惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

Y
Y Combinator Blog
博客园_首页
量子位
让小产品的独立变现更简单 - ezindie.com
让小产品的独立变现更简单 - ezindie.com
H
Help Net Security
The Cloudflare Blog
钛媒体:引领未来商业与生活新知
钛媒体:引领未来商业与生活新知
A
About on SuperTechFans
腾讯CDC
H
Hackread – Cybersecurity News, Data Breaches, AI and More
D
DataBreaches.Net
L
LangChain Blog
Stack Overflow Blog
Stack Overflow Blog
Jina AI
Jina AI
月光博客
月光博客
云风的 BLOG
云风的 BLOG
博客园 - 聂微东
Hugging Face - Blog
Hugging Face - Blog
雷峰网
雷峰网
C
Check Point Blog
V
V2EX
T
Tailwind CSS Blog
freeCodeCamp Programming Tutorials: Python, JavaScript, Git & More
Microsoft Security Blog
Microsoft Security Blog

DEV Community

Authentication Security Deep Dive: From Brute Force to Salted Hashing (With Java Examples) Why AI Systems Don’t Fail — They Drift Spilling beans for how i learn for exam😁"Reinforcement Learning Cheat Sheet" I Replaced Chrome with Safari for AI Browser Automation. Here's What Broke (and What Finally Worked) How Python Borrows Other People's Work The $40 Architecture: Processing 1 Billion API Requests with 99.99% Uptime Vibe Coding: A Workflow Guide (From Zero to SaaS) Most webhook security guides protect the wrong side. The scary part is delivery. Headless CMS for TanStack Start: Build a Blog with Cosmic EU Age Verification App "Hacked in 2 Minutes" — What Actually Happened Comfy Cloud’s delete function does not actually remove files Running AI Models on GPU Cloud Servers: A Beginner Guide Event-driven media intelligence with AWS Step Functions and Bedrock I scored 500 AI prompts across 8 quality dimensions — here's what broke How to Call Google Gemini API from Next.js (Free Tier, No Backend Needed) The Portal Protocol: Reclaiming Human Connection in the Age of AI How to Fix Your Team's Scattered Knowledge Problem With a Self-Hosted Forum Intro to tc Cloud Functors: A Graph-First Mental Model for the Modern Cloud Designing Multi-Tenant Backends With Both Ownership and Team Access I Built a Neumorphic CSS Library with 77+ Components — Here's What I Learned PostgreSQL Performance Optimization: Why Connection Pooling Is Critical at Scale Cómo construí un SaaS multi-rubro para gestionar expensas en Argentina con FastAPI + Vue 3 🚀 I Built an Ethical Hacking Scanner Tool – Open Source Project I Replaced /usage and /context in Claude Code With a Single Statusline A Pythonic Way to Handle Emails (IMAP/SMTP) with Auto-Discovery and AI-Ready Design I Collected 8.9 Million Polymarket Price Points — Here's What I Found About How Markets Really Move EcoTrack AI — Carbon Footprint Tracker & Dashboard Everyone's Using AI. No One Agrees How. 5 self-hosted ebook managers worth trying in 2026 Building Your First AI Agent with LangChain: From Chatbot to Autonomous Assistant
Stop Starting From Scratch: How Claude Projects Will Chan...
AbdlrahmanSa · 2026-05-03 · via DEV Community

What Is a Claude Project?

A Project is a persistent workspace inside Claude. Think of it as a dedicated room where you and Claude work together on a specific area of your life or work — and unlike a regular chat, it doesn't forget anything between sessions.

Each Project has three key parts:

Custom Instructions — a set of guidelines that shape how Claude behaves inside that project. Always. Every conversation.

Knowledge Files — documents, notes, and files you upload that Claude can reference in every conversation within the project.

Conversation History — all your past chats inside that project are preserved and accessible.

The context window for a Project is 200,000 tokens — roughly equivalent to a 500-page book. That's a lot of room to work with.

The mental model is simple: if a regular Claude chat is a sticky note, a Project is a notebook. One is for quick thoughts; the other is where real work happens.


Custom Instructions: The Feature Most People Skip

When you create a Project, you get a field called "Custom Instructions." Most people leave it blank or type something vague like "be helpful and professional." That's a waste.

Custom Instructions are how you give Claude a standing brief — the kind of context you'd give a new colleague on their first day. Except you only have to write it once.

What to include in strong custom instructions:

Your role and context. Tell Claude who you are. "I'm a freelance UX writer working with B2B SaaS clients" is more useful than nothing. Claude will calibrate its vocabulary, assumptions, and suggestions accordingly.

Tone and style preferences. Do you want formal or conversational? Bullet points or prose? Short and direct, or thorough? Spell it out. Claude will default to something reasonable, but "something reasonable" isn't the same as your preference.

