36 people have already signed up for my FREE webinar in 24 hours since I opened the spots.

It is limited to 100 people.

You'll leave knowing what's actually possible when you stop chasing tools and start building a system that runs the business for you.

Join Here

Now to the main topic…

I run my entire content engine through 2 commands. (Before you judge “one more guy who shares some clickbait shit” – stop it.)

I say I run; that doesn’t mean I post right away what AI gives me (LoL).

Give me 3 minutes of your time and you will understand how it works in practice (+ you have the video breakdown below)

The first command finds what’s worth posting this week. It’s automated. I drink coffee while it runs.

The second command writes the drafts, outlines, and visual prompts for YouTube, my newsletter, and LinkedIn. In my voice. Following my brand guidelines.

That second one is not automated. On purpose.

Look, I’ll be honest with you. I would never let AI write my whole post start to end. Never.

The day I do that is the day I lose the only thing that’s actually mine. My opinions. My edge. My way of seeing things.

So the system writes the draft. I rewrite the draft. That’s it.

Here’s what each command actually does.

Command 1: /content-research

Every morning, this command goes out and pulls 50 to 100 posts per platform across LinkedIn, Substack, YouTube, X, and Reddit. Across the channels and people I actually care about.

That’s around 351 posts in one run.

Then five sub-agents (one per platform) run in parallel. They aggregate. They dedupe. They pre-filter.

Then three more agents score every post on three things:

→ Does it match one of my content pillars?

→ Is it actually new, or already saturated?

→ Does it sound like something my voice could carry?

Out of 351 posts, the system picks 25 winners. Divided clean by my three pillars.

This is literally curated feed in my Notion, LoL, with full context about each post including links, engagements, angles etc…

The system also reads my last 7 days of coaching and sales calls (Fathom records them). It mines what people actually asked me. Then it pushes those questions into the same idea pool.

So the 25 winners aren’t just “what’s trending on the internet.”

The 25 winners are “what’s trending and what your buyers are asking you about right now.”

That’s a completely different filter.

Every winner shows up in Notion as its own subpage:

→ The hook (verbatim)

→ The full transcript or post copy

→ The repurpose path (YouTube? Newsletter? LinkedIn?)

→ A direct link to the original

I open Notion. I read 25 winners over coffee. I pick one or two.

That’s it for command one.

This step alone saves me 6 hours every week. Six hours of “what should I post next.” Gone.

Command 2: /content-production

I flip a winner from “Not Started” to “In Production” in Notion.

Then I run /content-production.

The system reads:

→ My business profile

→ My ICP (who I’m writing for)

→ My current priorities

→ My voice DNA file

→ My content pillars

→ My brand guidelines

→ The full transcript of whichever winner I picked

Then it produces every format that winner needs:

→ A YouTube brief, outline, script, packaging, title, description

→ A newsletter edition (the one you’re reading right now)

→ A Substack note

→ A LinkedIn carousel and infographic and contrarian post

All in my voice. All branded. All ready for me to rewrite.

Last week I picked one idea from Nate Herk. The post was “every layer of Claude Code explained.” Within 20 minutes the system gave me a LinkedIn carousel, an infographic, a contrarian post, and a draft newsletter on it.

I didn’t ship any of those drafts as-is. I opened Whisper Flow, talked into it, added my opinions, my framing, my Bosnia-shaped way of seeing the same idea. Took me 30 more minutes.

That’s 50 minutes from “winning idea” to “ready to publish across 3 channels.”

Without the system? A full afternoon. Maybe two.

Why I don’t let AI write end-to-end

This is the part I want you to actually hear.

Most LLMs are programmed to please you. Not to push you. ChatGPT gives you 7/10 when reality is 3/10. It tells you the post is great. It tells you the hook is strong. It’s not. It’s never that strong.

When I let AI write for me, I sound like everyone else. Same hooks. Same structures. Same “5 tips to...” soulless wallpaper that’s drowning every feed.

When I rewrite AI drafts, I sound like me.

The difference is the difference between getting likes and getting clients.

So the system does the research. The system does the first draft. The system does the visual prompts. The system does the packaging.

I do the thinking.

That’s the contract. That’s why this works.

The compound part

Here’s the thing nobody talks about.

Every time I run /content-research, the brain remembers. Every winner I pick. Every angle that worked. Every coaching-call question that turned into a post that landed.

Next week, the system is sharper than it was this week. The pillars get tighter. The voice scoring gets more accurate. The winners get more obviously mine.

That’s the part no tool on the market gives you. Every SaaS dashboard is the same on day 1 and day 365. This system gets sharper.

That’s why I keep saying: you don’t need another AI tool. You need to finish one system that remembers.

The receipts

→ 351 posts scanned per run

→ 25 winners ranked, with full context

→ 6 hours back every week

→ 0 generic AI-slop drafts published

→ My voice intact. Period.

That’s what 2 commands look like.

To your next client,

Sabahudin ‘one command, 5 channels’ Murtic

P.S.

What’s the one part of your content workflow that’s still eating your week? Reply and tell me. If five people say the same thing, I’ll write the next edition about exactly that.

P.P.S.

Massive shout-out to the Trigify team. Without their scan engine, none of this works.

Want to work with me?

Fill out the form for a FREE AI AUDIT CALL (30 min) and see if I can genuinely help you:

Fill out my qualifications form

This is a focused 30-minute call where I map your workflows, tools, and daily reality – then show you where AI would actually make a difference right now.

After this call you’ll have the following:
→ A clear picture of which parts of your business are ready for AI
→ The 2–3 highest-impact workflows to automate first
→ A high-level overview of how your second brain should operate.
→ A concrete next step - not a vague plan, an actual action

Before we go into the main topic…

On June 17 at 17:30 GMT+2, I’m running a free live webinar session where I open my actual setup and show you how one system runs my entire business: outreach, content, follow-ups, CRM, delivery, and money.

If you’ve tried 3+ AI tools and ended up with a folder of half-built workflows that don’t talk to each other, this session shows you what one connected system actually looks like.

You’ll see the dashboard I work from, how every routine logs to one place, and how the brain compounds week over week. Real commands with real output.

It's FREE → limited to 100 spots.

Save your spot

Today’s special guest edition…

3 minutes.

That is all it took for an on-brand infographic to land on your screen.

Right colors. Right fonts. Right vibe.

I sat on the other side of a call with Samy, impressed.

Here are some outputs:

Most of the people have been doing this the hard way.

Or they outsource designers for visuals

That’s why we are sharing the full design system for FREE with you guys (grab the repo at the end of the edition)

In case you prefer watching more, you can access the full edition below:

Quick favour, I am testing this type of format where I invite my close network like Samy to share some secret sauce with you guys. So, a question for you (will appreciate your vote because I do this for you):

A quick word on the guy who built it

The man behind that screen was Samy Chouaf.

You can read his substack newsletter here

Samy is a data and AI consultant out of France. Open-source builder. Posts some of the cleanest technical takes on AI workflows I have seen on LinkedIn. I follow him because he ships, not just talks.

We hopped on a call last Friday so he could walk me through his latest project. An open-source design system that runs on Claude Code and turns plain-English prompts into carousels, infographics, and slide decks. On brand. In minutes.

I came off that call rebuilding how I think about AI design.

The bill solo creators pay every week

Designing your own visuals is brutal math.

A carousel costs 2 hours if you fly and 4 if you care.

An infographic eats half a morning. Stack three a week, and your calendar is gone.

Most people pick one of two exits. Hire a designer and lose margin. Or post less and lose momentum.

Samy built a third door.

Why your AI design keeps drifting

Here is the part most people get wrong.

They drop a screenshot into an image model and hope it copies the brand. They paste a Figma link and pray the hex codes hold. They drag in a logo and watch the model paint a slightly-wrong version of it three slides later.

Vision models are not reading your brand. They are guessing it.

A guess is fine for inspiration. It is terrible for production.

Samy flipped the angle. Instead of pointing the model at a picture, he wrote his brand as code. Hex values, font tokens, border radii, spacing rules. All sitting in plain text Claude can read with zero interpretation.

The model stopped guessing. It started executing.

“This one I really like,” Samy said.

Drift killed. 95% done in one pass.

What sits inside the repo

Three files do most of the work.

DESIGN.md holds the brand. See it as the unique source of truth of your design system. Colors, fonts, shadows, spacing, border radii… Every visual decision your eye makes becomes predictable, and the AI agent stops guessing anymore. Google Labs open-sourced this format with their Stitch design tool.

To learn more about it:

DESIGN.md Repo

Samy built around it. Run his /setup command, point it at your site, and the file fills itself (using Firecrawl brand extraction).

CLAUDE.md is the operating manual. Folder map, scripts, which sub-agent handles which task. Kept under 50 lines because every line loads on every session, and bloat costs tokens.

components/ folder is the Lego set. JSX blocks for cards, headers, checklists, callouts. Claude grabs one and drops it in. No redrawing from scratch. No drift between posts.

On top, Samy wired isolated sub-agents. One handles each asset type. Each carries only the context it needs. Less noise. Cleaner output.

The move that saved my Friday afternoon

The piece that genuinely broke my brain was the design brief.

Run /infographic and the system asks four questions. Topic, audience, tone, structure. Then it stops.

It writes you a brief. Which icons it picked. Which layout is planned. Which copy lives where. You read it like a designer reading their own notes before they touch the canvas.

You approve. Or you redirect.

Then it builds.

That single pause kills the regenerative spiral most people lose entire afternoons in. You know the one. “Lighter. Smaller. Different icon. Move that. Start over.”

Approve the brief once. Ship the asset cleanly. Move on with your week.

Output lands in a local gallery. Every design you have ever made is scrollable. Hit download, or push the file to Figma for final polish.

Five things I am stealing this week

Five takeaways from those 46 minutes that I am wiring into my own system.

• Code beats vision for anything brand-critical. The model that reads instructions beats the one that interprets pictures, every single time.

• One brief, one pass. The plan-mode preview saves more hours than the design itself.

• DESIGN.md is the missing standard solo creators do not know exists. Every visual decision in one place a machine can read.

• Sub-agents stay sharp because they stay narrow. Same principle that makes my own 2nd Brain OS routines work.

• Same repo, different engine. When Claude caps for the day, Sami swaps over to Cursor or Codex on free tier. Identical output. That one I had never thought about.

Total monthly bill to run this whole thing? Claude Pro at $20. Figma at $20 if you want the final tweak layer. Everything else rides on free APIs.

Two paid subscriptions. One open-source repo. A LinkedIn feed you can finally fill without trading half your week for it.

Where to grab the whole thing?

Clone Samy’s repo: HERE

If you ship visuals alone, you owe yourself one evening with this repo.

To your next client,

Sabahudin ‘one command, 5 channels’ Murtic

P.S.

The lesson from this call was not the system. It was watching someone treat his brand as infrastructure, not decoration. Encode it once. Let the machine do the rest.

P.P.S.

If you clone the repo and get stuck, reply to this email. I will send you the exact prompts Sami used during our session.

Want to work with me?

Fill out the form for a FREE AI AUDIT CALL (30 min) and see if I can genuinely help you:

Fill out my qualifications form

This is a focused 30-minute call where I map your workflows, tools, and daily reality – then show you where AI would actually make a difference right now.

After this call you’ll have the following:
→ A clear picture of which parts of your business are ready for AI
→ The 2–3 highest-impact workflows to automate first
→ A high-level overview of how your second brain should operate.
→ A concrete next step - not a vague plan, an actual action

I have a problem.

Every time Claude does something I don’t like, I add a rule.

“Never X again.” “Always Y.” “Don’t do Z.”

Six months in, my CLAUDE.md is a graveyard of one-off fixes.

Last Tuesday I was about to add another one. The patch for yesterday’s weird output. My finger hovered over save. Then I stopped.

Because that file used to be 80 lines. Now it was pushing 300. And every rule I’d added was sitting somewhere else too. In context/. In .claude/rules/. In memory.

