built Trigli, an AI customer support SaaS, with Claude as my pair programmer. I generate animation code, Canva slide drafts, and half of my written ideation through Claude conversations. It’s in my daily workflow the way a text editor or a terminal is.
People ask me “why not ChatGPT?” constantly. Usually friends who are still on it, occasionally people who assume I must have tested both extensively with matched prompts and run spreadsheets.
I haven’t. This isn’t a benchmark post. It’s the honest story of why I switched, what kept me here, and the parts where I’m going to be careful not to over-claim.
If you’re on ChatGPT and it’s working for you, this post is not an argument to switch. It’s the answer to “why did you switch, and why do you still use Claude?” for anyone who wondered.
Where I started
ChatGPT, like a lot of people. I was using it for writing drafts, brainstorming, the occasional coding question, general-purpose “how do I phrase this” help. It was fine. It was the first one I tried, and at the time, it was the only one anyone was talking about.
Nothing dramatic happened to push me away. I just started trying Claude on a whim, mostly because the internet kept saying it wrote better prose, and I was writing a lot.
The writing switch
The first thing I noticed with Claude was that its writing sounded less like a college freshman trying to hit a word count.
I know that’s subjective. I know “less robotic” is the kind of thing everyone claims about their favorite model. But for me, specifically, the difference was in a handful of habits:
Fewer preamble paragraphs. ChatGPT output I was getting at the time loved to restate the question, explain what it was going to do, and then do the thing. Claude would just do the thing.
Less scaffolding language. You know the phrases. “It’s important to note that.” “In today’s fast-paced world.” “Here are some key considerations.” Claude’s drafts had fewer of those, more of them in the “actual content” department.
Better at tone-matching. When I gave Claude a voice sample from my own writing and asked it to help draft in that voice, the output felt closer to me than what I was getting elsewhere. That part is huge for someone publishing under their own name.
None of that is a benchmark. It’s “the output hit my ear better.” For writing, that’s the only benchmark that matters, because you’re the one publishing it.
The coding switch
The bigger shift was Claude Code.
I live in a terminal. When Claude Code came out (a CLI-based agentic coding assistant from Anthropic), it landed exactly where I already worked. I didn’t have to paste code into a chat window, get an answer, paste it back into my editor, run it, paste the error back. Claude Code could read my files, run commands, write code, check results, and iterate, all from the command line.
That workflow alone was the win for me. Most of my previous AI-coding pain was context-switching overhead. A tool that lives where the code lives solves most of it.
Since then, Claude Code keeps shipping updates that compound. Subagents. The CLAUDE.md project config file. Skills. Real filesystem access. Tool use that actually feels tool-use-y and not tool-use-cosplay. This blog’s entire multi-agent content pipeline runs on Claude Code: a planner agent, a researcher, a writer (hi), a fact checker, an SEO strategist, an image creator, a publisher. They hand work off to each other through markdown files and JSON status files, and Claude Code orchestrates the whole thing. (I’ve thought hard about the blast radius and guardrails for a setup like this, and that’s a separate post.)
I could not have built that on the tooling I was using two years ago. The primitives didn’t exist yet.
The ChatGPT side may have shipped equivalent tooling in this time. I haven’t gone and re-evaluated. Which brings me to the part where I have to be careful.
What I am NOT going to claim
Let me be very clear about what I’m not saying in this post:
I’m not saying Claude beats ChatGPT at coding benchmarks. I haven’t run the same prompts on both models on current tasks. The third-party benchmarks I’ve seen lean toward Claude on some tests and OpenAI on others, depending on the week and the test. I’m not a benchmarking authority and I’m not pretending to be.
I’m not saying ChatGPT is bad. A lot of very smart people use it daily and get great work out of it. If your workflow is running on ChatGPT and you’re happy, stop reading this post and go ship something.
I’m not saying you should switch. This is a “why I switched” post, not a “you should switch” post. The right tool is the one your brain clicks with and your workflow can accommodate.
The reason I’m being this careful is that I see a lot of AI-tool posts that overclaim. “Model X is better at everything.” It’s usually not true. Different models are better at different things for different people on different days. Claude works for me. That’s the whole argument.
What Claude actually does for me in 2026
Here’s the concrete list of what I reach for Claude (either the chat app or Claude Code) for on any given workday:
This blog. Claude Code coordinates the full editorial pipeline. From topic research to fact checks to WordPress publishing. Posts get drafted, reviewed, image-generated, and scheduled with Claude running the coordination between specialized agents. I review, approve, and ship. That’s the whole workflow.
pickmysquare.com development. When I added March Madness Bracket Pools to the site, Claude helped me build the entire promo animation. Not an image. Actual animated code that lives on the site. I described what I wanted, Claude wrote the HTML/CSS/JS, we iterated, and it shipped.
