Let's Talk About "Claude"

You love it, I love it. Claude is all over my feed and I love hearing stories about marketers nerding out in Claude Code.

But let’s dissect “Claude” a bit to make sure we understand what it is, and the other alternatives that exist.

When you peel back Anthropic, there are two parts to what you think of as Claude: the “harness” i.e. the user experience on top of the model, and then the models themselves like Fable, Opus, Sonnet, etc.

For most of us, when we express our love for Claude, we’re mostly talking about the harness. There are three ways to use Claude: the traditional chat experience, Claude Cowork, and Claude Code. Cowork adds a ton of useful connectors, scheduling, skills, live artifacts… all the stuff we need to complete business tasks. It’s probably the best general purpose harness available today. Code adds even more, like looping, goal setting, and all sorts of quality of life improvements for engineers.

But there are other great harnesses. On the business side, you’ve got companies like Optimizely, Profound, and HubSpot building marketing harnesses. On the code side, you’ve got Codex, Copilot, Cursor, and many more. For more general purpose agents, you’ve got OpenClaw and Hermes.

Cowork and Code are both fantastic, and are probably the best harnesses available today, but this space is ultra competitive, and others are closing the gap every week.

Next, you’ve got the models. Anthropic is most known for its “frontier models” like Fable and Opus that sit at the edge of what’s possible. These are also the most expensive and compute-intensive models. Are they great? Of course. Do you need them? Not quite as clear.

That’s because the gap between Anthropic’s best models and other closed and “open” model providers continues to narrow every few weeks. Most recently, Kimi K3 has produced benchmarks showing it’s nearly as good as Opus for most thinking use cases i.e. complex jobs like writing code.

And that’s important because models like Fable and Opus are expensive, and you can get similar performance from alternative models at a much lower cost. Or even free, if you can afford a cluster of Mac Studios to run the models locally :)

Chances are you’ve just set Cowork or Code to use Opus, and that’s fine because your company is probably paying for it, but it’s overkill. I’ve been using OpenClaw with a few different open models like GLM 5.2 + DeepSeek and… I really don’t see much of a difference.

“Claude” is great. The harnesses are the best in the biz at the moment and you can’t go wrong with any of the latest Anthropic models. But like everything else in marketing, when everyone jumps on the same bandwagon, that creates opportunities for the bold to innovate by doing something different. Are there way better alternatives to Anthropic at the moment? Not really.

Might that change next month? Maybe. That’s what I’m here to help you figure out 💜

Originally posted on LinkedIn.