what claude actually is and why most people are using the wrong surface

most people only know one version of claude.

Most people who use Claude are using one version of it. They open a tab, type a question, get an answer, close the tab. Maybe they come back tomorrow and do it again. Each time, they start from scratch, re-explaining who they are, what they’re working on, what kind of output they need.

That pattern is not wrong, exactly. It just misses most of what Claude actually is.

Claude exists across three distinct surfaces, and each one is built for a different kind of work. Using the wrong one is like using a sketchbook when you needed a studio, or hiring a contractor for a job that needed a full-time employee. The tool technically works. But you’re spending energy compensating for a mismatch that didn’t have to exist.

the three surfaces

The first is the general chat interface, the one most people default to. It’s conversational, contextless by default, and resets with each new conversation. It’s the right tool for thinking out loud, asking one-off questions, doing quick research, or talking through a decision. It’s not a production environment. It’s a whiteboard.

The second is Claude Projects. This is where the tool changes character entirely. A Project holds persistent standing instructions, a brief you write once that applies to every conversation inside it. You describe the voice, the standards, the workflow rules, the context. Claude reads that brief at the start of every session without you having to repeat it. You can also upload reference files: brand guidelines, templates, examples, documentation. Anything that should inform the work consistently lives here.

The practical effect is that a Project stops feeling like an AI assistant and starts feeling like a trained team member. Someone who already knows your client’s name, your preferred format, the one phrase you never want to see in a headline. You brief them once. After that, you just assign work.

The third surface is Cowork, Claude’s mode for autonomous, multi-step tasks. You hand off a complex job, describe the outcome you want, and Claude works through it independently. You come back to a finished result, not a conversation. It’s the right tool for anything that involves a sequence of actions rather than a single output: research and synthesis, content pipelines, batch processing, structured analysis across multiple inputs.

the mistake most people make

The default behavior, using general chat for everything, costs more than it looks like it does. Every session spent re-explaining your context is friction that compounds. Every output that sounds slightly off because Claude didn’t know your voice is a revision you didn’t need to do. Every task you abandoned because it felt too complex to prompt correctly was probably a Cowork job you tried to run in a chat window.

The issue isn’t capability. Claude in general chat is genuinely useful. The issue is that running production work through a tool with no memory is like running a restaurant where the chef forgets the menu between every table. The meal still arrives. But something is always slightly wrong, and you can never quite build on what worked yesterday.

what changes when you understand the difference

When you match the surface to the work, a few things shift quickly.

Production work moves faster because the context is already there. A Project built around a specific brand or workflow produces consistent output from the first message, not after several rounds of correction. The energy you were spending on setup goes into the actual work.

Complex tasks become manageable because you’re not trying to hold the whole sequence in a single prompt. You describe the goal, the constraints, and the output format. Cowork handles the steps. You review the result.

And the general chat interface becomes more useful too, because you stop asking it to do things it wasn’t designed for. It becomes a thinking space again, which is what it’s best at.

where to start

If you have any recurring work, a client you write for regularly, a content type you produce on a schedule, a workflow you run more than once, build a Project for it. Write the instructions the way you’d brief a new hire on their first day. Be specific about voice, format, what good looks like, and what you never want to see. Upload the reference files that matter. Then use it for a week and notice what you stop having to explain.

That’s the shift. Not a new feature. Not a more sophisticated prompt. Just using the right container for the work you’re already doing.

Everything I’m building on this site, the systems, the workflows, the integrations, runs across all three surfaces. General chat for planning and strategy. Projects for production. Cowork for the tasks that are too structured for conversation and too complex for a single prompt. The distinction is the foundation. It’s worth understanding before anything else.