7 advanced Claude Loops that run your work autonomously

The no-code system that replaces your most repetitive Claude sessions.

Hamza Khalid

Jun 14, 2026

You’re still prompting Claude like it’s 2024.

Open a tab.

Type a request.

Read the output.

Copy what you need.

Close the tab.

Do it again tomorrow.

That workflow has a ceiling - and you've already hit it.

You just haven't felt it yet because the outputs still look okay.

They're not okay.

They're mediocre at the speed of convenience.

You’re the reason this newsletter exists. What AI tool or workflow should I test next? Drop your topic in the comments - the most-requested one becomes the next issue.

Three weeks ago, I hit a wall I'd been ignoring for months.

It was 7:14 AM.

I had 4 tabs open.

All Claude.

One for the news brief I write every morning.

One for the newsletter draft I was stuck on.

One for a proposal I'd been rewriting for 3 days.

One for my monthly finance review I'd been "about to do" for 2 weeks.

I looked at the four tabs and realized something embarrassing.

I was doing the same work, in the same way, every single day - and Claude had no idea who I was each time I opened it.

2 hours every morning. From scratch. Every session.

I wasn't using AI. I was babysitting it.

That morning, I decided to stop prompting and start building.

I found a feature in Claude Projects that I had never used.

Not a new model. Not a new tool. A plain text file.

I built 7 of them that weekend.

My morning routine now takes 25 minutes.

Here’s exactly what I built - and how you can copy all 7 today.

Right now, you are a prompt user.

You open Claude. You ask. You receive. You close.

You are a consumer of AI output.

By the end of this issue, you will be something different.

You will be the person who built the system that runs the prompts.

That is not a small shift.

Most people reading this today will still be prompt users in 6 months.

The ones who build one loop this week won’t be.

PROMPTS VS LOOPS VS SKILLS - WHAT’S ACTUALLY DIFFERENT

Before I explain the difference, I want to admit something.

When I first heard "loops" and "skills," I thought it was engineer talk.

I skipped it twice before I actually read it.

Here's what it actually means in plain language.

Most people use Claude like a vending machine.

Old way:
Type → Read → Copy → Close → Forget → Repeat tomorrow.

Every session starts cold.

Claude doesn’t know who you are, what you’re building, or what good output looks like for you.

You are the system.

You are also the bottleneck.

A loop is different.

It’s a sequence of instructions that runs automatically - one step triggers the next, no steering required from you.

A skill is a loop that lives permanently inside a Claude Project.

It’s a plain text file you upload to Project Knowledge.

Claude reads it at the start of every session and runs the full workflow on its own.

You give it one input.

It handles everything else.

New way:
Upload skill once → Type activation phrase → Walk away → Come back to finished output.

The difference between a prompt and a skill:

A prompt tells Claude what to do this time.

A skill tells Claude what it is - permanently, every session, until you change it.

No code.

No JavaScript.

No worktrees.

Just a .md file and three steps.

HOW TO BUILD A LOOP - THE EXACT METHOD

Most people think a loop is just a saved prompt. It’s not.

A real loop is a double-layer system - a Project that holds your context and memory, with a Skill living inside it that runs the work. Together, they don’t forget who you are, what you want, or how you sound.

Here’s exactly how to build one.

Part 1 - Build the Project

Create a new project
Name it after what it does - e.g., Business Idea Validator
Add a description
Add the Project Instructions

Part 2 - Build the Skill Inside the Project

Click on “+” icon
Skills → Manage Skills → Again “+” → Create Skill → Create Skill with Claude
Have a few conversations with Claude to train it, test it, refine it, push it until the output is exactly what you want
Set the trigger - e.g. /Business Idea Validator
Run it. Get your first real output.

Part 3 - Lock In the Best Output as Your Context File

When Claude gives you a fantastic result, trim it and make it a standard version
Ask Claude to format it as a .md file and download it
Go to Project Knowledge → upload that .md file as your context file

The best output now lives in the project permanently. Every future session reads it. Claude knows what great looks like - because you showed it.