I had no idea how much dead weight I was carrying.

What is CLAUDE.md?

It’s one file at the root of your project. Claude reads it first every time you start a new conversation.

Think of it as the binder you’d leave for an assistant who shows up every morning with amnesia.

No memory of yesterday. No memory of the client. No memory of the rules.

So you leave one binder on the desk. Every morning the assistant reads it. Then gets to work.

That binder is your CLAUDE.md.

It tells Claude who you are. What you sell. What files matter. What to never do.

The better the binder, the better the work.

A bloated binder is just as bad as no binder. The assistant burns the first hour re-reading 300 pages of contradicting rules before they’ve even started.

That’s why the audit matters.

If you think your CLAUDE.md file is a good, think again

If you’ve been running Claude Code for a few months, this is happening to you too.

You ship something. You hit a bad output. You add a guardrail. You move on.

Three months later, your instruction layer is a museum of small fixes nobody curated.

Conflicts you can’t see. Rules that fight each other. Defaults you told Claude to do that Claude already does on its own.

Every line is a token. Every duplicate rule is a token Claude reads twice. Every conflict is a coin flip on which version wins.

Here is what I hear from last 20 calls with prospects

Most people never audit the instructions they wrote.

They write. They patch. They never read back.

Then they wonder why Claude feels slow. Why outputs feel bloated. Why the same rule shows up in two files with two different wordings. Why they spend tokens so fast.

That was me as well. So I built a prompt to do the audit for me.

The Anatomy of CLAUDE.md audit prompt

The prompt reads your CLAUDE.md.

Every file in .claude/rules/. Every file in context/. Your MEMORY.md and every file it points to. The frontmatter description of every SKILL.md. Nothing else.

Then it runs each rule through a 5-question filter:

→ Is this a default Claude would already do?

→ Does it conflict with another rule somewhere else?

→ Is the same thing stated in 2+ places?

→ Is this a reactive patch from one bad output?

→ Is this vague enough to be interpreted differently each run?

A rule can fail multiple filters. Every fail gets flagged.

The output is four sections. An audit table grouped by source file. A cut list with one-line reasons. A conflict map naming both sides of every contradiction. And a full rewrite of CLAUDE.md with the dead weight stripped out, ready to paste in.

The first time I ran it I found rules in three places saying the same thing in three different wordings. Defaults I’d written out that Claude already follows. A handful of reactive patches I’d forgotten the reason for. Vague lines that meant nothing useful.

I cut what failed. Kept what was load-bearing. My instruction layer got lighter and Claude got faster.

That’s it.

Here is the example of my audit suggestion:

Leave a comment

Here’s the prompt. Swap in your own memory path and run it.

# Task: Audit your Claude Code instruction layer for dead weight

You are auditing the rules and instructions that govern how I work with Claude Code in this repo. The goal is to surface redundancy, conflicts, and noise — then propose a cleaned CLAUDE.md.

## Scope — read in this exact order, nothing else

1. `CLAUDE.md` (project root)
2. Every file in `.claude/rules/` (the auto-loading rule layer)
3. Every file in `context/` (working memory — business profile, ICP, priorities, voice, patterns, offers, etc.)
4. The memory index at `[YOUR-MEMORY-PATH]/MEMORY.md` and each memory file it points to
5. The frontmatter `description` field of every `SKILL.md` under `.claude/skills/**/SKILL.md` (description only — NOT the full skill body)

DO NOT read skill bodies, agent files, hooks, or command files. If you find something obviously broken outside scope, note it once at the end — do not expand scope.

## What to evaluate

For every distinct rule, instruction, or preference you find across the files in scope, run it through this 5-question filter:

1. **Default behavior** — would Claude already do this without being told? (e.g., "be helpful", "respond in English", "don't lie")
2. **Conflict** — does it contradict another rule in a different file? Name both locations.
3. **Duplication** — is the same instruction stated in 2+ places? Name all locations.
4. **Reactive patch** — does it read like it was added to fix one specific bad output rather than improve the system overall? (Signal: hyper-specific scenarios, narrow caveats, "never do X again" framing)
5. **Vague** — would you interpret this differently on different runs? (e.g., "be more natural", "use a good tone", "write better")

A rule can fail multiple filters. Flag every filter it fails.

## Output — produce these 4 sections, in this order

### Section 1 — Audit Table
A markdown table with these columns: `Source File | Rule (paraphrased, ≤15 words) | Filters Failed (1–5 or "—") | Notes`
One row per rule. Group rows by source file. Skip rules that pass all 5 filters.

### Section 2 — Cut List
A flat list of every rule you'd cut, formatted as:
- `[source file] "rule excerpt" — one-line reason for cutting`
Order by source file, then by severity (rules failing 3+ filters first).

### Section 3 — Conflict Map
Every conflict you found, formatted as:
- `Conflict between [file A] and [file B]: [file A says X, file B says Y]. Resolution suggestion: [which to keep, why]`
If none found, write "No conflicts detected."

### Section 4 — Cleaned CLAUDE.md (proposed)
A full rewrite of `CLAUDE.md` with the dead weight removed. Preserve every load-bearing instruction. Preserve all tables. Strip vague instructions, defaults, redundancies, and reactive patches.
Output it inside a fenced markdown block so I can copy-paste it. Show line count before/after at the top of the block as a comment.

## Stop conditions — MANDATORY

- DO NOT modify any file. This is read-only analysis.
- DO NOT write the cleaned CLAUDE.md to disk. Output it in the chat as a code block only.
- DO NOT proceed to "phase 2" or offer to apply changes — finish the report and stop.
- If a file in scope is missing or empty, note it in the audit table and continue.
- Stay inside the 5 scope folders/files listed. No directory expansion.

## Success criteria

Done when all 4 output sections are produced, every flagged rule has a reason, every conflict names both source files, and the proposed CLAUDE.md is a complete drop-in replacement (not a diff, not a snippet).

Run it on your own setup.

You’ll find duplicates. Conflicts. Vague rules. Defaults you didn’t need to write.

The cleaned version reads like the CLAUDE.md you wished you’d written from day one. Less weight. Less drift. Same outcomes. And Claude stops re-reading the same instruction four times per conversation.

That’s the whole point.

P.S. Two tips before you run it. One, commit your current CLAUDE.md first so you can diff. Two, the prompt is read-only on purpose. Don’t let Claude apply the cleanup. Read it. Decide what stays. Then you paste the version you trust.

P.P.S. If you find more than 10 cuts in your first run, you waited too long. Put a reminder on your calendar to re-run it every 30 days. The instruction layer is the only file that compounds against you if you don’t curate it.

To your next client,

Sabahudin “less weight” Murtic

P.S.

Want to work with me?

Fill out the form for a FREE AI AUDIT CALL (30 min) and see if I can genuinely help you:

Fill out my qualifications form

This is a focused 30-minute call where I map your workflows, tools, and daily reality – then show you where AI would actually make a difference right now.

After this call you’ll have the following:
→ A clear picture of which parts of your business are ready for AI
→ The 2–3 highest-impact workflows to automate first
→ A high-level overview of how your second brain should operate.
→ A concrete next step - not a vague plan, an actual action

Before we go into the main topic, quick life update: I am attending the SaaStanak conference in Šibenik (Croatian coast). Amazing conference, sunny beach, and the best possible speakers from the GTM world.

If you are curious about sea vibes, check out this video to get some feeling about the atmosphere:

The guy at the end of the video said the best: triple S → Sun. Sea. SaaS

There are few tickets left; you can grab them here: https://www.entrio.hr/en/event/saastanak-2026-25090

(you can use my code “sabahudin” for a discount.)

Here are the keynote speakers:

Now to the main topic…

For the last several months, I’ve been running my entire business through a personal Claude Code 2nd-brain-OS.

Inbox. Calls. Proposals. Content. Client work. Every routine automated or assisted.

I never planned to release any of it publicly. These were tools I built for myself — rough in places, very specific to how I work.

But enough people have asked. So today I’m releasing 12 of these skills as a public GitHub repo.

Free. No opt-in. No course. Just the files.

Here’s what’s inside — and what each one actually does.

GRAB IT HERE

Operations

1. prompt-master — the one that makes everything else work better

You type something half-formed. You dictated it on the way to your desk. You know what you want but the words aren’t there yet.

Prompt-master takes the mess and turns it into a single, clean, production-ready prompt — formatted for whichever AI tool you’re running it on. Claude Code, ChatGPT, Cursor, Midjourney, n8n, ElevenLabs. 30+ tool profiles built in.

Example: Voice note — “I want to make like a script for a YouTube video about AI workflows, you know, something that actually converts” becomes a precise, scoped Claude Code prompt with output format, scope boundaries, and stop conditions in under 10 seconds.

There’s also a hook that auto-detects voice input and intercepts it before Claude executes anything. No more half-completed tasks from rambling dictation.

2. skill-improve — your AI gets smarter every time you correct it

You run a skill. The output isn’t quite right. You say “never do that again.”

Most people stop there. The correction dies in the conversation.

Skill-improve takes what you said and writes it permanently into the relevant skill’s file. Next time it runs, it already knows.

Example: “The email report is too long — just give me the action items, not the full table” gets turned into a rule inside the skill. The next 50 runs are different because of one correction.

This is what makes the system compound instead of repeating the same mistakes.

3. autoresearch — runs an optimization loop so you don’t have to

Give it a skill, a file, or any piece of your OS. Give it quality criteria. It iterates, tests, measures, and improves automatically.

Deterministic mode uses eval.py to score mechanically (structure, format, keywords). AI judge mode uses an LLM rubric to score subjectively (tone, completeness, authenticity).

No prompt tweaking. No “try this, see what happens.” A loop that runs until the metric improves.

Example: “Autoresearch my email-report skill — it needs to surface dropped balls better” runs 10 iterations with measurable pass rates. You see before/after and approve the winning version.

Zero setup. Works on any file in your project.

4. email-report — not a summary. a decision aid.

This is the first thing that runs every morning.

It scans 24 hours of Gmail. Categorizes every email into INBOUND / CLIENT / REVENUE / ADMIN / FYI. Scores each one with a priority rubric. Surfaces only the emails where you’re the bottleneck.

Then it finds every thread where you last replied more than 3 days ago and the other side is waiting.

Example output:

• “🔥 Simon — Tally form, CCO at 200-person company. ICP fit: HIGH. Reply within 24h.”

• “🟠 Dropped ball: proposal from 5 days ago, no follow-up.”

After reading it, you know in under 60 seconds which things to act on. That’s the whole point.

5. check-recent-calls — captures every commitment you made on a call

You finish a coaching call. You said “I’ll send you the workflow by Wednesday.” You said “next step is getting Apollo set up.”

Without this skill, you try to remember. You miss things.

This skill scans every call transcript from the last 3 days, extracts high-confidence commitments, and creates the Notion task row automatically. Every morning, before you open your laptop.

Example: A Fathom transcript from Tuesday’s coaching call → 2 auto-created Notion tasks, 1 flagged for review. Tagged with the exact quote that triggered the extraction so you can verify in one glance.

6. meeting-prep — full research before every discovery call

The morning before a discovery call, this runs automatically.

It pulls their LinkedIn profile, their last 5 posts, 30 days of web research (podcasts, funding, news), and builds a concise research brief. Then it writes a personalized 12-step sales script with questions drawn from their actual profile and language. Then it generates a PDF pitch deck for the specific offer they’re most likely a fit for.

Example: Discovery call in 2 hours with a founder of a 12-person ops consultancy → brief, script, and deck ready in 8 minutes. The script opens with a question about their specific LinkedIn post from last week.

Content

7. youtube-scripting — writes scripts that sound like you

Takes a video outline and turns it into filming-ready material — full script, bullet points, or hybrid.

The Voice Authenticity Test runs before outputting. It asks: does this sound like the creator’s actual voice, or a polished LinkedIn guru? If the answer is “guru,” it rewrites.

