Code 231

I used Claude Code for 8 months without commands

I knew maybe 12 commands. There are 100.

Here are the 30 I actually use:

(Save this. Your terminal will thank you.)

☑️ 1. SESSIONS & PROJECT CONTROL

→ /init · generate your CLAUDE. md guide
→ /rewind · roll back to a checkpoint
→ /branch · try a direction, original kept

☑️ 2. CONTEXT & MEMORY

→ /context · see usage as a grid
→ /compact · summarize to free context
→ /memory · edit your CLAUDE. md files

☑️ 3. MODELS & REASONING

→ /plan · plan mode before touching code
→ /goal · work until the condition is met
→ ultrathink · triggers extended thinking

☑️ 4. FILES, TOOLS & PERMISSIONS

→ /permissions · manage allow and deny rules
→ /hooks · view lifecycle hooks on tools
→ /sandbox · toggle sandboxed ex*****on

☑️ 5. CLI LAUNCH COMMANDS

→ claude -c · continue the recent conversation
→ claude -p · print the answer and exit
→ claude doctor · check installation health

☑️ 6. CLI FLAGS & OUTPUT

→ --max-budget-usd · hard cap on spend
→ --output-format · text, json, or stream-json
→ --safe-mode · troubleshoot with customizations off

☑️ 7. MCP, PLUGINS & SKILLS

→ /mcp · manage MCP servers and OAuth
→ /skills · list skills and token cost
→ /plugin · install and manage plugins

☑️ 8. GIT, REVIEW & AUTOMATION

→ /security-review · scan the diff for vulnerabilities
→ /code-review · multi-agent review of your diff
→ /batch · parallel changes across git worktrees

☑️ 9. CONFIGURATION & DIAGNOSTICS

→ /usage · plan limits and session cost
→ /doctor · diagnose and fix setup issues
→ /statusline · build a custom status line

☑️ 10. SHORTCUTS & NAVIGATION

→ Ctrl+B · send running task to background
→ Ctrl+R · search your prompt history
→ Esc Esc · clear draft or open rewind

Honestly, /rewind alone has saved me more time than any prompt I have ever written.

ChatGPT vs Grok vs Gemini vs Claude vs Perplexity
People keep debating which AI tool is the “best.”
But the reality is:

Build apps, websites, AI videos, and complete businesses with just prompts using GPT-5.6, Fable 5, and more check it out

These tools are not competing to replace each other.
Each one is designed for a different purpose.
→ Some are better for creativity.
→ Some are stronger at research.
→ Some improve productivity.
→ Some help with real-time insights.
→ Some simplify complex work.
The real advantage today is not choosing one AI over another.
It’s understanding which tool fits which task.
Once you learn that, AI becomes more than just software.
It becomes part of your workflow and daily system.

ChatGPT
Great for everyday tasks like writing, brainstorming, coding, planning, learning, and solving problems quickly.

Grok
Useful for trending topics, live discussions, internet culture, and real-time social conversations.

Gemini
Works best for users heavily connected with Google tools like Gmail, Docs, Sheets, Slides, and Drive.
Claude

Strong at handling long documents, reports, research papers, and detailed explanations with better readability.
Perplexity

Excellent for research, fact-checking, source-based answers, and finding reliable information fast.

The biggest change happening with AI is this:

We are shifting from:

“One AI tool can do everything.”

To:
“Different AI tools do different things exceptionally well.”

The advantage today is no longer just using AI.

The real edge comes from knowing:
Which AI to use,
For what task,
And at what time.

Most companies focus on reducing AI costs.

Leading companies focus on maximizing AI value per dollar spent.

The difference is governance, prioritization, and financial discipline.

𝐇𝐞𝐫𝐞 𝐚𝐫𝐞 8 𝐀𝐈 𝐜𝐨𝐬𝐭 𝐨𝐩𝐭𝐢𝐦𝐢𝐳𝐚𝐭𝐢𝐨𝐧 𝐬𝐭𝐫𝐚𝐭𝐞𝐠𝐢𝐞𝐬 𝐭𝐡𝐚𝐭 𝐚𝐜𝐭𝐮𝐚𝐥𝐥𝐲 𝐰𝐨𝐫𝐤:

𝟏. 𝐀𝐥𝐢𝐠𝐧 𝐀𝐈 𝐰𝐢𝐭𝐡 𝐁𝐮𝐬𝐢𝐧𝐞𝐬𝐬 𝐎𝐛𝐣𝐞𝐜𝐭𝐢𝐯𝐞𝐬:
Every AI initiative should solve a measurable business problem, not just showcase technology.