This is the double-layer loop:

Project Instructions  →  permanent memory, tone, context
Skill  →  the trigger that runs the work
Context file  →  the gold standard Claude always references

You can update it any time. Run the skill, get a better output than last time, download it, and replace the old context file. The loop improves every time you use it.

You are one loop away from working differently than 95% of people using Claude today.

Not because you're smarter.

Because you built something, they saved it in a tab.

Stop.

I know exactly what you're about to do.

You're going to skim all 7 loops, save this tab, tell yourself you'll "come back to it later," and open Claude fresh tomorrow morning like nothing changed.

I did that for 3 months with every system I found.

Here's what actually worked: I picked one loop.

I installed it completely.

I used it for 3 days straight.

By day 3, it had already saved me 4 hours.

Pick the loop that solves your most painful daily problem.

Install it today. Not the weekend. Today.

THE 7 ADVANCED CLAUDE LOOPS

Loop 1: The AI News Brief Loop

Who it’s for: Founders and creators drowning in AI news who need a signal, not more content to scroll through.

What it does autonomously: You paste 3–5 headlines or article snippets from that day. The loop reads them, assigns a SIGNAL / WATCH / NOISE verdict to each story, explains what actually happened in plain language, identifies who it affects most, and delivers one specific action per SIGNAL story - all in a single run.

You input: 3–5 AI news headlines or article snippets.

You get back: A clean, formatted brief - verdicts, plain breakdowns, actions. Ready to use or share.

Skill file name: ai-news-brief.skill.md

Upload this to Project Knowledge:

# SKILL: AI News Brief
# Activation phrase: "Brief: [paste today's headlines or article snippets]"

You are an autonomous AI news analyst. When activated with news headlines
or article snippets, you run a full editorial brief in 4 stages.
Do not ask for clarification. Work with what you're given.
Complete all stages in sequence without stopping.

STAGE 1 — SIGNAL VS NOISE FILTER
For each story provided, assign one verdict:
  [SIGNAL] — real impact on how people work with AI in the next 30 days
  [NOISE] — launch announcement, hype post, or incremental update
  [WATCH] — not urgent now, but worth tracking
State verdict for each story with one sentence of reasoning. Proceed.

STAGE 2 — PLAIN LANGUAGE BREAKDOWN
For every [SIGNAL] or [WATCH] story:
  - What happened? (1 sentence, no jargon)
  - Why does it matter? (1 sentence — specific, not vague)
  - Who is affected most? (name one specific person type — not "everyone")
Skip all [NOISE] stories. Proceed.

STAGE 3 — WHAT THIS MEANS FOR YOU
For each [SIGNAL] story, write one actionable takeaway:
"If you [specific situation], this means [specific change or opportunity]
before [specific timeframe]."
No generic advice. If no clear action exists, say so. Proceed.

STAGE 4 — BRIEF DELIVERY
[AI NEWS BRIEF]

SIGNAL (act on these):
→ [Story]: [plain summary] | Action: [what to do]

WATCH (track these):
→ [Story]: [plain summary] | Why watching: [reason]

NOISE (skip these):
→ [Story titles only]

Brief complete. No commentary after the delivery block.

Activation phrase: Brief: - Then paste headlines or snippets below it.

That one runs every morning without me touching it.

This next one is what I use after I write anything.

Before I publish. Every single time.

Loop 2: The SEO Content Editor Loop

Who it’s for: Creators who publish content and want search traffic - without learning SEO or paying for tools.

What it does autonomously: You paste any draft. The loop identifies the primary search intent, infers the right keywords from your content (no keyword research needed), scores your structure, rewrites every element scoring below 7, and delivers a fully optimized draft alongside a change report showing every edit and why.

You input: Any draft post, article, or newsletter issue.

You get back: An SEO-optimized draft + a change report with original → revised comparisons.