Example: Outline for “how I use Claude Code in my morning routine” → hybrid script with tight hook options, natural middle section bullet points, and a soft CTA calibrated to the pillar. Filming checklist included.

8. newsletter-subject-lines — 8-10 scored options from 100 frameworks

Takes a newsletter draft. Identifies the edition intent (educational, story, authority, roundup, launch). Loads only the relevant framework categories. Generates 8-10 subject lines, scores each one on 4 criteria, recommends the strongest pick with 2 A/B variants.

Standalone mode needs zero setup — just paste the draft and it runs.

Example: A newsletter about how you replaced your ops manager with Notion automations → “the ops manager i never hired”, “how i replaced $60k/year with 3 Notion databases”, “everyone said this would break. it didn’t.” — all scored, all with preview text.

9. linkedin-lead-magnet — giveaway posts that convert without begging

Writes LinkedIn posts that give away a real resource using the structure: result first, proof, breakdown, use cases, self-aware CTA.

The self-aware CTA is the detail that matters. Instead of “comment ‘Templates’ and I’ll send it” — “Comment Templ... I’m kidding, the link is below.” Breaks the trope. Gets more clicks.

The skill ships with a brand DNA template you fill in once. After that, every post sounds like you — not like every other consultant on LinkedIn.

Offer / Sales

10. pitch-deck — a branded PDF deck from your Offer Doc in one command

Takes your Offer Doc and builds a complete HTML+PDF pitch deck. Slide count is driven by the content, not a fixed template. Uses your brand assets (profile photo, logos, colors). Self-contained — no broken image links when you send the PDF.

Example: Offer Doc for a 6-week consulting engagement → 11-slide deck with problem, vision, process, proof, pricing, and CTA. Built in the brand. Ready to send.

Delivery

11. visual-coaching — co-designs Excalidraw visuals for coaching sessions

Not a chart generator. A guided co-design process.

It asks what the learner needs to understand, what type of visual fits (mindmap, roadmap, blueprint, process flow), and what the learning goal is. Then it illustrates — shapes that carry meaning before text is added. The design philosophy is explicit: illustrate, don’t decorate.

Example: Client needs to understand how their content system flows → hub-and-spoke diagram with the distribution channel at center, formats radiating out, one orange accent on the highest-leverage touchpoint.

Design

12. brand-guidelines — applies Anthropic’s official brand to any artifact

If you’re building anything for or about Claude — presentations, proposals, client materials — this applies Anthropic’s official colors (dark/light/orange/blue/green) and typography (Poppins + Lora) automatically.

Zero MCP setup needed. Python + python-pptx only.

How to get them

Everything is at: https://github.com/sabahudin-web/claude-skills-starter-pack

Clone → copy the skills you want → connect MCPs → run.

The three zero-setup skills work today with no configuration: prompt-master, skill-improve, autoresearch.

Full setup guide in INSTALL.md covers every MCP connection, Notion ID, and context file you need.

A note on what this is

These skills are tools I built to solve real problems in my business. They work because they were tested on actual work, not demos.

The repo won’t tell you what to do with your business. It gives you the machinery. What you build on top of it is yours.

The references/linkedin-content-universe.md file in the lead magnet skill is the clearest example of this. It ships as a template with bracketed slots — your mentor, your origin story, your community. Fill it in once. Every post after that pulls from your actual brand DNA.

That’s the whole design philosophy: infrastructure you own, not a black box you rent.

To your next client,

Sabahudin ‘one command, 5 channels’ Murtic

P.S.

Which of the 12 would you install first? I’m curious whether people start with the zero-setup ones or go straight for the complex ones, like meeting prep.

Leave a comment

Want to work with me?

Fill out the form for a FREE AI AUDIT CALL (30 min) and see if I can genuinely help you:

Fill out my qualifications form

This is a focused 30-minute call where I map your workflows, tools, and daily reality – then show you where AI would actually make a difference right now.

After this call you’ll have the following:
→ A clear picture of which parts of your business are ready for AI
→ The 2–3 highest-impact workflows to automate first
→ A high-level overview of how your second brain should operate.
→ A concrete next step - not a vague plan, an actual action

Outreach isn’t dead…

Bad outreach is (and it will be always dead)

But what we can do now with Claude code when it comes to prospecting → is crazy.

We just replaced clicking and building some lists with simply typing.

Old way: open Sales Navigator or Apollo, build filter combinations by hand, click the job title dropdown, click geography, click company size, hit search. Hope for the best.

Then copy the names to a spreadsheet. Manually check each profile. Write a generic message, send it and forget to follow up.

New way: type “find me consultants at small agencies in the US who are frustrated with manual outreach and building their practice” — and the system surfaces them, enriches them, scores them, and drafts the first message.

That’s the shift. People call it vibe prospecting.

You describe the person you’re looking for.

Claude Code translates it into targeting.

The leads show up. You talk to the right people.

The filter isn’t a form you fill out. It’s a conversation you have with your system.

Here’s how the whole thing works.

The /outreach Command

I run my entire business through 9 commands.

/outreach is the one that fills my pipeline.

One command. Four sub-routines.

Each sub-routine solves a different problem — and runs at a different cadence depending on what stage you’re in. You don’t run all four at once. You pick the one that fits the moment.

My results since I do this outreach routine?

Here is the Payoneer (my payment provider) from March to April:

Now let me break each routine, one by one…

Sub-routine A: Check Who Replied (when the campaign is already active and running)

Runs automatically every morning at 9 AM.

Why it exists

If someone replied to your first message, you need to know — and act within 24 hours. After that, the conversation goes cold fast. Most people miss this window because they’re not checking their LinkedIn inbox (in my case → Aimfox inbox) religiously.

This routine checks it for you.

How it works, step by step

→ Step 1: Claude connects to Aimfox and scans your active campaigns for leads who accepted your connection request AND replied to your first message.

It outputs a list of those lead IDs.

→ Step 2: Claude applies the “replied” label to each of those leads inside Aimfox. This moves them out of the cold pipeline and into the active conversation bucket.

→ Step 3: For leads who replied and match my coaching offer signal (cc-coaching), Claude pulls their enriched profile from Trigify, finds a recent AI-related signal from their industry, drafts a personalized nurture message, and pushes them to my Notion CRM with the draft already written.

I used their reply to nurture the conversation and make it more about them and less about me by asking simple questions about how AI (because my offer is attached to the AI) is affecting their industry or role. This way I create urgency to react.

+ that messaging angle is in my CRM, stored as notes so I can come back and revisit my active pipeline and move the conversation until I book a call.

Why it’s powerful

I never miss a reply. Before this, replies would sit in Aimfox for 48 hours. Now the system finds them, labels them, and drafts my next move before I’ve opened my laptop.

Sub-routine B: Launch a New Campaign

Run every two weeks.

Why it exists

Cold outreach needs fresh fuel. Every two weeks this routine replenishes the pipeline with qualified people who actually fit your offer.

How it works, step by step

→ Step 1: Claude searches Apollo for solopreneurs, consultants, and SMB owners across US, UK, EU, Singapore, and UAE. Max 50-200 leads per run. Apollo has MCP support → Sales Nav doesn’t, which means Claude can’t automate anything with it.

→ Step 2: Claude runs each lead through Trigify CLI in parallel batches — 10 leads at a time, 5 batches simultaneously. Pulls name, title, company, location, and most recent LinkedIn posts.

→ Step 3: Every lead gets scored against my ICP matrix. Only Tier 1 and Tier 2 make it through.

→ Step 4: Qualified leads land in an Aimfox draft campaign. Claude does NOT launch it. I review everything before it goes live.

→ Step 5: Claude writes DMs based on how AI impacts their actual role and industry.

Why it’s powerful

A 50-200-person qualified, enriched, scored pipeline — with draft DMs already written → in one command.

Sub-routine C: Intent Keywords Campaign

Run manually, recommended weekly.

Why it exists

The best leads aren’t the ones you find. They’re the ones already talking about the problem you solve.

How it works, step by step

→ Step 1: Claude sets up 5 keyword searches in Trigify: “Claude Code,” “AI for consultants,” “lead generation automation,” “build in public with AI,” “Cowork OS.”

→ Step 2: From top-ranked posts, Claude fetches 150 to 200 engagers.

→ Step 3: Headline filter against ICP — no external tool, just matching job titles.

→ Step 4: Trigify enriches each filtered lead.

→ Step 5: ICP score + intent label: Claude Code OS Build, Agent Routine OS, or Cowork OS.

→ Step 6: Into Aimfox draft campaign.

→ Step 7: DM variations by topic signal. I review. I launch.

Why it’s powerful

These people are already publishing or engaging with the topic that is relevant to my offer. The first message doesn’t have to do heavy lifting.

Sub-routine D: Intent Profiles Campaign

Run manually, recommended twice per week.

Why it exists

People who engaged with your content or the creators your ICP follows are the warmest possible starting point.

How it works, step by step

→ Step 1: Claude monitors 6 LinkedIn profiles: my own account plus 5 Claude Code creators my ICP follows.

→ Step 2: Fetches 150 to 200 engagers from top posts.

→ Step 3: ICP headline filter.

→ Step 4: Trigify enrichment.

→ Step 5: ICP scoring + warmth tags: “sab-warm” or “creator-warm.”

→ Step 6: Aimfox draft campaign.

→ Step 7: DMs by warmth track. I review both versions before anything launches.

Why it’s powerful

These people already showed a signal. The opener feels less cold because there’s something real to reference.

The First Message Strategy

Let me be real.

The first messages I send aren’t pitches. Not even close.

They’re questions. Framed like I’m running some light research. I’m opening conversations with people who might be relevant to what I offer, to see what’s possible.

After the first automated message gets a reply, I take over manually. I look at who replied, what they said, what they’re dealing with. Then I label them: relevant to my offer, or not.

The system surfaces the conversations. I decide who’s worth continuing with.

The Tool Stack (and What It Actually Costs)

• Trigify Essential $140 Lead prospecting, keyword monitoring, AND content inspiration

• Aimfox $49 LinkedIn outreach automation and inbox management

• Apollo $50 Lead database with MCP (replaces Sales Nav)

• Claude Max $100 Everything: all 9 commands, content, CRM, reporting

• Notion $12 Entire business database: CRM, tasks, KPIs, client deliverables

$351/month — but this isn’t an outreach-only stack.

Trigify powers my content inspiration.

Claude Max runs every routine.

Notion is my entire business frontend.

Total tool spend including other tools is around $400/month.

With Claude Code as my operational dashboard and Notion as the frontend, I save 15+ hours per week. That time goes toward real conversations and client work.

$400/month to reclaim 15 hours a week. The math works.

/outreach Is One of 9 Commands

• /morning Email scan, calendar review, client tasks, daily plan: done before you sit down

• /content YouTube brief, LinkedIn post, or newsletter draft: in your voice

• /outreach LinkedIn replies checked, leads surfaced, follow-up messages drafted

• /followup Proposals and conversations that need a reply: with drafts ready

• /crm Recent leads reviewed, gaps filled, records updated

• /finance Revenue vs goals, unpaid invoices, anything off

• /audit-brain OS health check: memory, skills, workflows audited and a prioritized fix report pushed to Notion

• /weekly Full week summary pushed to your Notion dashboard

• /monthly Month retrospective: what worked, what to change

• /eod Work logged, decisions captured, tomorrow’s context ready

That’s the whole business. 9 commands. The system runs.

If you want to see how this looks in practice — I’m doing free discovery calls this month.

No pitch. Just a conversation. Book it below.

→ [Book a free call with Sab]

To your next client,

Sabahudin “Vibe Prospector” Murtic

P.S.

Which of the 4 sub-routines would you run first if you had this system today? Reply and let me know. Genuinely curious which problem feels biggest right now.

• content repurposing in 1 claude code session (exact system)

• The AI Brain That Never Forgets

• Keep an eye on your competitors while you sleep.