𝟐. 𝐏𝐫𝐢𝐨𝐫𝐢𝐭𝐢𝐳𝐞 𝐇𝐢𝐠𝐡-𝐑𝐎𝐈 𝐔𝐬𝐞 𝐂𝐚𝐬𝐞𝐬:
Focus on projects with clear business impact, scalability, and faster time-to-value.

𝟑. 𝐀𝐝𝐨𝐩𝐭 𝐏𝐨𝐫𝐭𝐟𝐨𝐥𝐢𝐨-𝐁𝐚𝐬𝐞𝐝 𝐅𝐮𝐧𝐝𝐢𝐧𝐠:
Treat AI investments as a strategic portfolio rather than isolated experiments.

𝟒. 𝐁𝐮𝐢𝐥𝐝 𝐂𝐨𝐬𝐭 𝐓𝐫𝐚𝐧𝐬𝐩𝐚𝐫𝐞𝐧𝐜𝐲:
Track spending across teams, models, vendors, and cloud resources.

𝟓. 𝐁𝐮𝐝𝐠𝐞𝐭 𝐟𝐨𝐫 𝐭𝐡𝐞 𝐅𝐮𝐥𝐥 𝐋𝐢𝐟𝐞𝐜𝐲𝐜𝐥𝐞:
Deployment is only the beginning. Maintenance and governance drive long-term costs.

𝟔. 𝐑𝐞𝐝𝐮𝐜𝐞 𝐓𝐨𝐨𝐥 𝐒𝐩𝐫𝐚𝐰𝐥:
Too many AI platforms create unnecessary complexity and waste.

𝟕. 𝐎𝐩𝐭𝐢𝐦𝐢𝐳𝐞 𝐂𝐥𝐨𝐮𝐝 & 𝐂𝐨𝐦𝐩𝐮𝐭𝐞:
Monitor infrastructure continuously to prevent runaway spending.

𝟖. 𝐄𝐦𝐛𝐞𝐝 𝐆𝐨𝐯𝐞𝐫𝐧𝐚𝐧𝐜𝐞 𝐢𝐧𝐭𝐨 𝐀𝐈 𝐒𝐩𝐞𝐧𝐝𝐢𝐧𝐠:
Financial controls should exist throughout the AI lifecycle, not after deployment.

Organizations that win with AI are not the ones spending the most.

They are the ones connecting every AI dollar to measurable business outcomes and sustainable ROI.

P.S. Which AI cost optimization strategy do you think companies overlook most often?

-------------------------------------------------

The biggest advantage with Claude isn't having more prompts.
It's knowing how to give AI better instructions.

Most people use AI by typing a quick request.

"Write an email."
"Summarize this."
"Create a post."

Then they wonder why the output feels average.

The difference between basic and high quality AI results often comes down to the workflow behind the prompt.

The best AI users aren't just collecting prompts.
They're building repeatable systems.

Here are a few prompts that can save hours every week:

Turn Claude into your writing assistant.

Use it to draft emails, refine messages, adjust tone, and create content faster.

A simple instruction with clear context can turn a rough idea into a polished piece of work.

Repurpose content at scale.
One idea can become multiple assets.

A blog post can become LinkedIn posts, newsletters, social media captions, and video scripts.

Claude helps multiply your content without starting from zero.

Create systems, not just answers.

Use prompts to build SOPs, workflows, proposals, and repeatable processes.

AI becomes more valuable when it helps organize how work gets done.

Use AI for thinking, not only writing.

Ask Claude to analyze competitors, find patterns, compare options, and identify opportunities.

The strongest results come when AI supports decision making.

Automate repetitive tasks.
Small tasks repeated every day add up.

Finding ways to automate those workflows can free more time for higher value work.

The infographic below includes 10 Claude prompts that can help you write faster, plan better, research smarter, and improve daily workflows.

Better prompts don't just save time.
They create better systems.

Save this for the next time you want Claude to do more than just answer questions.

Your notes are dying in folders you never open again.

Most people capture everything and retrieve nothing.
They screenshot, highlight, save to Notion.
Then never look at it again.
That's not a second brain.
That's a digital junk drawer.