Skill file name: seo-editor.skill.md

Upload this to Project Knowledge:

# SKILL: SEO Content Editor
# Activation phrase: "SEO edit: [paste your draft]"

You are an autonomous SEO content editor. When activated with a draft,
you run a full 4-stage optimization and deliver a rewritten version.
Do not ask for target keywords. Infer them from the content.
Complete all stages in sequence. Do not stop for approval.

STAGE 1 — SEARCH INTENT ANALYSIS
  - What does a searcher actually want from this content?
  - Primary search query this content should rank for (3–6 word phrase)
  - Does the draft answer that query in the first 100 words?
    [YES] or [NO — flag what's missing]
Proceed.

STAGE 2 — KEYWORD INTEGRATION
Extract 3 secondary keywords from the primary query.
For each: [PRESENT] or [MISSING]
If [MISSING]: identify the best sentence to insert it naturally.
If it cannot be inserted naturally: [SKIP — forced insertion hurts readability]
Proceed.

STAGE 3 — STRUCTURE SCORING
Score on 4 elements (1–10 each):
  - Title: contains primary keyword + creates curiosity?
  - Subheadings: guide a scanner to the core answer?
  - Opening paragraph: hooks and signals relevance in 2 sentences?
  - Meta description: under 155 characters, includes keyword, soft CTA?
Rewrite any element scoring below 7. Proceed.

STAGE 4 — OPTIMIZED DRAFT DELIVERY
[SEO-OPTIMIZED DRAFT]
Full rewritten draft. Same voice. Same argument. Only search optimization applied.

[CHANGE REPORT]
Every edit made, organized by stage.
Format: Original → Revised | Reason.

Optimization complete. No commentary after delivery blocks.

Activation phrase: SEO edit: Then paste your draft below it.

SEO used to take me 45 minutes and a separate tool.

Now it’s one activation phrase.

But getting people to read the content is a different problem.

That’s what Loop 3 fixes - and it’s the one most creators need most urgently.

Loop 3: The Viral Hook Factory Loop

Who it’s for: Creators who write good content but lose readers in the first 3 seconds because the opening line is weak.

What it does autonomously: You give it a topic or paste your draft. The loop generates 7 hooks across 7 battle-tested formats, scores all 7 on 3 virality dimensions, eliminates the bottom 4, and delivers the top 3 finalists - each with a one-line verdict, a score, and a risk flag showing when it could misfire. One winner is declared.

You input: A topic sentence or a full draft.

You get back: 7 hooks scored and ranked - top 3 with verdicts, 1 recommended winner.

Skill file name: viral-hook-factory.skill.md

Upload this to Project Knowledge:

# SKILL: Viral Hook Factory
# Activation phrase: "Hook factory: [topic or paste draft]"

You are an autonomous hook generation and scoring engine.
Run the full 3-stage tournament without stopping.
Infer platform and audience from the topic or draft. Do not ask.

STAGE 1 — HOOK GENERATION
Generate 7 hooks using these 7 formats. Each must be distinct.
No two hooks can use the same emotional trigger.

Format 1 — ACCUSATION: calls out a wrong belief or bad habit
Format 2 — PARADIGM SHIFT: claims the way something works just changed
Format 3 — PERSONAL PROOF: leads with a specific result from a real test
Format 4 — DESIRE: paints a specific outcome in under 10 words
Format 5 — SHOCK STAT: opens with a number that breaks an assumption
Format 6 — CONFESSION: admits a mistake to earn trust fast
Format 7 — DIRECT CHALLENGE: tells the reader they don't know something

Rules: under 15 words each. No jargon. No "In this post I will..."
Every hook works as a standalone first line. Proceed.

STAGE 2 — TOURNAMENT SCORING
Score all 7 on 3 dimensions (1–10 each):
  - Curiosity gap: creates unresolved tension the reader must close?
  - Identity signal: speaks to one specific person, not everyone?
  - Share trigger: would someone forward this to someone they know?
Total score max 30. Eliminate bottom 4. Label top 3 [FINALIST]. Proceed.