• Try This Claude skill that turns your LinkedIn connections into buyers

• I just open-sourced my entire offer creation system

So there's no special introduction or story today. The only thing I can say is that I was busy in a good way setting up my Cowork OS and Second-Brain-OS for my clients (which I will break down in future editions).

If any of these are interesting for you, I am offering 5 1:1 coaching sessions plus setup. You can check if you are fit for my free discovery call

Now, to the main topic

How to use plan mode with Claude Code.

Here is what I learn from my own mistake…

I dictate my thoughts to Claude every single day. (which is great, talking instead typing)

But…

For a while, the same pattern followed me.

I’d open plan mode. Record myself rambling for 90 seconds. Paste the transcript in. Expect a sharp plan to come back.

It never did.

The reality is…

Claude Code plan mode doesn't matter → Your input does.

Claude is only as good as what you feed it.

Most people open plan mode, dump their voice notes in, and wonder why the plan comes back wrong.

The problem isn’t Claude.

The problem is that a voice note is not a prompt.

Voice notes are ingredients. Unorganized. Half-formed. Jumping between ideas. That’s how a human brain dictates. That’s confusing for Claude to understand it.

No amount of model quality fixes a bad input at the start.

So I built a small fix.

It sits in front of plan mode. Before Claude runs plan mode and sees my rambling, it intercepts it.

Here’s how it works:

I have a rule inside my Claude Code setup that detects voice transcript patterns. Filler words. Run-on sentences. Self-corrections. When it catches those signals, it automatically runs the operations-master-prompt skill on the raw text.

That skill has prompt frameworks built in. The real ones. The ones that actually work.

It reads my rambling, picks the right structure, and reshapes it into a clean, scoped prompt.

Audit and proof. Current state. Scope. Exact steps. Stop conditions. Output. Constraints.

All of it, pulled from my messy 90-second dictation.

Then it shows me the cleaned version and asks: “Would you like me to run this?”

I say yes. The plan comes back sharp. No back-and-forth. No corrections. No second prompts.

Subscribe now

Bonus Tip to maximise your plan mode

It is simple: when you record yourself, tell Claude to use the native tool inside of which is called “Ask User Question”.

This way Claude code will ask you every possible question that is related to your plan so it can be 100% sure what you want.

Probably you will be frustrated by how many questions you will be asked, but hey, it's better to fix it at the beginning when you still don't spend tokens.

Why are these 2 approaches important for plan mode?

Well, it is obvious. If you have a good plan, you won't have to go back and forth refining the actual plan, and the end result will be exactly what you envisioned.

Furthermore, you will save a large number of tokens because if you execute the plan and are dissatisfied with the results, you have already spent the tokens and will have to redo it. In other words, you're spending 2x, 3x, or 10x more tokens because you're constantly improving your final product.

However, with a structured prompt, Claude will know exactly where to find the resource, what to build, what the actual goal or value is.

"Ask User Question" will capture any additional information so that Claude is 100% certain you are on the same page.

Install this to your claude code sessions

If you want the setup file, I created a GitHub repo.

Drop it into your Claude Code and clone that repo. This way you will update your rules and hooks. Done.

GRAB IT HERE

Whenever you go into plan mode and start rambling, it catches you and restructures automatically.

To your next clean plan,

Sabahudin “your words are data, not a prompt” Murtic

P.S.

How do you currently talk to Claude? Do you type clean instructions or just dictate and hope? Reply or leave a comment and tell me. Genuinely curious how many people are living in the back-and-forth loop.

Leave a comment

Find this edition valuable? Share this with your network:

Want to work with me?

Fill out the form for a FREE AI AUDIT CALL + you can see my Claude Code 2nd-brain-OS in action that runs on 9 commands and one main notion dashboard.

Fill out my qualifications form

After this call you’ll have the following:

→ A clear picture of which parts of your business are ready for AI

→ The 2–3 highest-impact workflows to automate first

→ A high-level overview of how your Claude Code second brain should operate.

→ A concrete next step

I spent years learning things the hard way.

And forgetting them.

Every lesson. Every mistake. Every “I’ll never do that again.”

Gone. By the next month I was making the same mistake with a different name.

My friends and I would sit on the streets of Sarajevo talking about what we were going to build. Great ideas. Real energy.

None of it written down.
None of it captured.
None of it turned into anything real.

That was the pattern. Learn something. Lose it. Repeat.

Your AI does the exact same thing.

You open Claude.
Or ChatGPT.
Or whatever you use.

And the first thing you do is re-explain your entire business.

Who you are.
What you sell.
Who your clients are.
How you talk.

Every session starts from zero.

It’s like hiring someone brilliant who gets amnesia every night.

This drove me crazy until I found what Andrej Karpathy built.

Karpathy is the guy who led AI at Tesla. In April 2026, he shared something that clicked for me immediately.

He built a self-maintaining knowledge base. 400,000 words of structured, interlinked knowledge. Zero written manually.

Plain markdown files. No vector database. No embeddings. No complicated infrastructure.

The AI maintains the wiki. Writes it. Links it. Keeps it consistent.

He built it for AI research papers.

You can check his GitHub Gist here

I looked at it and thought: what if I built this for running a business?

So I did.

I call it the compounding brain.

And it runs on three operations.

1. INGEST runs after every task. The brain reads what just happened and extracts the key patterns. Adds them to the relevant knowledge area. Takes 30 seconds.

Example after I update my offer doc & design system:

2. GRADUATE runs for the richest outputs. Deep extraction into standalone notes with full cross-references. The insights are worth their own file.

Example: I run command manually /graduate acquisition area (but it also works in any other business area) and get asked which dates you want to graduate (I type "April 13 works best").

This way I keep only what I remember that works well, and what does not – I don’t want my 2nd brain to have everything (you can schedule this task as well)

3. LINT runs weekly. Scores every area of the brain for health. Flags what’s stale. Tells me exactly what I haven’t learnt yet.

This week I was lazy and didn’t update my brain, and here is what I got:

Sorry, Claude, I was busy living my life.

Side note: I help my friend Jasmin Alic to organize his LinkUp event + clients’ work (that is my excuse, so I have to update my brain health for this week).

You can see what happened here (which is why I was off from this newsletter as well for almost 13 days - sorry guys <3).

We are still humans – right?

However, the advantage of this approach is that it holds me accountable for updating my second brain. And it is funny, like I am playing the video game LoL.

The cycle compounds.

Day 1 the brain knows nothing. Day 30 it knows my patterns. Day 90 it knows my business better than any tool I’ve paid for.

To keep it simple here is the perfect (at some point funny) analogy for the approach above:

Your brain is like a kitchen.

INGEST is washing dishes after every meal so nothing piles up (30 seconds, done).

GRADUATE is when you finally nail grandma's soup recipe and write it down in your recipe book, but you also rip out the bad recipes so only the good stuff stays.

LINT is your Sunday fridge check where you toss expired food and write a shopping list for what's missing.

Do all 3 and your kitchen/brain gets cleaner and smarter every week instead of turning into a mess you're scared to open.

The difference between this and what most people do with AI?

RAG re-discovers knowledge from scratch every query. Nothing accumulates. Day 100 is the same as day 1.

A compounding brain accumulates. Every session adds. Every week the system gets smarter.

No database. No infrastructure. Just folders and markdown files on my laptop.

The setup takes 5 minutes. The compound effect takes 90 days.

But it starts the moment you create that first folder.

That daydreamer on the streets of Sarajevo had the right ideas. He just never wrote them down.

Now the system does it for him.

I call the Karpathy wiki approach my /brain folder, and it gets updated across the following:

• index.md (the hub)

• 7 area MOCs

• queue.md/log

To your next second brain,

Sabahudin “the brain remembers” Murtic

P.S.

What’s the one thing your AI keeps forgetting that drives you crazy? Hit reply. I read every one.

Want to work with me?

Fill out the form for a FREE AI AUDIT CALL (30 min) and see if I can genuinely help you:

Fill out my qualifications form

This is a focused 30-minute call where I map your workflows, tools, and daily reality – then show you where AI would actually make a difference right now.

After this call you’ll have the following:
→ A clear picture of which parts of your business are ready for AI
→ The 2–3 highest-impact workflows to automate first
→ A high-level overview of how your second brain should operate.
→ A concrete next step - not a vague plan, an actual action

On top of that, I am always for more creating than consuming - that’s the best way to build your business and actual dream life/freedom.

I don’t force you to post content or create, but if you do it, you will learn fast, attract the right people, and get a lot of opportunities. And if it is cringe? Who cares? People forget what they eat for breakfast today, so no worries.

But to actually track what to post, get inspiration (not copy) and then add your POV into your content, well that is what we will cover today.

I mean, most of people have usual problems when it comes to the content:
- They check what is trending.
- They open a blank doc.
- They write something.
- They move on to the next platform and start over.

The problem is not effort.

It is that every channel gets treated as a separate project instead of one idea built once and distributed everywhere.

That is what we will solve in today's edition.

Subscribe now

Here is my process from 1 winning idea to 10 repurposed content files …

Before anything, I connect my Trigify MCP to my Claude code so I can track winning topics

(No affiliate link here – it is a good tool thats’s it.)

It watches 4 YouTube channels, 2 Twitter searches, and 1 LinkedIn feed. Constantly.

Here is how to set up Trigify MCP and your social searches.

Then it goes through the full content process….

I type the command /content. Claude Code opens the content pipeline.

Then it ask me how you want to approach it

I pick A – autonomous approach.

And from there it starts the magic.

I can let it run completely autonomously, or I can check the outputs per platform and tell him to stop or continue (I prefer to check my outputs because I enjoy rewriting and editing my own content).

PHASE 1 - FIND WINNING IDEAS

Trigify runs the signals and keywords I previously set up.

Three results per source. Freshest first. Every item gets scored on three things: how strong the engagement is, how well it maps to my content pillars, and whether I have a unique angle nobody else is covering.

This run produced 8 ideas.

The best topic: context architecture versus prompt engineering.

Trending hard in March 2026.

Martin Fowler wrote about it.

BrightCoding published on it. Gartner declared 2026 the year of context.

And zero YouTube creators had covered it from the solopreneur angle. That is the gap.

Map the winning idea across every channel

Before touching a single script, Claude maps the winning idea across every channel. What is the YouTube concept? What LinkedIn post type fits? Which newsletter story connects? What is the Substack tease?

5 minutes. Full content calendar potential visible before committing to production.

PHASE 2 — YOUTUBE PIPELINE

Goal? Start with long-form content

You can record yourself once you get idea and repurpose it or you can use what claudes gives you in this phase and repurpose it (which is the option A that i pick once i run content command)

Then…

Six skills run in sequence. Each reads everything the previous one produced.

First the brief. Main outcome for the viewer.

Why this video deserves to exist. What proof or demos make it concrete.

Research pulled from the web to validate the angle.

Then the outline. Six sections. Timing per section. Demo checklist. Visual needs identified before scripting.

Then the script. Hybrid format. Full sentences for the hook and outro. Bullet points for the middle where I speak naturally. Inline cues for every screen recording and diagram.

Then the visuals. One Excalidraw diagram: the four-layer context architecture stack. Built to 16:9 spec, readable at mobile thumbnail size.

Then packaging. Thumbnail concepts. Title options ranked by estimated click-through rate. A Nano Banana prompt ready to generate the image.

Then the video description. SEO-optimized. Timestamps included. Hashtags. One offer mention.

Then the final title locked: “Why Claude Keeps Forgetting Your Business (and the Fix).”

Everything saves to Notion automatically. One record in the YouTube Content Dashboard. All fields populated.

PHASE 3 - REPURPOSE.

It starts with LinkedIn, and ask me which angles you want to go with?

2 posts are written in parallel

Post output example

Each one is pushed to Notion the moment it is done.

The lead magnet: giving away the two-file starter kit that ends the re-prompting loop.

The funny post: writing a CLAUDE.md is basically therapy. Both live in the LinkedIn Content Workspace, status Draft, ready to schedule. No copy-paste. No manual entry. Done.