Trigli features. Trigli is a real SaaS product with real customers. Claude Code helps me ship features, debug weird issues, refactor, and build out documentation. I’m still the one making architectural decisions and reviewing everything, but the implementation velocity is what it is because of the tooling.
Trigli’s landing page animation. Similar to pickmysquare, I’m building the trigli.com product animation with Claude’s help. Not generated images: runnable code I can edit forever.
Instagram Canva slides. I produce content for Instagram regularly, and Claude is good at drafting the copy and layout ideas for Canva slide sets before I move them into Canva for final polish.
Writing drafts. A lot of what you’re reading on this blog started as a conversation in Claude. Outline, first pass, rewrite, polish. The editorial pipeline I mentioned above catches the rest.
That’s a pretty ordinary week. None of this is special-case usage. It’s the boring, everyday “knowledge work meets code work” stuff that a lot of solo founders and creators deal with.
The price conversation
I’m on Claude Max. It’s a higher-tier subscription, more expensive than the base Pro plan, with substantially more usage and access to Claude Code at volume. For my workload (running a full content pipeline plus active coding on Trigli), the math works. I was blowing through the Pro tier’s limits in a single morning, so moving up wasn’t really a choice.
Anthropic’s pricing page is the source of truth for current Max pricing, and it has moved before. If you’re evaluating, check it fresh before you sign up. What I’ll say is: for a casual user, Pro is almost certainly enough. For someone using Claude Code agentically all day, Max starts making sense.
(Open question: if I were still on a day job and just chatting with Claude occasionally, I’d probably be on Pro and perfectly happy.)
What I give up
Claude isn’t a free lunch. A few honest downsides from my own use:
Rate limits during heavy coding sessions. Even on Max, if I’m running a long Claude Code session with lots of file edits and tool calls, I can hit usage limits mid-task. It’s rare but it happens. The workaround is pacing or splitting work across days. Long sessions also surface some interesting behavioral patterns worth knowing about.
Ecosystem plugins. OpenAI has a richer plugin and integration ecosystem with third-party apps. If your workflow depends on a specific integration that exists for ChatGPT and not Claude, that matters.
Voice mode. I don’t use it. If you do, ChatGPT’s voice experience has been the more polished one.
Availability outages. Every AI service has them. Anthropic has had some over the last year. So has OpenAI. If you’re dependent on any single model for your workflow, you should have a fallback plan regardless of which side you’re on.
When I’d tell you to use ChatGPT instead
I don’t want to sound like a fanboy, so here’s the honest list of when ChatGPT is the better call:
- If OpenAI’s plugin or integration ecosystem is load-bearing for you. If you need a specific third-party hook that only ships for ChatGPT, use ChatGPT.
- If you’re already fluent in it and don’t want to relearn. Workflow muscle memory is real. Switching tools has a cost. If you’re shipping work on ChatGPT, shipping > switching.
- Voice-first interaction. If voice mode is how you actually want to use an AI, ChatGPT’s voice experience is polished in a way I haven’t seen matched.
- Anything where OpenAI’s specific model characteristics fit your task better. Some users find GPT-5.5 or GPT-4o outputs better on specific writing styles or technical domains. If you’ve tested it on your actual work and it wins, trust your results over my story.
Why I still haven’t gone back
Short version: nothing has broken. Claude keeps getting better. My workflow is built around Claude Code at this point, and switching costs something real in terms of re-learning my own tooling. More importantly, the outputs are still hitting my ear right for writing and still solving the coding problems I actually face.
The moment that changes, I’ll re-evaluate. I’m not married to Anthropic. I’m using the tool that does the job. Right now, that’s Claude. In a year, it might be something else. The honest answer is “I will use whatever tool is best for my workflow,” and today’s answer is Claude.
What’s your daily driver?
If you’ve switched one way or the other (or you use both), I’d like to hear what made it stick. Which workflow clicked, what frustrated you about the alternative, and whether you’ve re-evaluated recently.
Drop a comment with your setup. Share this with a friend who’s on the fence. And if you want the deeper walkthrough of how this blog runs on Claude Code, that post is already up: Claude Code for bloggers.
Whatever you’re using, keep shipping. The tools matter less than the output. 🔧
Sources
- Anthropic Claude pricing (current subscription tiers and Max pricing)
- Anthropic Claude Code documentation (CLI coding tool features and agentic workflow)
- XDA Developers – I switched from ChatGPT Plus to Claude Pro and haven’t regretted it
- Medium / Ravindu Himansha – I Switched from ChatGPT to Claude: What Actually Changed
- Dale Dumbs It Down – 90 Days After Switching to Claude
- FindSkill.ai – Claude vs ChatGPT 2026: Why Devs Are Switching
- Towards AI – I Switched from ChatGPT to Claude: What Nobody Tells You