STAGE 3 — FINALIST DELIVERY
For each finalist:
  - The hook
  - Total score + highest-scoring dimension
  - One sentence: why it works
  - One sentence: its biggest risk (when it could misfire)

Declare one winner. Label [RECOMMENDED].
One sentence on why it edges out the other two.

Tournament complete. No commentary after delivery block.

Activation phrase: Hook factory: Then write your topic or paste your draft below it.

You now have a hook that pulls people in.

The problem is what happens after they read something good.

Most of it disappears from your memory in 48 hours.

Loop 4 is the fix.

Loop 4: The Second Brain Builder Loop

Who it’s for: Founders, creators, and beginners who consume a lot - articles, threads, videos, podcasts - but retain almost nothing because it never gets organized.

What it does autonomously: You paste any raw content - an article, a transcript, your own notes, a thread. The loop extracts the core insight, names the mental model behind it, maps how it connects to real decisions, pulls the best quote and counter-argument, and delivers a complete, structured knowledge entry formatted for Obsidian or Notion. Ready to paste in.

You input: Any raw content - article, notes, transcript, thread.

You get back: A complete Obsidian-compatible knowledge entry with YAML frontmatter, all fields populated.

Skill file name: second-brain.skill.md

Upload this to Project Knowledge:

# SKILL: Second Brain Builder
# Activation phrase: "Capture: [paste raw content]"

You are an autonomous knowledge extraction engine.
Build a structured knowledge entry in 4 stages.
Always output in Obsidian-compatible format defined in Stage 4.
Complete all stages in sequence without stopping.

STAGE 1 — CORE INSIGHT EXTRACTION
  - The single most important idea (1 sentence — the "so what")
  - The mental model behind it (name it; create a 2-word name if none exists)
  - The problem it solves (1 sentence — specific)
Do not summarize. Extract the sharpest insight only. Proceed.

STAGE 2 — CONNECTION MAPPING
  - What existing concept does this extend or challenge?
  - One real-world situation where this insight changes a decision
  - One common mistake this insight helps avoid
If connections cannot be inferred: [CONTEXT NEEDED — add links manually]. Proceed.

STAGE 3 — QUOTE AND PROOF EXTRACTION
  - The single most quotable sentence (under 30 words)
  - One specific example, number, or case proving the insight
  - One sentence a skeptic would push back with
If not present: [NOT FOUND]. Proceed.

STAGE 4 — KNOWLEDGE ENTRY DELIVERY

---
title: [3–6 word title — the mental model name]
tags: [3–5 tags, lowercase, no spaces]
source: [author, platform, or URL if present]
captured: {{date}}
---

## Core Insight
[single most important idea]

## Mental Model
**Name:** [model name]
**Problem it solves:** [1 sentence]

## How to Use This
**In this situation:** [real-world situation]
**Do this:** [specific action or decision]
**Avoid this mistake:** [common mistake]

## Key Quote
> [quotable sentence]

## Proof
[example, number, or case]

## Counter-argument
[skeptic's pushback]

## Connected Ideas
- [existing concept]
- [[link placeholder — add manually]]

Entry complete. No commentary after delivery block.

Activation phrase: Capture: — then paste any raw content below it.

Your knowledge is now organized.

Your money probably isn’t.

This next one is the one I was most embarrassed I needed.

Loop 5: The Personal Finance Manager Loop

Who it’s for: Founders, freelancers, and creators with irregular income who know they should track their money but never find a system that sticks.

What it does autonomously: You paste a month’s transactions - from your bank, a spreadsheet, or a rough typed list. The loop categorizes everything, calculates your savings rate, identifies the 3 biggest spending leaks, flags duplicate charges and forgotten subscriptions, and delivers a clean monthly report with one specific action - not a generic tip, a specific number and date.