After 16 years in marketing,
I've taken thousands of notes across campaigns, client calls, research.
For most of that time,
They went to die in folders.
Then I started running marketing at a global beauty brand and realised the problem wasn't capturing less.
It was retrieving badly.
AI fixed that almost overnight.

Here's the reframe.
A second brain isn't a filing cabinet.
It's a system that finds things for you, connects them, and turns them into output.
Search stops being the bottleneck.
Thinking becomes the whole job again.

The 7 layers that actually make this work:

→ Capture
Save the raw idea instantly, organise later

→ Organise
Four folders only: Projects, Areas, Resources, Archive

→ Understand
Ask AI to summarise, simplify, find the gaps

→ Connect
Let AI surface related notes and contradictions you missed

→ Retrieve
Stop searching folders, just ask "what did I learn about pricing?"

→ Create
Turn notes into posts, reports, SOPs, decks

→ Improve
Every week, ask AI what patterns it's noticing

I tested this exact structure with my own client notes last quarter.
Cut my prep time for a single campaign brief from two hours to twenty minutes.
Not because I typed faster.
Because I stopped hunting for the note I knew existed somewhere.

A few prompts worth stealing straight away:
✔ "Summarise this into five key points."
✔ "What related ideas already exist?"
✔ "Turn these notes into a LinkedIn post"
✔ "What did I learn this week?"

Small prompts.
Compounding return.
Month one is just a capture habit.
By month twelve,
AI genuinely thinks alongside you.

I've covered the full breakdown in the infographic.
Worth keeping.

If you're building something similar,
You'll probably enjoy what else I post here.
Send this to the person on your team who still has forty browser tabs open.
Where do you actually store your notes right now?
Curious what's working for people.

I'm building an Agentic Coding course, and the waitlist is open now. I don't like courses with fixed lessons. I always start knowing some of the material, and I want to skip straight to the parts I don't know. A fixed playlist can't do that.

So this course will work differently.

There's no lesson one. Your first session is a check-in: about 15 minutes of questions, asked and answered in your terminal inside of Claude Code itself. It finds what you already know and skips it. If it finds no gaps, you are Done.

After that, a knowledge graph tracks what you know. Your agent will file evidence after each lesson and exercise, and your next session is built around what you don't know yet.

Every lesson is a Claude Code session in your own terminal: a prepared project, a task, and Claude working through it with you. Run A/B tests to tweak configuration settings, explore how hooks and subagents change the behavior of your agent, and compare token costs between models.

The course is a subscription. Claude Code keeps shipping new features, and when it does, you get new lessons and updates to existing ones included. Lessons for Codex, Pi agent, and OpenCode land in the same subscription as they're released.

The course is still in development. Joining the waitlist now locks your early-access price and your place in line.

Seats 1 to 5 are free.

The next ones have a decreasing discount.

Link in the comments.

Not every AI assistant is an AI agent.

The difference comes down to how much responsibility the system can take.

A chatbot is built for conversation. It answers questions, handles FAQs, captures leads, and supports Q&A.

A copilot goes one step further. It works alongside a person, understands context, accesses approved tools or data, and suggests the next action. The human reviews the output.

An AI agent is designed for ex*****on. It receives a goal, creates a plan, uses tools, calls APIs, tracks progress, and completes a task with less guidance.

Agentic AI operates at a broader level. It coordinates specialist agents, systems, and feedback loops to achieve an end-to-end business outcome.

Think of the progression like this:

↳ Chatbot: talks with users
↳ Copilot: assists users
↳ AI agent: acts for users
↳ Agentic AI: coordinates multiple actors

The right choice depends on the problem.

Use a chatbot for support and internal Q&A.
Use a copilot for drafting, meeting preparation, and sales assistance.
Use an AI agent for research, ticket triage, onboarding, and data updates.
Use agentic AI for long-running workflows, operations orchestration, and cross-team automation.

Greater autonomy creates greater responsibility.

Security, governance, observability, human oversight, and ROI measurement must be built in from day one.

The goal is not to deploy the most advanced AI system.

It is to choose the right level of autonomy for the workflow, risk, and business outcome.

Claude Code and Codex can solve similar problems but they approach the work differently.

One operates like a context-aware coding partner inside your development workflow.

