AI Wrestle Mania
Are we calling it LLM or AI? Or is it a model?
I'm freaking out, man
It seems like everyone is freaking out about AI. OpenClaw, Claude, ChatGPT, models, agents, and everything. Is this FOMO (Fear Of Missing Out) or something else? Has the playing field of the tech world been leveled and we're all looking for an edge. Maybe we all just want to have the best helpers so we can brag about it to our friends.
You are freaking out... man
There is something inherently cool about the idea of having multiple providers/models/agents talking to each other. Like "Brick Tamland" is really good at telling the weather because he uses Kimi. And "Ron Burgundy" is actually terrible at the news, but people love him for some reason.
Slop
I love creating new projects. I love writing code to learn about how something works. Those projects usually end up dying in a private repo somewhere. The only difference is now I create and abandon those projects faster than before.
Frustration
In hopes to save money and work on a project efficiently, I'll sometimes use a "dumber" model. Then after several attempts, I threaten it by yelling "Let me talk to your boss." I've become an AI Karen. A kAIren.
I'll also spend hours to get something to work, like OpenClaw, only to realize the software is actually very buggy. This can be very frustrating. The product looks mature, they have great branding, everyone is talking about it, but the bugs bite you. It seems like OpenClaw is too busy trying to add more functionality than trying to fix the bugs. Maybe that's a harsh criticism because it is so new.
Reviews
These reviews are based on my own experiences. They're my opinions.
Claude Code
I've been primarily using Claude Code for just about everything. The terminal is where I was born, so it feels like home. It can do just about anything I need.
Right now, I'm having it search for neighborhoods in the Cincinnati area based on certain criteria. Then after it has collected all the data, it's going to build an HTML page that I can share with my wife and dogs.
I love the "opusplan" model right now. It'll use Opus to create the plan, then Sonnet or Haiku to make the changes. It feels more economical this way. It gives you a great plan, but then has the minions do the work.
OpenClaw
I tried it, got it working about as well as anyone else, but then gave up on it. I decided the energy required to keep it working properly wasn't worth what I was getting in return.
Also worth noting is that Anthropic is making some changes to how you pay for usage.
One note: starting April 4, third-party harnesses like OpenClaw connected to your Claude account will draw from extra usage instead of from your subscription. If you don’t use them, nothing changes. If you do, the credit and bundles above have you covered.
I'd rather see people give up OpenClaw and stick with Anthropic, but some of them will probably go crawling back to OpenAI.
OpenRouter
This turned out to be cooler than I thought. I really like how they'll route to multiple providers. For example, if I'm trying to use openai/gpt-oss-120b, it might send the request to DeepInfra or Clarifai.
OpenRouter provides a unified API that gives you access to hundreds of AI models through a single endpoint, while automatically handling fallbacks and selecting the most cost-effective options.
Price and Performance. OpenRouter scouts for the best prices, the lowest latencies, and the highest throughput across dozens of providers, and lets you choose how to prioritize them.
Cloudflare AI Gateway
I would stay away from Cloudflare AI Gateway. I want it to work well, but it doesn't. About 50% of my requests fail. It's extremely frustrating. The sad part is, they've known about this for over a year. Cloudflare AI Gateway is broken.
They host some models and will run them on the Cloudflare infra, but they also proxy requests to other providers (like OpenRouter does).
Gemini
Gemini works pretty well. I only recently started trying it out, so not much to report right now. The TUI is pretty.
ChatGPT
I've been using this one the longest, but I don't have much to say. It works pretty good. It's not amazing, but it's not bad either. If I'm running everything else to the max and I need a simple question answered, I'll ask ChatGPT.
Conclusion
A lot is happening. And it's happening very fast. Be sure to take a break every hour or so. And on your break, focus on something else. Don't think about AI on your break. Go for a walk if you can or phone a friend. Don't go outside and talk to a LLM on your phone; that's not taking a break.