Then Newsletter

Story angle selected from a personal story library. Three options surfaced. One chosen.

The edition written to a six-part structure. Saved locally and pushed to the Unstuck Newsletter Database.

Ten subject lines generated. One recommended.

A Substack note. Six lines. Pushed to the same Notion record.

A torn-paper collage thumbnail prompt generated for Nano Banana. Also in Notion.

One record. Five fields. All populated.

Conclusion?

One command. One session. YouTube, LinkedIn, newsletter — all from the same winning idea, built once, distributed everywhere.

(You can add Instagram, X, etc.) - for now, I am focusing on these LinkedIn, Substcak, and YouTube platforms to create content.

This is what connecting social listening to content creation actually looks like in practice.

The system finds what is already moving.

It tests the idea against your pillars.

It builds once and distributes to every channel without reinventing the angle each time.

If you want to see how the Trigify setup works, watch the video above.

If you want to build the full system inside your own Claude Code session, book a call below

To your next client,

Sabahudin ‘one command, 5 channels’ Murtic

P.S.

The newsletter you are reading right now was produced inside this exact pipeline but optimised with my POV.

2 ways to work with me:

1) Getting clients through AI-powered social selling on LinkedIn

-> Offer, ideal clients, outreach, content, DMs — all systematised
-> A repeatable system that brings qualified leads every month

⚒️ SETUP: Claude Code + Your Tool Stack + MCP = AI Social Selling OS
♦️ How we work together: DWY = Coaching/Implementation

Book a Free Discovery Call.

2) Fixing what's broken in your business with AI

-> Marketing, sales, onboarding, delivery, retention → wherever you're leaking
-> Real AI workflows built around your actual process

⚒️ SETUP: Relay App or Claude Code or Both
♦️ How we work together: DWY = Consulting & Implementation

Fill out my qualifications form.

Let me ask you something.

How many times have you told yourself:

“I should keep an eye on what competitors are posting.”

“I should track what’s working in my niche.”

“I should build a system for this.”

And then… opened LinkedIn. Scrolled for 40 minutes. Closed the app.

Same result. Zero insight. Full guilt.

Subscribe now

The real problem isn’t information. It’s the system.

Most people don’t have a competitor research problem.

They have a consistency problem.

They check manually when they feel like it.

Which means they check almost never.

And when they do, it’s random. Unstructured. Easy to forget.

Here’s what I had to learn the hard way:

Manual research isn’t research. It’s noise with extra steps.

You don’t need more discipline.

You need a system that runs without you.

So I built one. Inside Claude Code.

I’ll be honest — I’m not a developer.

I installed VS Code. I installed the Claude extension. That’s literally it.

No Python environment setup. No API keys wired through 7 different services. No n8n flows that break every Tuesday.

Just… VS Code. Claude extension. Done.

Then I did something simple.

I asked Claude to research people who are building content around Claude Code on X and LinkedIn.

Not a surface-level search.

A deep analysis. How often they post. What formats they use. What topics get engagement. What’s working. What’s clearly not.

Claude went and pulled it all together.

And then I said: “Turn this into a skill. Save it as a markdown file.”

Boom.

That’s it.

Now I have a reusable research skill that understands exactly what to look for, how to format the output, and what questions to answer — every single time it runs.

What a skill actually means (and why it saves you time)

Here’s the part that changed how I think about Claude Code.

Once a skill exists as a markdown file, Claude reads it automatically when relevant.

I don’t re-explain the context. I don’t paste the same prompt again. I don’t remind it what my competitors’ names are.

It already knows.

The skill holds all of that.

One setup. Infinite reuse.

If you’ve ever spent 10 minutes re-explaining your ICP to an AI tool before every session… you know exactly why this matters.

Then I scheduled it. 10am. Every day.

This is where it gets actually useful.

Every morning at 10:00 AM, Claude automatically runs the research skill.

It checks what my competitors posted on X and LinkedIn.

It pulls the latest content.

Summarizes what’s trending, what formats are getting traction, what topics are showing up repeatedly.

By the time I open my laptop…

I already have a competitive briefing waiting for me.

No manual scrolling. No guesswork. No “I wonder what they’ve been posting lately.”

Just clarity. Every morning.

Here’s how to build it (step by step)

Step 1 — Set up VS Code + Claude

Install VS Code.

Install the Claude Code extension.

(Quick tip: if you’ve had VS Code installed for a while and you’re hitting errors — just uninstall everything and reinstall from scratch. Don’t troubleshoot. Don’t Google. Fresh install fixes it 9 times out of 10.)

Step 2 - Run the research manually first

Open Claude Code and give it a prompt like this:

Research the top creators posting about [your topic / competitors] 
on X and LinkedIn.

For each person, analyze:
- How often they post
- What formats they use (text, carousels, videos, threads)
- What topics get the most engagement
- What their writing style looks like
- What's clearly working vs. falling flat

Give me a full analysis. Be specific. No fluff.

Let it run. Review the output.

Step 3 — Turn it into a skill

Once you’re happy with the output, tell Claude:

Save this research approach as a reusable skill. 
Create a markdown file called research-skill.md 
that I can use to run this same analysis anytime.

Claude will write the skill file for you.

It captures your format, your focus areas, your competitors — everything.

From this point on, Claude reads it automatically. You never start from zero again.

Step 4 — Schedule it

Go to Claude Co-Work and set up a scheduled task.

Set it to run every day at 10:00 AM.

Point it to your research skills.

That’s it.

Every morning, your AI has already done the work.

Step 5 — The output lands in Google Sheets. Automatically.

This is the part that makes it actually usable.

Claude doesn’t just summarize into a chat window you’ll forget about.

It saves everything into a structured xlsx file - synced straight to Google Sheets.

Every morning you get a clean summary table with four columns:

Topic Format EngagementNotes What they posted about: text / carousel / video / thread High / Medium / Low What stood out, what flopped

No digging through chat history. No copy-pasting into a spreadsheet manually.

Just open Google Sheets. Your briefing is already there. Updated. Dated. Ready.

To set this up, add this to your skill file when Claude creates it:

After completing the research, save the output as a structured table 
with these columns: Topic, Format, Engagement, Notes.

Export it to an xlsx file and save it to Google Sheets 
in a sheet called "Daily Competitor Research".

Append new rows each day. Do not overwrite previous entries.

That last line matters.

Append, not overwrite.

Because after 30 days, you’ll have a pattern library. After 90 days, you’ll see things your competitors don’t even know about themselves.

Share

Why this actually matters

You’re not just automating a task.

You’re building an unfair advantage.

While your competitors are manually scrolling and guessing what to post…

You have a daily briefing on exactly what’s working in your space.

Patterns you’d never catch manually.

Trends before they peak.

Formats your audience is already responding to - just not for you yet.

This is the difference between posting on instinct and posting with data.

The bigger shift

Claude Code is slowly replacing the entire stack I used to need.

No n8n. No Make. No Zapier. No separate research tools.

One environment. One interface. Skills that remember. Tasks that run on schedule.

We’re not at “AI does everything” yet.

But we’re close enough that the gap between people who’ve built systems and people who haven’t…

Is getting very wide, very fast.

Build the skill this week.

Schedule the task.

Let it run tomorrow morning.

And see what’s waiting for you when you wake up.

Talk soon, Sabahudin & Aman

P.S. What are you currently doing to track competitors? Manually, automated, or… not at all? Reply and tell me — I’m genuinely curious where people are at with this.

Leave a comment

Thank you.

I merged two Beehiiv publications → Sabahudin’s System and AI Case Study → into one home. Unstuck. Here on Substack. Those of you who came up from both publications and stay (THANK YOU)

A lot of you followed the move. Some of you sent messages saying you actually prefer it here and that last few editions are amazing. That meant more than you think.

One newsletter. One place. Everything I’m learning, building, and testing - right here.

Now let me catch you up on what’s been happening.

What I’ve Been Up To

It’s been a loud few weeks. Here’s the short version with links if you want to go deeper:

→ Let AI handle the soul-crushing tasks so you can do real work

→ How I planned a 30-day challenge using 2 AIs in one conversation

→ The AI presentation tool that actually adapts to YOUR brand

→ 17 free resources to master Claude Skills

→ A letter to myself 2 years ago about LinkedIn

→ Why cold DMs still work (with the receipts)

Podcast

I sat down with Multiplai AI and we went deep on 3 real AI workflows — lead gen, sales call prep, and newsletters — without losing your voice.

Not theory. Actual walkthroughs.

→ Watch: AI That Sells — Automate Lead Gen, Sales Calls & Newsletters Without Losing Your Voice

Collabs with Charlie Hills & Walid Boulanouar

Two pieces I’m proud of. Both with Charlie and one with Walid as well - and both packed with real workflows, not fluff.

Claude Code: Beginner to Advanced - I wrote the intermediate section on running parallel agent workflows for real client delivery. Charlie covered beginner setup, Walid Boulanouar covered the advanced OS-level stuff. Three perspectives, one guide.

→ Read it on Charlie’s Substack

I Fired My Team and Hired Claude Opus 4.6 - The complete guide to Skills, Cowork, and Plugins. We broke down what Skills actually are, how to build your first one, and 6 workflows for content, sales, legal, and visualisation.

→ Read it on Charlie’s Substack

Attend To My LinkedIn Live — This Thursday, March 26

Topic?

How to use AI social selling to consistently find new clients in 1 hour a day.

→ Attend here → it’s FREE

Show up. Bring your questions.

Quick Poll

I’m building a lot with Claude Code right now. And I want to make sure the stuff I share actually helps you.

My Gift To You: I Open-Sourced My Offer Creation System

Here’s the thing.

I spent few days building an offer creation pipeline inside Claude Code. It walks you through everything - from defining your offer to having a complete pitch deck, VSL script, landing page, and DM openers. All in one guided conversation.

Then I thought: why keep this to myself?

So I put it on GitHub. For free.

What it actually does:

You run one command. It asks you about your business, your brand, your audience. Then it builds:

→ Offer Doc — Your full offer blueprint using my framework + Hormozi’s Value Equation

→ HTML Pitch Deck - slides with your brand colors. Opens in any browser. Arrow keys to present.

→ VSL Script - Video sales letter with production cues. Ready to read on camera.

→ Landing Page - Full sales page copy. Markdown or branded HTML.

→ DM Openers - Curiosity-trigger messages organized by lead type and awareness level.

Five deliverables. One conversation. Zero design or copywriting skills needed.

Why this matters:

Most founders and consultants spend weeks stitching together their offer materials. Or they pay thousands for copywriters, designers, and strategists to do it.
This way they not test the actual offer fast - because that’s the whole point.

You will spend days polishing instead of reaching out the people, promote the offers, close deals or refine it and try again → fail, but fail fast → that’s the game.

This does it all in one sitting.

And it adapts to YOUR brand, voice, and business.
It’s not a template you fill in.
It’s a system that thinks with you.
You ship fast

How to install it:

Open your Claude Code in your IDE or your desktop version and paste this link:

```

https://github.com/sabahudin-web/offer-creation

```

That’s it.

Claude detects the plugin and installs it. Run offer-creation and follow the prompts.

Where this is going & how you should improve it:

Here’s what gets me excited. This plugin is one layer. You can build on top of it.

Imagine: offer doc → lead magnet posts → ad copy → funnel pages → newsletter email sequences → launch campaign. All as one agentic workflow. Start to finish.

I’m already building this. An offer-launch pipeline where one command takes you from “I have an idea” to “my launch is live.”

That’s the direction. And this GitHub repo is the first piece.

Leave a comment

→ Don’t forget to test it and give me your feedback: github.com/sabahudin-web/offer-creation

To your next client,

Sabahudin

---

P.S.

What would YOU build on top of this?

If you could add one more step to the pipeline - lead magnets, ads, funnels, emails → what would it be?

Reply and tell me.

I’m building the next layer and your input shapes it.