What to always do. If you always want sources cited, say so. If you always want a summary at the top of long answers, say so. If you want Claude to push back when your ideas have holes, say so.

What to never do. "Don't use bullet points unless I ask" or "never suggest I use a different tool" — these kinds of guardrails save a lot of back-and-forth over time.

An example of weak vs. strong custom instructions:

Weak:

"You are a helpful assistant. Be professional and concise."

Strong:

"I'm a product manager at a fintech startup. I'm working on reducing user churn. I prefer direct answers — lead with the conclusion, then explain. When reviewing my writing, give honest feedback and don't soften criticism. Never use corporate buzzwords. If I ask for a framework or process, give me one I can actually use in a meeting, not a textbook definition."

The second version saves you from re-explaining yourself in every single conversation. Over weeks of use, that compounds into a noticeably smarter working relationship.


Knowledge Files: Give Claude Something to Work With

The second pillar of Projects is Knowledge Files — the documents and materials you upload to a Project for Claude to draw from.

This is what turns Claude from a general assistant into something that actually understands your work.

Here's the idea: any document you find yourself referencing repeatedly in conversations — a style guide, a brief, a set of notes, a template — should live in your Project's knowledge base instead of being re-pasted into every chat.

What to upload:

  • Style guides or voice docs — if you're using Claude for writing, upload the brand voice guidelines or your personal writing principles. Claude will match the tone without you having to describe it from scratch.
  • SOPs or process docs — upload your team's workflows or your own operating procedures. Ask Claude how to handle a situation and it'll answer based on how you actually work, not a generic answer.
  • Research or reference material — working on a long-form project? Upload your notes, interview transcripts, or source documents. Claude can summarize, cross-reference, and synthesize across all of it.
  • Past work samples — give Claude examples of writing, analyses, or outputs you're proud of. It gives Claude a calibration point for the quality and style you're aiming for.

One important tip: be selective. It's tempting to dump everything into a Project and let Claude sort it out. Don't. Focused, high-quality context produces better results than a pile of loosely related documents. If you're not sure a file belongs, it probably doesn't — yet.


Five Projects Worth Creating Today

If you're not sure where to start, here are five concrete Project setups that work well for most people.

Job Search Project
Upload your current CV, a list of target roles, any cover letter templates you've used, and notes on companies you're watching. Custom instruction: "I'm actively job hunting. Help me tailor applications, prep for interviews, and research companies. Be direct about what's weak in my materials."

Writing / Content Project
Upload your style guide, a few past pieces you're proud of, and any editorial brief you're working from. Custom instruction: "Match my voice. Don't clean up my sentence fragments unless they're unclear. Push back if a headline is weak."

Work Project
Upload your team's internal docs, a glossary of terms, your org chart if relevant, and any recurring templates. Custom instruction: "I work in [role] at [company type]. Use our internal terminology. When I ask for an email or document, match our internal style."

Learning Project
Upload lecture notes, reading summaries, or course materials. Custom instruction: "I'm trying to understand [topic] deeply. Use the Socratic method when I get things wrong. Give me examples from [domain I already know]."

Side Project / Startup
Upload your product spec, user research notes, competitor notes, and any pitch material. Custom instruction: "We're building [brief description]. I'm the [role]. Prioritize speed and practical thinking over polish."


Tips for Getting the Most Out of Projects

Write your custom instructions like you're onboarding someone. Not vague adjectives ("be concise") but actual behavioral guidance ("when I ask for a plan, give me numbered steps with time estimates").

One Project per context, not per task. Don't create a new Project every time you start a new piece of work. A "Writing" Project covers all your writing work. A "Job Search" Project covers the whole job hunt. Projects are contexts, not folders.

Update your instructions when something annoys you. If Claude keeps doing something you don't want, that's a signal — add a line to your custom instructions. The relationship improves as you refine it.

Keep knowledge files current. If you update a style guide or process doc, re-upload it. Stale context produces stale answers.

Use conversation history deliberately. Because your history is preserved, you can ask Claude to refer back to decisions made in earlier sessions. "Based on the brief we discussed last week, draft the intro" is a perfectly valid prompt.


The Bigger Shift

The underlying idea here is simple but easy to miss: Claude is only as useful as the context you give it.

A regular chat gives Claude nothing but your most recent message. Projects give Claude your role, your preferences, your documents, and your history — all at once, every time. That's not a small upgrade. It fundamentally changes what the tool can do for you.

The people getting the most out of Claude aren't necessarily using better prompts. They're working inside Projects, with well-written instructions and a curated knowledge base. They're treating the relationship like a collaboration, not a search engine.

So: what would your first Project be?

If you already have one set up, drop your custom instructions in the comments — I'd genuinely love to see what's working for other people.


If you found this useful, follow me here for more practical pieces on working with AI tools. I write for practitioners, not hype.