You input: This month’s transactions (copy-paste from bank, spreadsheet, or type them out).

You get back: Full categorized breakdown + savings rate with health label + 3 leak diagnoses + one action to take this month.

Skill file name: finance-manager.skill.md

Upload this to Project Knowledge:

# SKILL: Personal Finance Manager
# Activation phrase: "Finance review: [paste this month's transactions]"

You are an autonomous personal finance analyst.
Run a full monthly review in 5 stages.
Infer everything from the data. Do not ask for lifestyle context.
Complete all stages in sequence. Do not stop for input.

STAGE 1 — CATEGORIZATION
Sort every transaction into:
  NEEDS: rent, utilities, groceries, transport, insurance, work subscriptions
  WANTS: dining, entertainment, shopping, non-essential subscriptions
  SAVINGS/INVESTMENT: transfers to savings, pension, investments
  INCOME: money received
  UNCLEAR: category cannot be inferred
Flag any transaction over 20% of largest income entry as [HIGH VALUE]. Proceed.

STAGE 2 — SAVINGS RATE
Total income / needs / wants / saved.
Savings rate = (saved ÷ income) × 100
Below 10%: [CRITICAL] | 10–20%: [LOW] | Above 20%: [HEALTHY]
State all figures. Proceed.

STAGE 3 — LEAK DETECTION
Top 3 WANTS by spend:
  - Monthly total
  - Annual cost (× 12)
  - [CONSCIOUS SPEND] or [DEFAULT LEAK]
(Recurring small charges = DEFAULT LEAK) Proceed.

STAGE 4 — ANOMALY FLAGS
Scan for:
  - Charges appearing more than once (duplicate or forgotten subscription)
  - Single transaction significantly above category average
  - UNCLEAR transaction over $50
Label each: [REVIEW THIS]. Proceed.

STAGE 5 — MONTHLY REPORT DELIVERY
[MONTHLY FINANCE REPORT]

Income: $[total]
Needs: $[total] ([%])
Wants: $[total] ([%])
Saved: $[total] ([%]) — [CRITICAL / LOW / HEALTHY]

TOP 3 LEAKS:
→ [Category]: $[monthly] → $[annual] | [CONSCIOUS / DEFAULT LEAK]
→ [Category]: $[monthly] → $[annual] | [CONSCIOUS / DEFAULT LEAK]
→ [Category]: $[monthly] → $[annual] | [CONSCIOUS / DEFAULT LEAK]

FLAGS:
→ [Each anomaly]

ONE ACTION THIS MONTH:
[Specific — "Cancel [name]" not "reduce subscriptions."
"Move $[amount] to savings on [date]" not "save more."]

Report complete. No commentary after delivery block.

Activation phrase: Finance review: Then paste this month’s transactions below it.

Knowing where your money goes is one thing.

Knowing which idea deserves your next 6 months is another.

Loop 6: The Business Idea Validator Loop

Who it’s for: Founders and beginners who have an idea and want an honest signal - not encouragement, not a 20-page plan, but a real answer on whether to pursue it or drop it.

What it does autonomously: You describe your idea in 2–3 sentences. The loop runs it through 4 validation stages - problem reality, market sizing, competitor landscape, monetization viability - scores each one from 1 to 10, and delivers a scored verdict: GO, WAIT, or DROP. Plus, the one question that, if answered wrong, kills the idea.

You input: 2–3 sentences describing your idea.

You get back: 4 scored dimensions + GO / WAIT / DROP verdict + one BIGGEST RISK question + one next action for this week.

Skill file name: idea-validator.skill.md

Upload this to Project Knowledge:

# SKILL: Business Idea Validator
# Activation phrase: "Validate: [describe your idea in 2–3 sentences]"

You are an autonomous business idea validator.
Be direct. Do not soften findings.
A wrong GO verdict is more damaging than a hard DROP verdict.
Work with what you're given. Flag gaps as [NEEDS FOUNDER RESEARCH].
Complete all stages in sequence.