2 ways to work with me:

1) Getting clients through AI-powered social selling on LinkedIn

-> Offer, ideal clients, outreach, content, DMs — all systematised
-> A repeatable system that brings qualified leads every month

⚒️ SETUP: Claude Code + Your Tool Stack + MCP = AI Social Selling OS
♦️ How we work together: DWY = Coaching/Implementation

Book a Free Discovery Call.

2) Fixing what's broken in your business with AI

-> Marketing, sales, onboarding, delivery, retention → wherever you're leaking
-> Real AI workflows built around your actual process

⚒️ SETUP: Relay App or Claude Code or Both
♦️ How we work together: DWY = Consulting & Implementation

Fill out my qualifications form.

“We should start this.”

“We should build that.”

We never did.

Not because we lacked ideas.

You already know this feeling

Just kept talking.

Same thing in business.

You know your outreach is broken. You say it in calls. In DMs. In your own head at 11pm.

“I should build a real pipeline.”

“I should stop sending generic connection requests.”

“I should have a system for this.”

Months pass. Nothing changes.

Here’s what I had to learn the hard way.

Saying “I should” feels productive. Building the system feels uncomfortable. But it’s fast.

The real problem isn’t the pipeline

Most consultants don’t have a pipeline problem. They have a starting problem.

They imagine it’s complicated. A big technical project. Needs a developer. Needs months of setup.

It’s not.

The mistake I see everywhere: waiting to fully understand the system before building it.

That’s not strategy. That’s fear.

You only understand it by building it.

Here’s what one command looks like

Subscribe now

Here’s what it actually looks like, step-by-step:

1) I take a CSV. Eight LinkedIn leads.

One command: build-qualify-list.

Claude asks which file to qualify. I point to my inputs folder. It starts enriching every lead using BrightData — real data from LinkedIn profiles, company pages, recent activity.

BrightData team approach me to give my readers $20 in credit.

Use code “sabahudin” and add it to the "add coupon" section in your account

(this can be enough to enrich over 1000 leads – so use it.)

8 leads are enriched. Zero skipped. (Bonus Tip – always test on small lists then move on larger batch runs)

2) Then Phase 2. → ICP scoring. Every lead gets scored automatically.

5 not a fit. 3 qualified. No guessing involved.

3) From there, the DM phase starts.

Parallel agents scrape the last LinkedIn post from each of those three leads. Then write a personalized connection note — using their profile, their recent activity, and my DM writing rules.

Not templated. **Actually personalized.**

4) OPTIONAL: Once I review and approve, the contacts push to HubSpot. Already in the CRM? It skips them. No duplicates.

Then the Aimfox campaign gets created automatically from Claude Code. Audience added. You just need to write it {{icebreaker}} as part of the connection note, and it will pull out all personalized notes from the sheet I show you in the video.

Eight leads in. Three qualified. DMs written. CRM updated. Campaign live.

For 100 leads? Fifteen parallel agents run at the same time. It scales.

You don’t.

Download the full pipeline implementation guide as a zip file.

I packaged everything into one ZIP.

The commands. The agents. The skills. The full pipeline — from CSV to campaign.

It also includes an onboarding command that runs first.

9 steps. It asks about your ICP, your offers, your voice, your outreach tool, your CRM. Writes a settings file. From that point on, every command reads your setup and runs for your specific use case - NOT MINE.

No Saba-specific hardcoding. No locked integrations. Aimfox or any other outreach tool. HubSpot or any other CRM. BrightData for enrichment. Everything is configurable.

→ Download the ZIP

Install it in Claude Desktop or Claude Code.

Run /onboard first. That’s it.

What am I building next?

Share Unstuck

This is my next addition:

• Someone books a call with me.

• BrightData fetches their LinkedIn profile, company page, recent posts, and website.

• Combines it with my sales script. Gives me a full research brief before the meeting.

All from one command.

Ten years of “we should.” One command to replace all of it.

If you are curious about how to build an AI-powered data enrichment agent in 10 minutes using LangGraph + Bright Data

Check this guide as well

I am preparing a 30-day Claude Code challenge in which I will cover how to build your Claude Code OS as your second brain AI Social Selling system (stay tuned) - would love to hear your response if you will follow my Claude Code challenge along the way. :)

To your next client,

Sabahudin ‘pipeline, not promises’ Murtic

P.S.

How many hours a week are you spending on outreach this could replace? Reply and tell me - I’m curious what your current setup looks like.

2 ways to work with me:

1) Getting clients through AI-powered social selling on LinkedIn

-> Offer, ideal clients, outreach, content, DMs — all systematised
-> A repeatable system that brings qualified leads every month

⚒️ SETUP: Claude Code + Your Tool Stack + MCP = AI Social Selling OS
♦️ How we work together: DWY = Coaching/Implementation

Book a Free Discovery Call.

2) Fixing what’s broken in your business with AI

-> Marketing, sales, onboarding, delivery, retention → wherever you’re leaking
-> Real AI workflows built around your actual process

⚒️ SETUP: Relay App or Claude Code or Both
♦️ How we work together: DWY = Consulting & Implementation

Fill out my qualifications form.

Hey,

Life can be busy, and if you missed some of my previous editions, here are the links:

1. Run better sales calls with this AI workflow

2. Filter your 5000 LinkedIn connections and find your target audience with Claude ICP skill and Excel

3. Full Claude Skills & Co-Work setup (Collaboration Newsletter with Charlie Hills)

My free guides for you:

1. Create a personalized lead magnet Claude plugin

2. Create your Lead Magnets from scratch

3. Paste this prompt into the Claude and ask him to create infographic skill

4. Use these custom instructions for any LLM to challenge you rather than to please you.

Now, let’s cover our main topic for today….Scheduled Tasks in Claude Co-Work and Claude Code

I read something this week that made me smile like a nerd.

Claude Code & Co-Work is now a 24/7 employee.

Which means…

It’s always Claude Code o’clock.

Yeah… that might be the nerdiest thing I’ve written in a while.

But the update behind it is actually a big shift for anyone building AI workflows.

Anthropic just released scheduled tasks inside Claude Code & Co-Work.

Which means your AI agents can now work on a schedule… without you touching anything.

In other words, if you learnt a skill in n8n or Make, you can forget about it because we are entering a new phase of development in which you can do everything from a single chat interface.

(hopefully they will not yet launch:“run your AI agents with smart sunglasses” haha.)

What's the difference between Claude Co-Work and Code?

Even though I'm not a technical person, I enjoy using Claude Code in the VS Code environment because Co-Work is still too slow, annoying, and full of bugs. (But I'm sure it'll get better soon)

Now, with scheduled tasks, it's pretty amazing.

The only current limitation of Claude Code compared to Co-Work is that scheduled tasks cannot be built directly into Claude Code; however, they can be built in Co-Work and used from Claude Code.

(I know, I know… this sounds like some joke, but hey, they are improving it every day)

In next editions, I'll write more guides about Claude Code's full setup and environment. - which you can see in the image below….

But for now, let’s look at what can be done as a scheduled task.

Your AI can now work while you sleep

Even if this subtitle sounds like a clickbait sh…t, we are not that far away from that “AI dream” LoL.

Back to the main topic, until now, using Claude Code looked like this:

You open the app → ask it to run something → watch it execute.

Helpful.

But still manual.

Now you can schedule things like:

• Every morning at 9:00 AM
• Every hour
• Every day
• Every week

Claude automatically starts a new session and runs the task.

Think about that for a second.

Instead of asking an AI assistant to do something…

You now hire it to do it every day.

Example: email triage every morning

One simple automation I love is Email Triage.

Imagine this.

Every morning at 9:00 AM Claude:

• checks your inbox
• reviews emails from the last 24 hours
• summarizes important conversations
• highlights urgent items
• suggests replies

So when you open your laptop…

You already know what matters today.

No inbox chaos. Just clarity.

+ You can run it from either Claude Co-Work or Claude Code (it doesn't matter).

Basic example with Co-Work in action:

Prompt example for the task

Here’s a simple prompt you could use for a scheduled task called email-triage:

You are my executive assistant.

Every time this task runs, analyze my emails from the last 24 hours.

Your goal is to help me quickly understand what matters without reading everything.

Steps:

1. Retrieve emails received in the last 24 hours.
2. Categorize them into these sections:

- Urgent / Requires Response
- Important but Not Urgent
- Newsletters / Low Priority
- Spam or irrelevant

3. Provide a short summary of the most important conversations.

4. Highlight any emails that require my reply today.

5. For urgent emails, suggest a short draft response I could send.

Output format:

Daily Email Briefing

1. Urgent emails
- sender
- subject
- one sentence summary
- suggested reply

2. Important but not urgent
- sender
- subject
- short summary

3. Newsletters / low priority
- bullet list

4. Recommended actions for today
- top 3 priorities from the inbox

Be concise and prioritize clarity.

Then schedule it to run every morning at 9:00 AM.

You need only to connect your Gmail in Claude directly and you will have access in Co-Work and Claude Code as well.

Of course, if you want to go more specific, create a skill for this task using official anthropic skill builder and…

Boom.

Your inbox gets pre-processed before you start your day.

+ You dont pay anymore for some email tools (and I am afraid, you will start canceling more and more tools, and build yours, internally with Claude)

Share

Why this update matters

The real shift here isn’t scheduling.

It’s agent workflows.

Traditional automation:

Step 1 → Step 2 → Step 3.

If something breaks?

The system crashes.

AI agents work differently.

If Claude runs into an issue, it might say the following:

“That didn’t work. Let me try another approach.”

It can adapt.

Test alternatives.

Fix errors.

Your workflows become self-healing systems and fix, autonomously, any mistake.

Subscribe now

A few things to know

There are some early limitations when it comes to scheduling these tasks.

1️⃣ Your computer must be on

Scheduled tasks currently run in the desktop app.

If your laptop is off, the task won’t run.

2️⃣ Every run starts fresh

Each scheduled task launches in a new session.

So if you want memory, store it in files.

3️⃣ Permissions matter

Because tasks run automatically, you should control things like:

• deleting files
• running risky commands
• making major changes to projects

But again, you can always use official Anthropic guides, such as best practices and improve your overall experience with Claude.

Pro AI bro tip: Just paste any official guide from Anthropic and tell Claude to follow that in your specific task, workflow, action, etc. (this way you are an AI expert LoL)

The bigger shift

Claude Code is moving from:

AI assistant

→ to

AI employee

Something that can:

• run workflows
• monitor projects
• fix issues
• analyze information
• operate on a schedule

Without you prompting it.

And once you start stacking these tasks…

You essentially build your own AI operations team.

I’m curious what people start building with this.

Because the future of AI isn’t:

“Ask me anything.”

It’s:

“I already handled that for you.”

Talk soon,
Sabahudin

Share Unstuck

When I remember my childhood, I never collected football stickers.

If I had the best player card, I sold it.

Same with Yu-Gi-Oh decks. My best ones left my hands.

Everyone else had binders full of cards. Stacks. Collections.

I had money and a reputation.

Kids knew I had what they needed. Not because I had everything. Because I only kept what mattered.

Same thing on LinkedIn.

You think you have 5,000 opportunities.

In reality, you probably have 100 to 200 people who are truly your best fit.

The rest are dead weight. They drag down your reach. They waste your time. They burn your enrichment credits.

Subscribe now

Here is what most ppl get wrong.

They treat their network like a trophy collection.

5,000 connections looks nice.

Your pipeline does not care about connection count.

It cares about fit.

If your list is messy, you will waste time. You will waste money. You will DM people who will never buy.

You will enrich everyone and convert no one.

Here is the exact workflow I used to turn 5,000 connections into 370 ICPs and 120 Tier A best-fit leads.

Step 1: Export your LinkedIn connections

Go to Settings, Data privacy, Get a copy of your data; request archive.

LinkedIn sends the file in three to four hours.

Step 2: Import into Google Sheets, download as Excel

Sheets handles the import. Excel is where you run the AI.

