You are here:Home»KB»Programming»My Claude Notes
Monday, 01 June 2026 09:24

My Claude Notes

Written by

my claude aI stuff

 

TL;DR

  • Claude is stateless = Every time you respond, the whole conversation is sent and re-read so your token usage gets exponentially more on every response.
  • Use /compact to reduce your conversation size
  • /context will show your current conversation token usage.
  • Use CLAUDE.md to reduce token burn.
  • Plan Mode is very useful
  • **need to chase this** Jobs that don't need an immediate response can run at a 50% discounted token rate, useful for conversions.
  • Pro or Enterprise plan is needed to use Claude in your IDE (i.e. Claude Code)

 

General AI

These will all apply to claude, but also other AI environments.

  • My Thoughts
    • AI is not always right, even though it says it is.
      • Claude will tell you when it cannot give an answer, which is an imporvement.
    • A new conversation is not always best.
      • Compact your conversation to prevent token burn however this can cause loss of important information.
      • New feature, new conversation
    • 15 to 20 messages is sweet spot.
      • AI can hallucinate after this
      • Circular thinking can also start occuring.
    • The better the prompt, the better the result.
      • Be concise even if you think what your are saying is stupid and it should know.
      • Don't be lazy
      • In `Exclusive File Locking` I had a lot of issues because Claude decided (on it's own) to use stripped Scintilla headers.
      • Don't let Claude cut corners, specify what libraries to use and be as precise as you can to prevent random thinking.
    • Claude is good at stuff everyone does because it is trained on that sort of data, but weird edge cases, it will need you to point stuff out (i.e. the Eureka moment) or it will get confused and give poor outputs.
  • Tutorials
    • How AI agents & Claude skills work (Clearly Explained) - YouTube | Greg Isenberg - I sit down with Ras Mic to break down how AI agents actually work and why most people are using them wrong. Ras Mic explains the mechanics of context windows, makes the case that agent md files are largely unnecessary, and shares his step-by-step methodology for building custom skills that make agents dramatically more productive. Whether you're coding with Claude Code or automating workflows with OpenClaw, this episode gives you the foundational knowledge to stop wasting tokens and start getting real results from your AI tools.
    • 18 Months of AI Coding Lessons in 18 Minutes (10 Tips) - YouTube | GritI Studio
      • AI coding tips that actually work. After 18 months of daily AI coding with Cursor, Claude Code, and other tools, I have distilled everything I learned into 10 tips you can learn in 15 minutes. These are battle-tested workflows I use every single day—no hype, just what actually works.
    • The models we use today are probably the worst models you will use for the rest of your life. In the past year alone, we have gone from AI that could barely autocomplete a function to agents that write entire features, create branches, and open pull requests while you drink your coffee. Hate it or embrace it—AI-assisted coding done the right way is a superpower!
    • Free AI Courses with Certificate | FreeAcademy.ai - Free AI courses with certificate. Learn ChatGPT, Claude & practical AI skills — 100% free, no coding required.
  • Swap Enter and Shift+Enter key behaviour / Disable Enter sending message
    • GitHub - deerhnt/chat-key-changer - "Chat-Key-Changer" is a Chrome extension that swaps Enter and Shift+Enter key behavior in AI chat services for improved usability.
      • Solving the Enter Key Frustration in AI Chat: "Chat-Key-Changer" Chrome Extension - DEV Community
        • Hello everyone! Have you ever experienced the frustration of accidentally sending an incomplete message while chatting with AI?
        • I regularly use ChatGPT, Claude, GitHub Copilot, and other AI services, but I often found myself accidentally hitting Enter while typing long messages, thinking I was adding a new line but ending up sending an incomplete message instead.
        • So I built a Chrome extension to solve this small but persistent annoyance - let me introduce it to you!
    • Chatgpt No Enter Chrome Extension - Amperly - Stop sending AI chat messages by accident. This Chrome extension prevents accidental sends in ChatGPT, Claude, Google Gemini, Grok, and Microsoft Copilot.
  • AI Comparisons
  • 3rd Party Libraries / Software
    • GitHub - pytorch/pytorch
      • Tensors and Dynamic neural networks in Python with strong GPU acceleration
      • Has extensive Claude integration
      • This is an AI neural net thing.

