H≡ICOS

The Claude Code Pyramid of Success

2026-04-13T10:00:00+00:00

This is the pyramid that ships. Built block by block. Every layer below is a prerequisite for the one above. Shipped code sits at the apex.

“The cornerstones of Claude Code mastery are a sharp CLAUDE.md and a fast dev loop. One is the constitution. The other is the feedback engine.”

The Foundation

The five cornerstones everything else rests on.

CLAUDE.md — No substitute for the master prompt. Worthwhile sessions come from a sharp constitution and careful planning.

MCP Servers — Your tool belt. Mutual respect between Claude and every API, DB, and service you ship with.

Auto-Commit — Loyal to your future self. Every change committed. Never lose a line of work.

Hooks — Automation that cooperates at every level. Pre-commit, post-tool, session-start. Invisible teammates.

Fast Dev Loop — Your feedback loop must be sub-second. A fast loop makes you want to iterate.

Layer Two — Execution Discipline

Plan Mode — Resist the urge to ship unplanned code. Delicate balance between thinking and doing. Keep judgment.

Verification — Watch every tool output. Spot regressions fast. Never claim done without proof.

Parallel Tools — Batch independent reads, writes, searches. Act, don’t wait. Desire to ship faster.

Scope Discipline — Resist the refactor side-quest. Stay on the task. One template, one commit, one ship.

Layer Three — Scale

Context Hygiene — Prompt, Memory, History. Keep the context window clean. Compact often. Dissipation of tokens must be eliminated.

Prompt Craft — Knowing the fundamentals and executing them. Clear specs, exact file paths, cover every edge case.

Subagent Orchestration — Spawn parallel agents instead of doing it yourself. “The swarm comes first.” Speed is a feature.

Layer Four — Flow State

Checkpoints — Session restore points. Every state is a save file. Pick up exactly where you left off. Nothing lost, nothing rushed.

Trust — Let Claude run without hovering. Trust comes from a sharp CLAUDE.md, tight hooks, and clean git state.

The Apex — Agentic Flow

“When the bug is gnarly, Claude is at its best.” Be ready when the problem is hard. Real love of a hard ticket.

“Shipped code is peace of mind as a direct result of knowing you set Claude up to do its best work — and then shipped what only you could have shipped.”

— Manthan Patel

Pyramid originally created by Manthan Patel

The Claude Code Cheatsheet

2026-04-13T08:00:00+00:00

60+ Commands · 6 Environments · 1M Token Context · 20 Voice Languages

Installation

macOS: brew install claude-code

Windows: winget install Anthropic.ClaudeCode

npm: npm i -g @anthropic-ai/claude-code

Then navigate to your project and launch:

cd your-project && claude

CLI Flags

claude "prompt" — start with a prompt
claude -p — one-shot, non-interactive
claude -c — continue last session
claude --resume — resume by ID or name
claude -n "name" — name a new session
claude --worktree — run in isolated git worktree
claude --effort high — set max reasoning effort
claude --remote-control — remote terminal access
claude --debug — enable debug logging

CLAUDE.md — Agent’s Constitution

CLAUDE.md is the configuration file that defines Claude Code’s behavior per project.

Where to put it:

~/.claude/CLAUDE.md — global
./CLAUDE.md — project
./src/CLAUDE.md — folder overrides

What to put in it:

Build / lint / test commands
Code conventions and patterns
Commit message format
Architecture decisions and tech stack
What NOT to do (anti-patterns)

Pro tip: Use /init to scaffold a CLAUDE.md, then curate it manually.

6 Environments

Terminal CLI · VS Code Extension · JetBrains Plugin · Jetmins Plugin · Desktop App · Web · claude.ai/code · iOS App

Slash Commands

Session

/compact — compress context
/clear — clear session
/rewind — undo last action
/diff — show diff
/cost — show token cost
/usage — show usage stats
/context — show current context

Model

/model — switch model
/effort low|med|high — set reasoning effort
"ultrathink" keyword — maximum reasoning on next turn

Workflow

/plan — plan mode
/voice — voice interaction
/loop — scheduled cron-like tasks
/btw — background task while you work
/init — scaffold project

Advanced

/review — multi-agent code review pipeline
/simplify — 3-agent architecture review before PR
/batch — parallel changes across multiple isolated worktrees
/debug — systematic debugging workflow
/claude-api — API and SDK documentation, patterns, SDK examples on demand

Extensibility Stack

Skills (.claude/skills/) — AI-driven skill definitions, automatically invoked when Claude matches the task. Uses YAML frontmatter.

Subagents (.claude/agents/) — instances with isolated context windows. Explore (read-only) and Code (tool access) types.

Hooks (settings.json → hooks{}) — deterministic code that runs at specific events, e.g., PreToolUse, PostToolUse.

MCP Servers (claude mcp add) — connects Claude Code to external tools/data via Model Context Protocol (GitHub, Jira, Slack, Linear, Postgres, Notion…). Lazy-loads tools.