If you are a Mac user, download free Excel here

Step 3: Use Claude inside Excel to classify roles

Install the Claude Excel extension. Connect your account.

Use the Position or Headline column to label ICP versus not ICP.

Step 4: Tier the ICP list

Create your own ICP Claude skill in the Claude environment.

Download this skill, then recreate it to your specific ICP criteria.

Ask Claude in the Excel environment to split ICPs into Tier A for best fit and Tier B for second best.

Create a new sheet for Tier A only.

Step 5: Build a simple dashboard

Count roles inside Tier A. Coaches versus consultants. Track your micro ICP.

Here is what happened in a backend once I start running this Claude skill + Excel (FASTER VERSION)

Bonus rule: Enrich after you filter.

Clay, Apollo, Relay, whatever you use.

Enrich only Tier A so you do not burn credits on people you will never contact.

I went from 5,000 random connections to 120 people I want to talk to.

Filter first. Then act.

If you want a simple system to find best-fit buyers without burning hours or tool credits, I am launching my paid community soon.

Weekly implementation calls. Real workflows. No theory.

Reply with LOCKED if you want in.

To your next client,

Sabahudin ‘filter first’ Murtic

These are 2 ways we can work together:

1. I run an agency for LinkedIn outreach & cold calling (best fit: Cybersec & SaaS companies) → Apply Here

2. Book an AI audit Power Hour with me - ideal for Coaches, Consultants, Solopreneurs, and SMB owners.
   

Hey there,

I write deep newsletter editions every week.

Real systems. Real examples. Real results.

But here's the thing that kept me up last night.

They're invisible to AI.

The brutal realization

Last week, I asked ChatGPT: "Who teaches LinkedIn monetization for B2B founders?"

My name didn't show up.

Not once.

I've written 104 newsletter editions. Thousands of words on LinkedIn systems, AI-powered outreach, and revenue generation.

All sitting on Beehiiv.

Great for inbox trust.

Useless for discovery.

Because LLMs don't crawl newsletters.

They crawl websites.

And my Beehiiv subdomain? Not optimized for AI at all.

So while I'm helping 30+ clients book qualified calls and hit $10K months, my entire content library is doing zero work for inbound discovery.

37% of your prospects now start with AI, not Google

When someone asks ChatGPT for a LinkedIn agency recommendation, there's no second page.

You're either the answer or you're invisible.

I've been invisible.

For two years.

The fix (without changing how I write)

I'm not rewriting anything.

I'm not doing SEO gymnastics.

I'm not hiring a content team.

Here's what I'm doing instead:

1. Move my newsletter to my own domain

2. Keep the newsletter tab for subscribers

3. Add a blog section

4. Repurpose each edition into an AI-readable blog post

5. Structure it so LLMs can actually cite it

Same ideas. Same voice. Same value.

Different surface area.

Now every edition does two jobs:

Inbox = trust + subscribers
Blog = AI visibility + inbound traffic

That's how newsletters start compounding.

The system I'm using

I found outwrite.ai after seeing Danish Lead Co. close their first ChatGPT-sourced client in 17 days.

Zero blog presence to paying customer.

Because they published 7 LLM-optimized articles that AI systems could confidently cite.

The platform does three things I can't do manually:

1. Turns existing newsletter content into AI-readable blog posts

2. Uses proven AEO structure (not guesswork)

3. Publishes directly to my site with one click

I recorded a 1-minute screen recording showing the full process:
From signup to finished blog post using an existing newsletter edition.

No rewriting. No complexity. No change to how I write.

Why this matters for you

If you write newsletters, you're probably in the same boat.

You create value weekly.

Your readers appreciate it.

But AI has no idea you exist.

So when your ideal client asks ChatGPT for a recommendation in your category, you're not even in the conversation.

That's leaving money on the table.

The difference between being cited and being ignored isn't quality.

It's structure.

AI needs:

• Answer fragments it can extract

• Clear Q&A blocks

• Schema markup

• Citation-ready content

Your newsletter probably has the insights.

It just needs the right packaging.

My 30-day experiment

Starting today, I'm tracking:

• AI citations across ChatGPT, Perplexity, and Gemini

• Specific prompts that mention my brand

• Inbound traffic from AI-sourced searches

I'll share the results in 30 days.

Either this works and I'll show you the exact system.

Or it doesn't and I'll tell you what I learned.

But doing nothing while 37% of buyers start with AI?

That's not an option.

Want to join me?

I put together the full outwrite.ai playbook I'm using.

It's the same process Danish Lead Co. used to go from zero to ChatGPT-sourced clients in 17 days.

Inside you'll get:

• Visibility audit process

• Content transformation workflow

• AI citation tracking setup

• 30-day publishing strategy

Get the full playbook here

If you write valuable content but feel invisible when prospects ask AI for help, this is for you.

Power to you,
Sabahudin "AI Visible" Murtic

I spent 10 years saying one sentence.

"We should start something."

It sounded smart.
It felt like progress.

But nothing changed. Because talking is free.

Same thing with lead research.

You open 47 tabs.

LinkedIn profile. Company website. Last 10 posts. Competitors. Google Doc.

You copy. You paste. You guess their pain. You write talking points.

Three hours later, you walk into the call with messy notes and zero confidence.

That's not research. That's busywork.

Here's what most ppl get wrong:

They're not losing deals because they're bad at sales.

They're losing deals because they walk into calls with zero context.

So they compensate.

They pitch more. They talk more. They guess the pain.

That's not confidence. That's unprepared.

And the worst part?

You burn 3 hours doing "research" and it's still messy.

Here's the simple automation workflow system:

1. Capture the right inputs in your form
   Problem, goal in 90 days, budget readiness, LinkedIn and website.

2. Auto-collect context
   Normalize the website URL, scrape the site, pull last 3 LinkedIn posts.

3. Run AI research on the data
   Summary, likely pains, objections, angles to mention on the call.

4. Merge it into your sales script template
   The output is a personalized call plan plus your offer doc. Not random text.

5. Send it where you live
   Notion plus email plus create or update the contact in your CRM. Avoid duplicates.

That's it.

90 seconds. What used to take half a day.

I tested this with my friend. Here's what happened:

The workflow scraped his website. Pulled his last 3 LinkedIn posts. Did the research.

Then it created a personalized 12-step sales script.

Mentioned his podcast. Called out the exact pain he selected in the form. Asked questions based on his inputs.

Not generic. Personalized.

All automated. I received the full report via email and on Notion, and I updated my CRM in HubSpot.

Here is the short preview of that output :

You don't need more people. You need one workflow.

Build it once. Use it forever.

How does 1:1 coaching to help you and your team to use AI in your business sound to you?

I'm offering 3 spots for 1:1 coaching and implementation calls. These sessions will help you save time, boost your revenue, and focus on what truly matters for your business.

We will:

• Audit your current business process to see main priority that can be improved with AI workflows

• Find the gaps where you loosing the time, money and focus.

• Came up with the first automation that we can build together.

This is a done-with-you offer where, right in front of you, we build a live automation together that solves the problem we previously identified.

You will leave with:

• AI implementation knowledge,

• a fully implemented workflow, and

• clarity on the direction your business can take.

SIDE NOTE: your team members are welcome; you can pay for one person and invite whole team for the same price

If you're interested, reply with "AUDIT" and we can schedule a discovery call to discuss details and pricing for your specific needs and situation.

If not, no pressure at all.

Meanwhile, I am always generous with my audience → grab the automation template from this edition here and try to implement for yourself

Bonus tip: If you record your calls, take the best ones you close and ask any LLM to analyze the transcripts to provide you with a sales script template optimized for your specific needs. Then, just import it into the workflow I just shared with you.

To your next client,

Sabahudin 'ship it' Murtic

Before you get ICP skill (which will be free - you have a link at the end of this edition), here are the results of booked calls and payments we received from Nov to Jan:

Booked calls in November:

Payments from Nov to Jan:

Now how do we get there?

We used to burn through 3,000 leads in two months.

Reply rate? 2%.

Booked calls? Maybe 5.

The problem wasn't my outreach.

It was my list.

Half were B2C companies. Quarter were non-English profiles. The rest? People who hadn't touched LinkedIn in months.

I was paying Clay credits to enrich garbage.

Paying Lemlist to send emails that bounced.

Spending hours crafting personalized messages to people who would never buy.

Why This Happened

We were building lists backwards.

Sales Navigator search: "CEO" + "Company size 1-10,000" + "Worldwide"

Result? 5 million profiles.

Then we'd try to narrow it down manually.

Scroll. Click. Read. Disqualify.

5-10 minutes per lead.

70% rejection rate.

It was like panning for gold in a sewage plant.

How It Got Worse

The more leads we burned, the harder it became to find new ones.

Our outreach reputation tanked.

Sales Nav started limiting our searches.

And here's the killer:

We started doubting our offer.

"Maybe our service isn't valuable?"

"Maybe LinkedIn growth doesn't work for these people?"

Wrong.

The restaurant owner I pitched doesn't need LinkedIn strategy.

The fashion brand founder doesn't have customers on LinkedIn.

No amount of "better messaging" would fix that.

The Claude ICP Skill as a Solution

I built a qualification system directly in Claude.

Not a generic prompt. A proper skill.

Here's exactly what it does:

Step 1: ICP Discovery

I feed Claude our business context:

• What we do (LinkedIn growth for B2B founders)

• How we deliver (90-day done-with-you system)

• What we charge (€10K single profile, €30K for 3-profile teams)

Claude builds five detailed personas:

• Scaling B2B Founder (10-250 employees)

• Growth-Stage CMO/VP Marketing

• Expert Consultant/Coach

• Agency Owner

• Sales Leader

Each persona includes:

• Demographics

• Emotional drivers (fears + desires)

• Intent signals

• Where to find them

• Qualifying questions

Step 2: Three-Stage Qualification

Then Claude creates the filtering system:

Stage 1: Algorithmic Rules (instant)

• Company size check

• Geography/language filter

• Industry exclusions (B2C, healthcare, retail)

• Role exclusions (students, job seekers)

Stage 2: Quick Screening (2 seconds per lead)

• Role match verification

• B2B business model check

• Decision authority confirmation

Stage 3: Deep Qualification (5 seconds per lead)

• Full context analysis

• Multi-profile potential scoring

• Lead categorization (High Priority vs Qualified vs Disqualified)

Step 3: Direct CSV Processing

I upload my lead list directly into Claude.

76 leads in → Processed in 45 seconds.

Output:

• 20 qualified (26% pass rate)

• 6 HIGH PRIORITY multi-profile opportunities

• 56 disqualified with reasons

Step 4: Outreach Angle Generation

For the 6 high-priority leads, Claude researches:

• Recent company activity

• Funding announcements

• LinkedIn presence gaps

• Trigger events

Then writes personalized outreach angles for each.

No generic "I saw your post" templates.

Actual business context that shows we did our homework.

The Results

Last 4 months using this system:

100+ qualified sales calls booked

Average list quality:

• 26% qualification rate (vs 5% before)

• 30% of qualified leads = high-priority multi-profile deals

• 15% reply rate (vs 2% before)

Time saved:

• Manual review: 8 hours per 100 leads

• Claude processing: 5 minutes per 100 leads

• 95% time reduction

Bonus tip: if you use the Claude extension, you can do relevance check of the list directly from Claude code (insane)

Money saved:

• No more enriching garbage leads

• No more burning domains on bad outreach

• ROI on closed deals: 15x

Here's what changed:

We stopped chasing volume.

Started chasing precision.

500 perfect-fit leads beats 5,000 random CEOs.

Every. Single. Time.

What This Means

You don't need more leads.

You need better leads.

And you don't need a VA spending 40 hours reviewing profiles.

You need a system that:

1. Knows your ICP inside out

2. Filters algorithmically (instant elimination)

3. Qualifies intelligently (AI-powered context analysis)

4. Prioritizes strategically (high-value opportunities first)

5. Personalizes automatically (research-backed outreach angles)

That's what this Claude skill does.