Official Claude Sites

Claude Plans

  • Claude Free Plan Limits 2026: What You Get, What You Don't, and When to Upgrade | FreeAcademy.ai
    • Everything about Claude free plan limits in 2026 — message caps, model access, which features are free vs paid, and exactly when upgrading to Pro or Max is worth it.
    • Limits are dynamic: during peak hours or heavy server load, you may hit caps sooner
  • Claude Pro vs Max vs Free (2026): Usage Limits Compared | FreeAcademy.ai
    • Claude Free vs Pro vs Max compared for 2026: usage limits, pricing, the real difference between Pro and Max, and which plan is actually worth it.
    • Claude Code is Anthropic's terminal-based AI coding agent. It reads your entire codebase, writes and edits files, runs commands, and handles multi-step development tasks autonomously. It requires a Pro subscription or higher.
  • Claude pricing and plan limits explained: Full guide to free, Pro, Team, and Max tiers | datastudios.org
    • Anthropic’s Claude platform offers a structured set of subscription plans designed for individuals, professionals, and teams.
    • Each tier differs in terms of model access, context window sizes, usage quotas, and pricing options, making it important to choose the right plan based on workload and expected performance.
    • Claude’s free plan offers Sonnet 4 with shorter session windows.The free plan provides access to Claude Sonnet 4, optimized for balanced performance and speed. Users can open multimodal

Claude Products/Software

  • Claude Code
    • Is an agentic coding tool that handles entire development workflows through natural language.
  • Claude Desktop
  • Claude Cowork
    • Claude Cowork: Claude Code power for knowledge work | Claude by Anthropic
      • Give Claude access to your local files and let it complete tasks autonomously.
      • Claude Cowork brings Claude Code's agentic capabilities to the desktop app for non-technical work.
      • Unlike Chat, Cowork lets Claude complete work on its own. Describe the outcome and cadence, and it takes action and keeps you informed. Come back to the result.
    • Claude Cowork | Anthropic’s agentic AI for knowledge work Anthropic
      • Claude Cowork is a system built by Anthropic that executes multi-step knowledge work on a user's behalf, including research synthesis, document preparation, and file management.
      • It is not a chat assistant.
      • Claude Cowork handles tasks autonomously. Give it a goal and Claude works on your computer, local files, and applications to return a finished deliverable. If it’s repetitive, messy, or just taking too long, assign it to Claude.
    • Claude Cowork, a research preview | Claude by Anthropic
      • Claude Code's power without the terminal.
      • Cowork brings Claude Code’s execution power to Claude Desktop for knowledge work, instead of just code. Delegate complex tasks that pull from multiple sources, need sustained effort, and produce real deliverables. The terminal happened to be where Claude Code lived. Now, that power runs directly on your desktop through a familiar interface.
  • Other
    • Claude MCP Clients - Explore various client applications implementing the Model Context Protocol (MCP)

Using Claude

Things I should do

  • Disable enter sending message
  • Goto Privacy Settings --> Toggle "Help improve Claude" to OFF
  • When developing, after each successful feature is added, make sure you do a commit as Claude has a habit of breaking stuff.
  • Don't let conversations get too long otherwise you might be burning tokens for reading the history when you don't need to.
  • On windows, tell Claude to use PowerShell and not BASH because windows does not have this shell.
    • **where should i put this? A local system config?
  • Disable Enter sending message