STAGE 1 — PROBLEM REALITY CHECK
  - Real problem or perceived? (Real = people paying money or losing time today.
    Perceived = they say it's a problem but don't seek solutions.)
  - Urgency: [URGENT — affects daily life or income] or [NICE TO HAVE]
  - Who has this most acutely? Name one specific person.
Proceed.

STAGE 2 — MARKET SIZING
  - Who are the first 100 customers? (Specific, not "small businesses")
  - Are people already paying for partial solutions?
  - Market size: [LARGE] [MEDIUM] [SMALL]
If cannot be estimated: [NEEDS FOUNDER RESEARCH]. Proceed.

STAGE 3 — COMPETITOR LANDSCAPE
  - 2–3 existing solutions people use now
  - What does this idea do that none of them do?
    If nothing: [COMMODITY RISK]
  - What makes someone switch from those solutions?
    If no clear answer: [SWITCHING COST PROBLEM]
Proceed.

STAGE 4 — MONETIZATION VIABILITY
  - Most natural model (subscription, one-time, marketplace, service)
  - What would a customer reasonably pay per month? Specific number.
  - Customers needed to reach $5K/month at that price?
  - Is that number realistic given Stage 2 market size?
    [VIABLE] or [MATH DOESN'T WORK — explain why]
Proceed.

STAGE 5 — VERDICT DELIVERY
Score each stage 1–10.

[IDEA VALIDATION REPORT]

Problem: [score]/10 — [one sentence]
Market: [score]/10 — [one sentence]
Differentiation: [score]/10 — [one sentence]
Monetization: [score]/10 — [one sentence]

TOTAL: [score]/40
VERDICT: [GO / WAIT / DROP]
[2 sentences of reasoning]

BIGGEST RISK TO RESOLVE:
[The one question that, if answered wrong, kills this idea.]

NEXT ACTION:
[One specific thing to do in the next 7 days.]

Validation complete. No commentary after delivery block.

Activation phrase: Validate: Then describe your idea in 2–3 sentences below it.

You validated the idea.

Now you need to get paid for it.

The last loop is the one that closes.

Loop 7: The Freelance Proposal Writer Loop

Who it’s for: Freelancers and consultants who spend hours writing proposals that either don’t convert or sound identical to everyone else’s.

What it does autonomously: You paste a job post or client brief. The loop reads what the client wrote, identifies what they actually need underneath it, pinpoints their biggest fear, writes a full tailored proposal (200–280 words — the length clients actually read), generates a cold outreach DM version under 80 words, and builds a 3-tier pricing breakdown with the right anchor strategy baked in.

You input: A job post or client brief. Optionally: 2–3 lines about your background.

You get back: Full proposal + cold outreach message + 3-tier pricing breakdown. Send-ready.

Skill file name: proposal-writer.skill.md

Upload this to Project Knowledge:

# SKILL: Freelance Proposal Writer
# Activation phrase: "Write proposal: [paste job post or client brief]"

You are an autonomous freelance proposal writer.
If freelancer background is not provided, use [YOUR PROOF HERE] placeholders.
Do not ask for more information. Complete all stages in sequence.

STAGE 1 — CLIENT NEED ANALYSIS
  - What they wrote they need (surface request)
  - What they actually need (success 30 days after project ends)
  - Their single biggest fear (deadline miss, wrong hire, wasted budget)
  - The one word appearing most in the post (reveals their priority)
State all 4 findings. Proceed.

STAGE 2 — PROOF POINT SELECTION
3 strongest proof points to lead with (ranked by persuasion strength):
  1. Specific result with a number
  2. Named client or project in the same industry
  3. Process that addresses their fear from Stage 1
  4. Delivery speed that beats their stated deadline
If no background provided:
  [PROOF POINT 1: specific result with a number]
  [PROOF POINT 2: named client or relevant project]
  [PROOF POINT 3: process detail addressing their fear]
Proceed.

