What Claude Projects Actually Does That Regular Claude Cannot

I was halfway through explaining my company's style guide to Claude for the fifth time that week when I finally caved and tried Projects. Should have done it months ago.

Here's the thing about Claude's regular interface: it forgets you exist the moment you close that tab. Every conversation starts from zero. You're a stranger every single time. Projects changes that in ways I didn't fully appreciate until I'd been using it for about six weeks.

The Actual Difference Nobody Explains Well

Regular Claude gives you a blank canvas every conversation. Projects gives you a pre-loaded canvas with your stuff already on it. That's the entire concept, but the implications go deeper than most people realize.

With a Project, you upload documents, set custom instructions, and create a persistent workspace. Claude reads those materials before you even type your first message. It's not memory in the way humans have memory — it's more like Claude waking up with notes already on its desk.

I tested this by creating a Project for my newsletter. Uploaded my last 20 issues, added instructions about my tone and audience, included a doc with topics I've already covered. Now when I ask for help brainstorming, Claude doesn't suggest obvious stuff I wrote about six months ago. It knows the territory.

Regular Claude would have given me generic newsletter advice. Project Claude suggested building on a specific point from issue #14 that readers had responded well to. That's a different kind of useful.

What I Was Wrong About

I assumed Projects was basically just a way to avoid re-uploading the same PDF every time. Like a convenience feature, nothing more. Turns out the custom instructions are the actual genuinely useful, not the file uploads.

You get this field where you can write essentially a briefing document for Claude. I started with basic stuff: "Write in casual tone, avoid corporate jargon." But after a few weeks, my instructions got weird and specific.

Now my main writing Project has instructions like: "When I share a draft, assume the first paragraph is always too long. I overwrite openings. Point this out even if I don't ask." And: "If I'm working after 10pm, I'm probably tired and making dumb mistakes — be more critical than usual."

This is stuff I'd never think to tell Claude in a regular conversation. But because I built it up over time in a persistent space, the interactions got tighter. Claude in my Projects feels calibrated. Regular Claude feels like a first date where you're still explaining your job.

The Kick: Projects Have a Ceiling Nobody Mentions

Here's what I only figured out after hitting the wall repeatedly: Projects have a context limit that your uploaded files eat into. This sounds obvious but the implications are brutal.

I uploaded maybe 30 documents to my research Project. Style guides, source materials, reference docs. Felt organized. Then I tried having a long conversation within that Project and Claude started losing the thread way faster than usual. Kept forgetting things I'd said three messages ago.

What's happening: those uploaded files take up space in Claude's context window. Every file you add means less room for the actual conversation. So if you upload a 50-page document, you've basically traded conversation depth for reference material.

The fix I found: be ruthless about what actually needs to be in there. I went from 30 documents down to 8. Pulled out anything Claude could figure out from context. Kept only the stuff that would be genuinely impossible to recreate — proprietary info, specific examples, exact wording that matters.

Nobody mentions this tradeoff. The feature descriptions make it sound like "upload everything, Claude will handle it." In reality, you're playing a resource allocation game. More prep materials means shorter useful conversations before Claude starts getting confused.

Where Projects Actually Falls Short

Can't share a Project with someone else. Can't have two people work in the same Project space. Can't export the custom instructions and files as a bundle to move to another account. These all feel like obvious features that don't exist.

Also, the files you upload aren't searchable by you. Claude can reference them, but you can't ctrl+F through your own uploaded docs within the interface. Had to keep a separate folder on my computer with copies of everything, which defeats some of the organizational appeal.

The mobile app situation is weird too. Projects exist there but the experience is clunky enough that I mostly just use it on desktop. Starting a conversation in a Project on mobile felt like I was fighting the interface.

And the artifacts — the code previews and document outputs — don't automatically save to your Project. They vanish with the conversation unless you manually copy them out. Lost a decent chunk of work this way before I figured out the workflow.

Who Should Actually Use This

If you talk to Claude about the same general topic more than twice a week, Projects makes sense. The setup time — maybe 20 minutes to upload files and write good instructions — pays back fast.

If you're using Claude for random one-off questions, regular Claude is probably fine. The overhead of maintaining Projects isn't worth it for casual use.

I now have four Projects running: one for this blog, one for client work, one for a side project, one experimental space where I test weird prompts. The boundaries help. Keeps things from bleeding together.

The strange thing is how quickly you forget what it was like before. I opened regular Claude last week to test something and felt genuinely annoyed at having to explain context. Like going back to a flip phone after—

Heads up: Some links in this post may be affiliate links. I only recommend tools I've personally tested. Opinions are entirely my own.