I've packaged the entire system into a guide you can implement yourself.

Includes:

• Full ICP discovery framework

• Three-stage qualification prompts

• CSV processing workflow

• Outreach angle generation template

• Sales Navigator search formulas

You can install your skill below:

Grab the full guide and skill implementation

Let me know if you are ready for my community launch in February. If yes, reply "Ready." I want to see who is really excited and who is "meh" (you can reply "meh" as well, haha).

Power to you,

Sabahudin "precision over volume" Murtic

P.S.

The biggest mistake? Blaming your messaging when your list is broken. Fix the list first. Everything else gets easier.

In a high school, a classmate leans over my desk.

"Will you draw this architectural project for me?"

(yep, I was in architectural high school.)

an up-and-coming architect who never made it haha

When I get that question from a classmate, in that moment – I'm not thinking about "design".

I'm thinking: make it clean. Make it clear. Make it fast.

I sketch out his requirements.
Turn messy notes into something visual.
A floor plan anyone understands at a glance.

He pays me. Word spreads.

More classmates show up with projects.

(I was social seller in my heart.)

I used the same process every time: take what they need, translate it into something clear, deliver fast.

This is not about talent.

It’s about having a repeatable way to turn words into visuals.

I get this infographic with the guide I will share at the end of this edition

It’s the same thing with your content…

Your best content is already written.

LinkedIn posts. DM conversations. Call notes. Transcripts.

But if people need more than 3 seconds to see the point, you lose them.

The reality is most people don't have a content problem.

They have a translation problem.

They write good posts and leave them as walls of text.

Or they try to make "one perfect infographic" and get stuck for 2 weeks.

The truth: consistency comes from having a simple system keeping you on-brand every time.

Here's the system I'm using now:

1. Take one LinkedIn post/transcript you already have.

2. Run it through my infographic prompt generator.

3. Open a new chat and paste the prompt into NanoBanana.

OUTCOME: You'll get an infographic that shows your ideas using your brand's colors and fonts.

Here is the infographic layout logic…

If you're talking about before vs after, it becomes a two-column layout.

If you're listing 3 things, it structures them as a vertical flow.

And so on…

The prompt does the translation work for you.

No designer needed. No 2-week delays. No fancy tools.

I used this last week to turn a post about generic LinkedIn DMs into an infographic showing bad vs good personalization.

Side by side, on-brand, done in minutes.

90% match to my brand colors. 100% match.

Same thing I did with those architectural drawings in high school: take messy input, turn it into something visual anyone understands fast.

You can go even further….

Why stop only at infographic style?

You can try the visual analogy style.

Here are few examples I get

I tweak my infographic prompt into this style (pretty epic results)

As my favorite reader, you know I always have a complete guide prepared for you, right?

Well, the truth is, I have…

I made an easy guide on how to create a Gem that generates prompts for infographics.

Want to see exactly how this works?

Watch the full breakdown with implementation here

If you want more simple workflows like this, real systems helping you ship content faster without losing your brand, we build them inside Intersection AI.

Weekly implementation calls. Real workflows. No theory.

Reply with 'I'm in' and include your LinkedIn profile URL.

I'll add you to my early list. The community will launch soon.

To your next client,

Sabahudin 'make it visual' Murtic

I spent 10 years talking instead of doing

I was the guy with big ideas.

Big plans.

I’d talk for 30 minutes about what we need to build. What we need to do.

Then I’d go home.

Nothing changed.

Same thing happens on sales calls.

Most consultants don’t lose deals because their offer is bad.

They lose because they talk too much.

They walk into discovery calls cold. No prep. No plan.

So they fill the silence with talking.

They pitch too early.

They don’t ask the questions to expose the real problem.

After the call? Stuck.

No clean notes. No clear next step. Vibes only.

Here’s what changed for me.

I stopped talking. Started preparing.

Not with scripts. With better questions.

Subscribe now

I built a simple system for every call:

Step 1:

Form submission captures their LinkedIn URL (I use Tally)

Step 2:

Scrape their LinkedIn and pull role, company, location, keywords (Relay has this built in)

Step 3:

AI step combines their LinkedIn data, your offer doc, and your sales script template. Then generates hyper-personalized call prep with specific questions.

Step 4:

Save to Notion as one page per prospect, ready before the call

I don’t read the script on the call.

I use this to ask better questions. And listen.

Here’s what this looks like in practice.

Let’s say Jenny books a call. She runs Trackfinity and Rock Ahead. Busy between both companies.

The automation scrapes her profile. Then generates questions like:

“How did you become known as ‘the tracking girl’ in the industry?”

“How are you structuring your affiliate programs right now?”

These questions are tailored to her. Not generic. Not something I’d ask everyone.

After the call?

I have two options:

1. Use the transcript to send an automated proposal

2. Use the transcript to create a content strategy roadmap for when we start working together

Every call is unique.

This gives me the structure to ask the right questions.

To understand where they are now. Where they need to be. What’s stopping them.

You’re not reading from a script.

You’re preparing so you listen.

Want to see exactly how this works? Watch the full breakdown:

If you want more simple workflows like this, real systems not theory, we build these inside Intersection AI.

Reply if you are interested to join I will add you to my list

To your next client,

Sabahudin ‘ask better’ Murtic

I'm staring at my revenue dashboard.

Best year ever. Numbers I never thought I'd hit.

And I felt... confused…

Not because the money didn't matter.

But because I knew the truth:

I'd scattered my focus across 5 different projects, said yes to clients that didn't fit, chased every new AI tool like it was the answer, and burned myself out pretending "busy" meant "winning."

Here's what actually happened:

I used to start 3 automation systems. Finished zero.

Launched ideas I never validated.

Posted on LinkedIn without strategy (just "consistency").

Told myself I was "exploring opportunities."

Bullshit.

I was scattered. And I paid for it with my energy, my focus, and my sanity.

The Wake-Up Call:

What you do TODAY determines what you win in 5-10 years.

Not what you plan.

Not what you talk about.

What you actually EXECUTE.

And I wasn't executing. I was surviving.

So here's what I'm doing differently in 2026:

Going back to basics. Not the motivational kind. The ruthless kind.

ONE focus. ONE system. ONE daily win.

That's it.

My 2026 Framework:

1. Daily Wins = Long-Term Freedom

Every day has ONE priority that moves the needle.

Not 10 tasks. ONE.

And I'm making it binary so I can't BS myself:

A daily win = Booking 1 qualified consultation call OR shipping 1 complete workflow.

Yes or no. No gray area.

I'll track it publicly every week in this newsletter.

2. ONE Target. ONE Problem. ONE Solution.

I'm done trying to help everyone.

I help B2B consultants & founders who have proven offers but inconsistent pipelines.

People who know their stuff but can't fill their calendar predictably.

So I focus more on common problem rather the role or industry (or those basics ICP criteria)

How?

Intent-based prospecting + value-first selling + AI automation that actually saves time = one social selling ecosystem

That's it.

3. ONE Core Offer (For Now)

Starting Q1 2026, I'm leading with ONE thing:

My community - Intersection AI.

Weekly implementation calls. Complete workflows. Real systems you build with me in real time.

Not a course library. A working lab.

Once that's proven and running smooth, I'll add 1:1 coaching and DFY services in Q2.

But right now? ONE focus - launch my private paid community

4. Build In Public. Ship Fast. No Perfectionism.

I'm shipping real workflows every week.

Sharing what works and what bombs.

Building in public so you see the process, not just the polish.

positioning one-pager

What This Means For You (And This Newsletter):

Starting today, this is Sabahudin's Systems.

Here's what you can expect:

ONE edition per week for the first 2 months.

Not 3. Not daily. ONE.

Deep value. Practical systems. Real workflows you can implement immediately.

No fluff. No recycled motivation. Just execution.

Real Resources. Real Implementation.

Every edition includes something you can use TODAY:

Templates. Workflows. Frameworks. The stuff I'm actually using in my business.

My 2026 Build In Public:

You'll see exactly how I'm building:

• Intent-based prospecting (8-step Top 200 framework)

• Value-first outreach that gets responses

• AI automation that saves 10 hours/week

• Content systems that convert

And I'll track my daily wins publicly every week.

The wins. The misses. The honest truth.

Here's What I Need From You:

I'm not asking you to buy anything.

I'm asking you one question:

Are you in?

Not "in" like you'll passively read emails.

"In" like you're ready to focus in 2026.

"In" like you're tired of scattered strategies.

"In" like you're willing to win your days instead of surviving them.

If that's you, reply with one word: LOCKED

I just want to know who's actually here.

Who's ready to execute, not just explore.

Not everyone will reply. That's fine.

But if you do, you'll get everything I build this year.

First resource drops next week…

No opt-in walls. No hidden costs. Just value.

Let's make 2026 the year we stop talking and start winning.

One day at a time.

Power to you,

Sabahudin "daily wins" Murtic

P.S.

I almost committed to 3 newsletters per week in this email. Then I caught myself repeating the same pattern that burned me out last year. So I'm starting with one. Proving I can sustain it. Then expanding if it makes sense. That's focus. Not more output pretending to be more impact.

Hey legend, happy holidays…

It is the end of the year, and I hope you have an amazing year.

I'm sitting in my home office in Sarajevo, staring at my Q4 revenue dashboard.

Best year ever.
30+ clients helped.
First consistent 10K+ months.

And I felt... empty.

Because I spent 8 months solving the wrong problem.

Let me explain.

In January 2025, I convinced myself the problem was systems.

I needed:

• Better automation workflows

• More sophisticated funnels

• Fancy AI integrations

• Complex multi-touch sequences

So I built them.

Relay workflows.
N8n automations.
Multi-step nurture sequences.

And you know what happened?

I never owned the one thing I should be best at.

I clearly explain the main problmessaaing fix.

Meanwhile, my buddy Ahmed was crushing it with:

• Plain text emails

• Simple Google Docs

• Basic follow-up sequences

Making MORE than me.

With LESS complexity.

I realized…

The problem was never the systems.

The problem was I didn't know what to say.

I had automation architecture. But I didn't have message clarity.

I could build workflows. But I couldn't articulate transformation.

I had tools. But I didn't have truth.

So I spent Q4 completely differently.

Instead of building more systems, I studied what made people say yes.

I analyzed 50+ successful client conversations.

I tracked which messages got responses.

I documented which stories created trust.

And I found the pattern.

The clients who said yes fastest weren't convinced by:

• My technical capabilities

• My automation expertise

• My LinkedIn following

They said yes because I made them feel SEEN.

I articulated their specific frustration before they could.

I named the exact problem they were losing sleep over.

I showed them I understood the gap between where they are and where they want to be.

That's not a systems problem.

That's a message problem.

And once I fixed the message, everything else accelerated.

Same automation tools. Same simple workflows.

But now they converted because people actually resonated with what I was saying.

Here's what changed in Q4:

The lesson isn't "systems don't matter."

The lesson is: systems amplify your message.

If your message is unclear, automation just scales confusion faster.

But when your message is dialed in?

Then automation becomes rocket fuel.

That's what I'm focused on in 2026.

Not building more complex systems.

Building crystal-clear messages that make the right people immediately understand: "This person gets exactly what I'm dealing with."

And then letting simple automation do its job.

If you're a founder or tech leader who's been building systems but not getting the conversion rates you want...

The problem might not be your automation.

It might be your message.

Happy new year with a small gift for you.

I put most of the important stuff that helps me to grow this year into the one compelling guide.

Grab it here

Since I recap 2025, I would also add 2026 plans:

→ Building community in public - with you
→ Value first in outreach & content - for you
→ Earning trust and keeping it - because of you

Power to you,

Sabahudin "Value First" Murtic

P.S.

The founders who get this right in Q1 won't need to chase leads for the rest of the year. Their message does the heavy lifting while automation handles the repetition.

That's the difference between working harder and working smarter.