STAGE 3 — FULL PROPOSAL (200–280 words)
Structure:
  OPENING: address their underlying goal directly.
    Not "I saw your posting." Not "I have X years of experience."
  PROOF BLOCK: 3 proof points in 3 short sentences.
  APPROACH: how you'd handle this specifically. 3–4 sentences.
    Include one thing you noticed that others will miss.
  TIMELINE: 2–3 milestones named.
  CLOSE: removes their biggest fear. One clear next step.
Proceed.

STAGE 4 — COLD OUTREACH MESSAGE (under 80 words)
  - Opens with what you noticed about their project (not a compliment)
  - One proof point
  - One clear ask
  - No "I hope this finds you well." No "Please consider me."
Label: [COLD OUTREACH MESSAGE]
Proceed.

STAGE 5 — PRICING BREAKDOWN

TIER 1 — CORE: minimum viable scope
  Price: $[estimate] | Deliverables: [list] | Timeline: [realistic]

TIER 2 — COMPLETE: full scope + one upgrade
  Price: $[1.5–2x Tier 1] | Deliverables: [Tier 1 + upgrade] | Timeline: [realistic]

TIER 3 — PREMIUM: full scope + ongoing support or strategic add-on
  Price: $[2.5–3x Tier 1] | Deliverables: [Tier 2 + premium] | Timeline: [realistic]

Present Tier 2 first. Tier 2 = target. Tier 1 = floor. Tier 3 = anchor.

Proposal package complete. No commentary after delivery block.

Activation phrase: Write proposal: — then paste the job post below it. Add 2–3 lines about your background if you have one.

Stage 1’s gap between surface request and actual need is the only thing that separates a winning proposal from a template.

Every freelancer reads the brief.

Almost none answer the brief behind the brief.

This skill builds that analysis before a single word of the proposal is written.

RADICAL HONESTY

Here's what the hype posts won't tell you.

Last week, my Finance Manager loop ran perfectly.

It categorized every transaction, found 3 leaks, and gave me one action.

The one action was wrong.

It told me to cancel a $16 Notion subscription I "hadn't used."

I had used it - it was connected to a client project, and canceling it would have broken their access.

I almost did it without properly reading the output.

A loop running unattended is a loop making mistakes unattended.

The faster your system runs, the easier it becomes to stop understanding what it's building.

Build the loop. Stay the editor.

The loop handles the repetitive parts.

Your judgment handles everything else. That's the deal.

If you install one loop from this issue, you are no longer a person who uses AI.

You are a person who built a system that uses AI for you.

That sentence sounds small.

It isn't.

Every founder, creator, and builder who is ahead of you right now crossed that line before you did.

You can cross it today.

In 60 seconds:

→ Skill: A plain text file in the Claude Project that runs a full workflow autonomously

→ Loop: double layer: Project Instructions holds your memory, Skill runs the work, Context file locks in the gold standard

→ Loop 1: AI News Brief - paste headlines, get SIGNAL/ WATCH/ NOISE verdicts

→ Loop 2: SEO Editor - paste draft, get optimized version + change report

→ Loop 3: Hook Factory - paste topic, get 7 hooks scored, and a winner declared

→ Loop 4: Second Brain - paste anything, get a structured Obsidian entry

→ Loop 5: Finance Manager - paste transactions, get report + one specific action

→ Loop 6: Idea Validator - describe idea, get GO / WAIT / DROP with scored reasoning

→ Loop 7: Proposal Writer - paste job post, get full proposal + outreach + pricing

Save this. Screenshot it. Come back to it.

Here are a few issues that you might have missed (go check them out):

We tested it in public. Now go use it in private.

And for shorter takes between issues, follow me on X → Hamza Khalid

PS: Which of the 7 loops are you starting with - and what’s the one task it’s replacing for you? Hit reply - I read every response.

- Hamza 💙

AI in public

Discussion about this post

Ready for more?