The other behaves more like a delegated engineering agent executing tasks inside an isolated environment.

Here is the workflow:

→ 𝗖𝗹𝗮𝘂𝗱𝗲 𝗖𝗼𝗱𝗲

It begins by loading project instructions and exploring the repository before creating a plan.

The agentic tool loop can then work across:

• Files
• Shell commands
• Git
• MCP integrations
• Subagents

It edits the repository, runs tests, checks linting, builds the project, and prepares the changes for human review and a commit or pull request.

Claude Code is best understood as a context-aware coding partner working directly across the developer’s project, tools, and active workflow.

→ 𝗖𝗵𝗮𝘁𝗚𝗣𝗧 𝗖𝗼𝗱𝗲𝘅

It loads agent instructions and task context, then prepares an isolated sandbox for ex*****on.

Its agentic loop can work with:

• Files
• Shell commands
• Git
• MCP connections
• Plugins
• Web access

It edits a worktree, runs tests, performs lint and build checks, and returns a diff, commit, or pull request for human review.

Codex is best understood as a delegated software-engineering agent that can execute complex tasks across isolated local or cloud environments.

The difference is not simply which model writes better code.

It is how each system gathers context, executes work, collaborates with the developer, and isolates tasks.

Which approach better fits your development workflow: interactive collaboration or delegated ex*****on?

If you’re building your brand as a founder,
Remember: people trust people - not tools.

And LinkedIn is where that trust starts.

Let’s build yours.

Follow Denis Panjuta for more such Insights!!

The same prompt can fail in one AI and land perfectly in another.

__________

Most teams still treat prompting like one universal skill.
That is where quality starts leaking.

ChatGPT, Claude, Perplexity, and Gemini do not behave the same.
They reward different instructions.
They are built around different jobs.

→ ChatGPT is the instructor.
Use it when you need a plan, draft, outline, or first pass.
Give it a role.
Name the task.
Ask for step-by-step structure.
Then ask for formats like a table, checklist, or options list.

→ Claude is the careful teammate.
Use it when the work needs judgment, writing, coding, or deep review.
Tell Claude who you are.
Explain how you work.
Hand it real files and real constraints.
For harder tasks, use stronger models and extended thinking.

→ Perplexity is the research analyst.
Use it when trust matters more than speed.
Ask one specific question.
Add names, dates, companies, or places.
Tell it to show sources.
Then use follow-ups to dig deeper instead of starting over.

→ Gemini is the project planner.
Use it when the answer needs to connect to Google work.
Ask for the output format first.
Name the audience.
Request tables, day-by-day plans, or project breakdowns.
Pull from Gmail, Drive, and Calendar when context matters.

The mistake is not “bad prompting.”
The mistake is using one prompt style for four different systems.

A leader should know the default job of each tool.
ChatGPT helps shape the work.
Claude improves and executes the work.
Perplexity checks the work against sources.
Gemini turns scattered Google context into plans.

That is the useful mental model.
Not prompt hacks.
Not magic words.
Not longer instructions.

Pick the AI based on the job.
Then prompt it like the job is already clear.

If your team uses all four tools, audit the handoff.
Where does planning start?
Where does research get checked?
Where does writing get improved?
Where does Google context enter the workflow?

The prompt is only half the system.
The tool choice does the other half.

Which AI tool is your team using for the wrong job right now?

AI tools are evolving faster than ever.

Many tools we used daily in 2025
are already being replaced by smarter AI workflows.

Here’s a simple comparison 👇

Then vs Now.

Search
2025 → Google
2026 → Gemini

Spreadsheets
2025 → Excel
2026 → Claude in Excel

Browser
2025 → Chrome
2026 → Perplexity Comet

Image Editing
2025 → Photoshop
2026 → Nano Banana

Video Editing
2025 → Premiere Pro
2026 → Kling

Writing
2025 → ChatGPT
2026 → Claude

Note-taking
2025 → Fireflies
2026 → Granola

Presentations
2025 → PowerPoint
2026 → Gamma

Design
2025 → Canva
2026 → Vislo

Email
2025 → Gmail
2026 → Google Workspace Studio

Research
2025 → Google
2026 → Perplexity

Image Generation
2025 → Midjourney
2026 → Imagen

The real shift isn’t just tools.

It’s how AI is replacing complex workflows with simple prompts.

Which AI tool from this list do you use the most right now? 👇