Keyboard Shortcuts

Shift+Tab — cycle permission modes (normal, auto-accept, plan mode)
Option+T — toggle extended thinking on/off
Option+P — open model picker
Ctrl+G — open current file/diff in external editor
Esc+Esc — open rewind/undo menu
Tab — toggle extended thinking for current turn only

Cost Tips

/effort low for quick iteration — saves thinking tokens
/model haiku for mechanical/repetitive tasks (formatting, renaming)
/compact proactively — larger contexts cost more per token
Batch API 50% off for non-urgent async workloads
Prompt Caching 90% savings for repetitive inputs

What’s New — March 2026

Opus 4.6 Default — Opus 4.6 is now the default model; effort defaults to medium.

1M Token Context — supports 1 million token context window for Max/Team/Enterprise plans.

Voice Mode — native voice interaction via /voice, push-to-talk, 20 languages.

/loop Command — scheduled recurring tasks (cron) within a session, e.g., for monitoring.

ultrathink — special keyword for maximum possible reasoning on the next turn only.

Remote Control — terminal access from claude.ai/code, iPhone, iPad, Mac — code stays local.

Computer Use — enhanced visual reasoning for desktop automation, research preview for Pro/Max.

HTTP Hooks — deterministic hooks can POST JSON to any URL, custom headers, env var interpolation.

Resources

Cheatsheet originally created by Brij Pandey · @brijpandeyji

Welcome to H≡ICOS

2026-04-13T07:00:00+00:00

A matter for the CEO

The hybrid organization

You know your monolithic organization with human intelligence only quite well. It’s the organization-type “MoHI”. But you never set up or get in touch with a mono-organization focused solely on artificial intelligence only (MoAI). Welcome to the reality of most enterprises worldwide. So the good news: You are not alone with this topic on your organization’s development roadmap.

Your absolutely new challenge, never had such a challenge before, is to transform your MoHI enterprise into a new type of organization, a hybrid-organization, adjusting your enterprise with AI capabilities, becoming an AInterprise.

You can stop reading here, as far as this hybrid-organization, to be an AInterprise, is not one of your targets for organization development.

Otherwise, and if you are convinced that the leadership role in such a change cannot be delegated to the technical level of the organization, you’ve come to the right place.

AI has no knowledge as Human Science

As a target, the hybrid organization sounds right, and most leadership teams are not short on information about AI. They are short on clardity about the MoAI and the actors, the Artificals.

And the root cause of the missing clarity is quite simple. It’s the AI undefined noise.

Let’s have a look at the Human way of knowledge engineering: It’s science by humans. From the processing of knowledge in this science, we can derive everything. It’s a stack with a fundamental, profound base. We take out noise, best guesses, and statistics to replace them with the groundwork of proof, well-defined, and eliminate guessing.

What is the science model of the Artificials? Right: It does not exist. Therefore, don’t search for anything reliable on the quality level of human science. Accept the noise of a mathematical statistical model, with the consequence that knowledge does not exist in this world. The best you can get is a best guess, and your hope that action on it does not end in a deep mess for your enterprise. This is to manage your expectations as well as to give you an idea where you start from.

-> Artificals are not stupid. They have just other habits and therefore another kind of identity than humans. Let’s learn how to deal with them.

Massive FOMO activities without strategical grounding

Too many AI options
Too much vendor noise
Too little internal alignment
No shared language between business and technology
No structured picture to sort through the different AI aspects
No practical experience matching the indidivual requirements

The result: decisions get delayed. Initiatives start without direction. Money gets spent on the wrong things.

The risk: Your well-defined Mono-Organization will be destroyed by a Hybrid-Organization approach “on demand”.

->FOMO is a bad guide. Discover the AI world with a map and a compass for your AInterprise targets.

The CEO’s transformer skill

A new CEO skill is required. The skill to transform a Mono-Organization into a Hybrid-Organization, which is not a simple merge of two types of Mono-Organizations. It’s a change to the Corporate Identity with all its consequences, and, explicitly, it’s not a question of technology.

Way of transformation

There is no best practice existing for your individual MoHI. You need to create your own MoAI blueprint to make the right decisions to help your target become a hybrid organization.

Our concept is to teach you to set up a Mono-Organization of type AI-Only and use the learnings as a blueprint. We define the Corporate Identity, the organizational structure, and we hire the Artificals - the employees of a MOAI. We have defined a business plan, and we are running this business in the MOAI. The deployment of the business to the market will be optional.

And after this step, we have an understanding of the standalone MOAI. And only after this step do we know a valid way to transform our Mono-Organization of Humans, in reality, into a Hybrid Organization, so-called AInterprise with Humans and Artificals.

The triple bar in our name—H≡ICOS—represents the 3-level structure and logic required to lead through the change by AI.

YOU CxO - navigate, decide, lead

What changes?

AI The Artificial Intelligence

HI The Human Intelligence

-> AI is a new world. For navigation, a Map and a compass are required.