Tutorials

  • Full Claude Code Tutorials
    • Claude Code - Full Tutorial for Beginners - YouTube | Tech With Tim
      • This video will teach you Claude Code. I'll go over everything and assume no prior knowledge. I'll walk you through the setup and installation. I'll show you how to utilize the tool, the best practices, multiple features, and by the end of the video, you'll be comfortable using this tool to generate some pretty insane outputs and awesome coding projects.
    • Full Claude Code Tutorial for Non-Technical Beginners in 2026 (step-by-step) - YouTube | Futurepedia
      • Uses the Desktop App.
      • This video is a comprehensive, non-technical guide to Claude Code, showing you how to build an app step-by-step without any coding.
      • We cover six different projects, including a website, a game, an automation connecting external tools, and a mobile app with AI Vision.
      • This is all done through the Claude Code desktop app with no coding and at a beginner level starting from zero knowledge.
      • Utilize the power of AI agents and AI automation even if it's your first time using it. Discover the possibilities with Claude AI and a no code app builder to bring your ideas to life, and learn how to build an app with ease.
    • CLAUDE CODE FULL COURSE 4 HOURS: Build & Sell (2026) - YouTube | Nick Saraev
      • The end-to-end, definitive course on Claude Code for beginners! I'll take you through a full four-hour masterclass where I start by teaching you how to set up and install Claude Code in a tutorial, how to configure your IDE or integrated development environment (we'll use Antigravity), how to utilize your CLAUDE.md file as your project brain, how to build your first project in Antigravity using Claude Code in under 15 minutes, advanced Claude Code functionality including hooks, slash commands, and more.
      • I also teach you how to spin up multiple Claude Code instances and have them work on your behalf; how to parallelize work using sub-agents; how to use Git work trees to accomplish many hours of work in just a few minutes; how to conserve tokens and use context management to crush your coding and software projects; how to deploy things to the cloud using Modal and related services, and in general... how to be awesome at Claude Code! Hope you enjoy the tutorial.
    • Master 95% of Claude Code in 36 Mins (as a beginner) - YouTube | Nate Herk
      • In this video, I break down everything you need to know to use Claude Code effectively for building automations.
      • I cover the interface and what actually matters, the WAT framework we use to build every automation, how to communicate clearly with Claude Code and get it to ask the right questions during planning, the superpowers like MCP servers and Claude skills, and how to test, optimize, and deploy your workflows so they run automatically.
      • No fluff, I build a real workflow live so you can see exactly how it all comes together. By the end of this video, you'll understand Claude Code better than most people using it and have a clear framework for building your own workflows.
    • Claude Code: Rules, Memory, and Skills - YouTube | ExamPro
    • This video covers how to use claude.md, rules, memory, and skills to create better project context, more consistent behaviour, and stronger day-to-day workflows.
  • VSCode

Cheat Sheets

Commands

Operational Settings

Project / Persistent Memory

Files

  • Skills / SKILLS.md
    • What is SKILLS.md
    • What are Skills? Python scripts that AI knows how to run?
  • MEMORY.md
    • MEMORY.md acts as an index of the memory directory.
  • CLAUDE.md vs AGENTS.md
    • is calude.md and agents.md the same - Bing Search
      • Yes, CLAUDE.md and AGENTS.md are related but serve different purposes.
        • CLAUDE.md is the project context file for Claude Code, used by Claude Code to understand the project context and its rules.
        • AGENTS.md is a universal standard for AI coding agents, used by various tools like OpenAI Codex to provide context and instructions for AI agents.
        • Both files are plain markdown files located at the repository root, but they are read by different AI agents and follow different discovery rules.
        • In summary, while they are related, they are not the same and are used by different AI tools for different purposes.
      • Yes, CLAUDE.md and AGENTS.md are related but serve different purposes.
        • AGENTS.md is a cross-tool standard maintained by the Agentic AI Foundation, used by various AI coding tools to provide instructions and context to the AI agent.
        • CLAUDE.md is the native context file for Claude Code, which is used by Claude Code itself to define its behaviour and instructions within a specific project.
        • In summary, while both files are related to AI coding, AGENTS.md is a universal standard for AI tools, whereas CLAUDE.md is specific to Claude Code.
  • Writing CLAUDE.md
    • Updating CLAUDE.md
      • Self-improving CLAUDE.md files - Martin Alderson
        • A simple trick to keep your CLAUDE.md and AGENTS.md files updated using the agent's own chat logs - turning a tedious chore into a 30 second job.
      • Q: Are there any commands to update the CLAUDE.md , I already have one but need to know if Claude can check if the content is still relevant and update it if needed?
        • A: Claude does not have a built-in command to update a CLAUDE.md file, but can help you review and update it!
      • Commands / Instructions (pick one)
        • Read and review it 
          claude "Review my CLAUDE.md file and check if the content is still relevant and up to date"
        • From within an interactive session (claude to start), then: 
          /read CLAUDE.md

then follow up with:

Is this content still relevant? Suggest any updates.
        • To have Claude update it directly: 
          claude "Read CLAUDE.md, check for outdated content, and update it in place with any improvements"
        • Read, Review Changes, Apply changes 
          Please read CLAUDE.md and tell me if anything is outdated or missing

Review it's suggestions,

then ask Claude to apply the changes
  • Excluding sensitive files
    • To prevent Claude Code from accessing files containing sensitive information like API keys, secrets, and environment files, use the permissions.deny setting in your .claude/settings.json file: 
      {
  "permissions": {
    "deny": [
      "Read(./.env)",
      "Read(./.env.*)",
      "Read(./secrets/**)",
      "Read(./config/credentials.json)",
      "Read(./build)"
    ]
  }
}

