The Playbooks AI
WEEKLY·ISSUE 007·July 1, 2026
Welcome

Stop treating every task with AI like a brand-new chat.

Hey there! This week I really want to show you a practical starter kit for Claude Projects: what it is, when to use one, and how to actually start.

A normal Claude chat is still great for one-off questions; that is what most people use. A Claude Project, on the other hand, is better when the same topic, files, instructions, or goal will come up more than once. If you keep pasting the same background into Claude, that is a sign the work probably belongs in a Project.

I recorded three pieces for this: set up Claude, what Claude Projects are, and how to use a Project to start a scientific research report. Go in order if you are new to Claude, or jump to the research-report playbook if you want to see what using a Project actually looks like.

What's in this issue
  • Beginner: set up Claude and understand Projects. Learn what Projects are, when to use them instead of a normal chat, and how to create one with documents Claude can reuse.
  • Intermediate: use a Project for a research report. Add notes and sample data to one Project, then use Claude to summarize, visualize, and draft the start of a report.
  • Advanced: know when to switch from Claude Chat to Claude Cowork. Regular Project Chat is enough for most readers. Cowork is the next step when a Project needs to become a more task-oriented workspace with richer file and folder context.
  • Also this week: new AI models, access, and chips. A few quick AI updates are worth tracking this week: frontier AI model access, AI cost control options, and the hardware race underneath faster AI tools.
The Playbook
BeginnerLevel 1

Start with the basics: one topic, one Project.

Video thumbnail for the Claude Projects walkthrough, showing files inside an Ocean Waves research project.
Video: create an Ocean Waves research Project and ask Claude to use the files.

If you have not set up Claude yet, start with the Claude setup video. If Claude is already installed and you know where Projects live, you can start here.

Think of a Claude Project as a reusable workspace for one topic. Instead of opening a blank chat every time, you give Claude the relevant files and instructions once, then keep related conversations inside that Project.

  • Use a normal chat when: the question is one-off and you do not need Claude to remember a specific file set or workflow.
  • Use a Project when: you keep referencing the same topic, notes, source files, instructions, rubric, audience, or output format.
  • Bring the right context: add documents, notes, examples, or files that you want Claude to use repeatedly.
  • Keep the instructions simple: tell Claude the goal of the Project, what files to trust, what tone to use, and when it should ask questions instead of guessing.

For the beginner path, the goal is not to build a giant AI workspace. The goal is to understand the difference between a chat and a Project, then make one Project that proves the value of shared context.

IntermediateLevel 2

Use the Project to start a scientific research report.

Video thumbnail showing Claude creating a chart from synthetic wave-height data inside a research project.
Video: use Claude Projects to jumpstart a scientific research report.

Once you have a basic Project created in Claude, the next step is to use it for actual work. In this demo, we use a beginner-friendly research report about ocean waves to show what that looks like.

The playbook follows the Ocean Waves demo directly and includes the files and Project instructions so you can follow along. After that, the structure is reusable for your own topic: add context, add instructions, check Claude's understanding, then ask for the first useful output.

Here is the simple structure for getting value out of a Project:

  • Add the context: upload the files that define the work. In the demo, that is the ocean-wave research packet and a small sample dataset.
  • Add Project instructions: tell Claude the goal, audience, source rules, and output style once so future chats start from the same frame.
  • Check understanding first: have Claude summarize what files and data it can see before asking for a chart, outline, or draft.
  • Create a useful output: ask Claude to turn the files into something you can use, like a report introduction, draft structure, summary, graph, or visual idea.

That is the core playbook: put the research materials, instructions, and starter prompt in one Project so every chat can work from the same source base.

AdvancedLevel 3

Go further: switch the Project from Chat to Cowork.

For most readers, regular Project Chat in Claude is enough. Upload the right files, add instructions, ask better questions, and keep related chats in one Project.

If you want to take Projects further, then try Cowork. It is the more advanced, plan-dependent lane where a Project can become more task-oriented, especially when you want Claude to work with richer context from real files, folders, and connected workspaces. Claude has help docs for Projects in Cowork and recurring Cowork tasks if you want the official details.

Here are the cases where Cowork starts to make more sense than regular Project Chat:

  • Stay in Project Chat when: you are asking questions, summarizing uploaded files, drafting from Project knowledge, or learning the workflow.
  • Try Cowork when: the work lives in real folders, local files, connected docs, or a broader project workspace that needs more than uploaded Project knowledge.
  • Start review-only: Cowork can access files and do more than regular chat, so begin by asking it to inspect context, explain what it can do, and propose a plan before you approve file changes or bigger tasks.
  • Use scheduled tasks later: once the workspace is organized, recurring Cowork tasks can help with check-ins, research updates, or project maintenance.
Safe Cowork starter prompt
Using Cowork, inspect the project or folder context you can see and help me understand what work this workspace is ready for.

Do not edit, move, delete, upload, or publish files unless I explicitly approve it.

First, inspect the project context you can see.
Then propose:
1. What files, folders, or connected context seem important
2. The best tasks you can help with from this workspace
3. Any missing files, permissions, or instructions
4. A safe first task you recommend I approve

If your Claude screen looks different from the video, do not worry. The principle is the same: Project Chat is the basic reusable-context workflow; Cowork is the more advanced task/workspace layer.

Also this week

Was this week's issue useful?

Got a sec? Tell me what to write more of.

See you next week,

Ky Tomita, The Playbooks AI