Tokens and Limits

Misc

Troubleshooting

  • When I send a prompt I get no response
    • Q:
      • When I send a prompt I get no response, it returns empty immediately with my prompt still in the box?
      • It just started 3 minutes ago. It happens on both claude.ai windows browser chat and Android Claude app.
      • Using the desktop app or a different browser made no differences.
      • Not taking prompts.
    • A:
  • Pressing an arrow key cycles through the message history and wipes my current context window, how do i disable this?
  • If a response gets cut off because it hit a length limit before finishing, here's what actually works:
    • Just say "continue" (or "keep going") — Claude will pick up where it left off in the same conversation.
      • Alternatively "continue where you left off" works pretty well.
    • If it restarts from scratch instead of continuing cleanly, be explicit: "Continue from where you stopped, don't repeat what you already gave me."
    • For code or long documents specifically, ask it to continue from a specific marker, e.g. "continue from the save_config function onward" — that avoids ambiguity about where the cutoff was.
    • If the answer got cut off partway through something structured (a table, a numbered list, a file), it's often cleaner to ask for just the remaining section rather than "continue," since that guarantees no duplication.
    • If it's a genuinely huge task (e.g. a very long document or large piece of code), it can help to ask upfront for it in chunks rather than all at once — that avoids hitting the limit mid-flow in the first place.

Claude Plugins

  • Development
  • .claudeignore
    • .claudeignore is not a native Claude Code feature.
    • There is a third-party npm package called claude-ignore
    • GitHub - li-zhixin/claude-ignore A Claude Code PreToolUse hook that prevents Claude from reading files that match patterns in .claudeignore files, similar to how .gitignore works. - li-zhixin/claude-ignore
  • claude-mem
    • GitHub - thedotmack/claude-mem
      • Persistent Context Across Sessions for Every Agent – Captures everything your agent does during sessions, compresses it with AI, and injects relevant context back into future sessions. Works with Claude Code, OpenClaw, Codex, Gemini, Hermes, Copilot, OpenCode + More
    • claude-mem + cmem — AI agent memory, everywhere
      • AI agent memory, everywhere. claude-mem is open-source memory for Claude Code; cmem links it to any MCP client — Cursor, Codex, more — via one private link.
      • claude-mem takes notes while your agent works — then syncs them to the cloud behind one private MCP link. Pick up exactly where you left off. Every session, every machine.
    • Adding Persistent Memory to Claude Code with claude-mem — Plus a DIY Lightweight Alternative - DEV Community
      • The Problem: Claude Code Forgets Everything. Every time you start a new Claude Code session, the slate is wiped clean. Your coding style preferences, project architecture decisions, yesterday's debugging session — all gone.
      • claude-mem: A plugin that gives Claude Code a long-term memory. It automatically captures what you do during sessions and injects relevant context into future conversations.

Claude Legal

Terms and privacy stuff

  • TL;DR
    • Free, Pro and Max are classed as consumer accounts.
    • Enterprise accounts do not retain  inputs and outputs for the use of Anthropic.
    • All inputs and outputs are retained by default for Free accounts for the use of Anthropic, this cannot be turned off. (need to check for Europe)
    • All inputs and outputs are retained by default for Pro and Max accounts for the use of Anthropic, but this can be turned off.
      • Privacy Settings --> Toggle "Help improve Claude": OFF
      • This ensures your future conversations are not used to train Anthropic’s models, though it does not automatically remove data that was already processed.
      • Data can be retained for 5 years if you do not opt-out.
      • NB: when I created a Pro subscription, I was already opted out for helping Claude. I am in the UK, so this might also be a European thing.
    • Pro and Free allow Non-commercial use only.
      • You agree not to use our Services for any commercial or business purposes and we (and our Providers) have no liability to you for any loss of profit, loss of business, business interruption, or loss of business opportunity.
    • You cannot use claude to make products or algorithms that will complete against claude.
      • You may not access or use, or help another person to access or use, our Services in the following ways: :To develop any products or services that compete with our Services, including to develop or train any artificial intelligence or machine learning algorithms or models or resell the Services.
    • If you cancel during a period, your service will run to the end of the period. You do have to give 24+ hours before the renewal day, so do it 48 hours before your renewal date to ensure no re-billing.
  • Official
  • 3rd-Party

Other AI (ChatGPT, CoPilot, etc..)

 

Read 335 times Last modified on Sunday, 05 July 2026 13:52
Published in Programming
back to top