Co-Founder @Kodebusters - we bring Products to Market 5x faster and for a fraction of the cost

I help brands bring their customer experience into ChatGPT.

As more customers search, compare, and decide inside AI tools, brands risk losing control over how their products and services are represented. The best answer wins, and most brands aren’t shaping it yet.

At Kodebusters, we help teams extend their existing sales, booking, and advisory experience into ChatGPT - so customers can discover, choose, and convert with confidence, without changing core systems or workflows.

Before Kodebusters, I spent over a decade in product leadership roles at Mercedes-Benz, Amazon, Delivery Hero, and UPPER, building and scaling products used by millions.

Today, I work with founders, operators, and enterprise teams exploring how AI-driven discovery impacts customer journeys through focused advisory, pilots, and custom implementations.

If you’re thinking about how your brand should show up in ChatGPT, happy to compare notes or share a short demo.

Gleb Gordeev

Mission → 1k teams to adopt weekly prod. discovery. 📈 curr: 15/1000

I use AI tools 40h/week for real work. I share what actually works.

Most AI content is either too basic ("copy this prompt!") or too technical (47-node agent workflows). I'm in the middle — practical workflows for smart people who don't want to code.

**I post about:**
• Claude Code for non-developers
• MCP integrations that extend what AI can do
• Productivity workflows I actually use

**What I'm building:**
juttu — helping SaaS companies grow revenue from existing customers.

SWE → PM → Founder. Munich.

Niko Noll

use this weekend to setup claude code.
made you a how to.
and show you some use cases.

if you haven't already this is your chance to see what this hype wave is all about.

it's neither THE END OF ALL KNOWLEDGE WORK.
Nor is it to ignore.

I spend the last 3 months mostly in Claude Code.
Like 70% of my workday.

It requires some setup, and some getting used to.
It's a new way of working.
But damn is is powerful.

Let me know what you want me to cover next!
Or question you have.

I share the AI workflows I use to get work done | Founder @ Juttu

I got the most honest question form a busy founder friend.
What the F do people actually do in Claude Code.
I had my doubts.

Even uninstalled Claude again.
Now I live in it.

Let's
- find what actual search trends are
- identify available .com domains that can rank easily
- scrape competitors to find gaps in their content (our SEO opportunity)

What would you want me to automate next week?
This week I already promised a Tella -> claude -> LinkedIn tutorial to Louise. 

Send this to someone who doubts Claude Code.
And if you are sitting on the toilet right now (i know a lot of you are), bookmark this to watch later at your desk.

What are your (non-tech) questions about Claude Code? 

I’ll do a video on them.

Joining Louise and others in doing more video in January.
 
I've been working with Claude Code for some pretty advanced non-technical workflows over the last months.

And I can really say that it revolutionized the way I work even though that sounds hyperbolic.

I'm happy to screen share how I set them up and what they do for me.

If you're not getting into these new tools and ways of working, streamlining workflows end to end, I think you will be falling behind professionally.

The only way to avoid that is by adopting a learning mindset. There are no stupid questions so just put your question below.

No alternative text description for this image

I tested the hottest open-source AI agent.

It wanted my email, calendar, and root access.

ClawdBot runs on your hardware. Open source. Orchestrates multiple AI agents in parallel from your own machine.

Sounds great, right?

Then you look at the setup.

Email access. Calendar. WhatsApp. Root permissions on your entire machine.

And the agents... still call remote models.

So your data leaves. Your credentials are exposed. An autonomous process gets full control of your system.

That part honestly got me kind of scared.

But here's what's actually interesting.

It's one of the first real explorations of what multi-agent orchestration UX looks like. How do you manage dozens of agents running at once? How do you even stay in control?

The bottleneck isn't the AI.
It's us. Human orchestration is the constraint nobody's designing for yet.

The security needs work. But the question it's asking?

That's the right one.

What's your take... are you the bottleneck yet? :) And if so, how are you dealing?

P.S: Clawd was renamed to OpenClaw cause Anthropic complained that Clawed is too close to Claude.

P.P.S.
Hey, I'm Niko.
I share how I use AI workflows and product analytics to build better products.
I also make walkthroughs and tutorials on YouTube.
Follow along if that's your jam.

Stop feeling FOMO about Claude Agents.

Start simple.  

Do the work in a normal chat first. Track repeat steps in a notebook. When you catch yourself doing the same micro workflow 5 to 7 times, that is your signal to go full agent.

Generalize on the 7th.  

Tell Claude, we have done this seven times, write a concise prompt to turn it into a sub-agent. Open /agents in Claude Code. Paste the prompt. Pick Sonnet, not Opus. Cheaper, fast, good for routine tasks.

Now ... why? ... you ask

Well, these sub-agents run in parallel, keep context clean.  

Ask the main chat to launch the sub-agent for each target. Example, competitor pricing for seven sites, churn reason summaries in a fixed format, or file to summary transforms. Claude spins parallel workers, returns only the outputs, keeps the thread unpolluted.


Have you done the same task 7 times in your ChatGpt/Gemini/Claude?
Where would this save you the most time this week?

For me it was checking competitor pricing and fencing (what features are pro/premium).

My literal prompt:
```Launch competitor-research-subagent for each of the following urls. Create a summary table in competitor-landscape file```

P.S.  

Hey, I'm Niko.  

I'm building a retention tooling SaaS in public. Talking about AI, B2B subscriptions, pricing, and all that good stuff.  

Follow along if that's your jam.

Stop feeling FOMO about Claude Agents.

Start simple.  

Do the work in a normal chat first. Track repeat steps in a notebook. When you catch yourself doing the same micro workflow 5 to 7 times, t

The real meta-skill in startups is learning to swim in overwhelm.


What feels true today: in a job you juggle a lot, plus try to get deep work in

As a founder you juggle more balls and they keep multiplying, and quality is expected in everything you do.

I’ve been waiting to share a tidy build-in-public update, but turns out, I’m never there :)


Reality at juttu

We onboarded customers, built a pipeline, and the product still needs love for today and tomorrow. The hardest part is not tasks, it is emotion and expectation management.

Notice the anxiety, keep moving, drop fewer balls, pick them up fast.


My simple system

Pick one morning intent.

1) Find more customers.

2) Make existing customers happier.

Block time for it. Afternoon is ad hoc, support, emails, calls. Friday is the ops block.

How do you structure your day when everything is urgent? 



P.S.

Hey, I'm Niko. I'm building a retention tooling SaaS in public. Talking about AI, B2B subscriptions, pricing, and all that good stuff. Follow along if that's your jam.

"We can't predict when churn risk goes up."

  You can. You're just not looking.

Your users are telling you. Not with words. With behavior.

The real problem isn't missing data. It's that most teams don't have a system to read the signals in real time.

Here's a simple version you can set up this week:

Step 1: Pick your 3 core post-onboarding actions.

Not signup steps. The things activated users do regularly. Could be creating a report, inviting a teammate, running an export. Define 2-3 tracking events for each.

Step 2: Set a 14-day silence alarm.

When an account stops doing those actions for 14 days, flag it. You want those users top of mind, not buried in a dashboard.

Step 3: Reach out with outcome, not features.

Don't email them "Hey, you haven't used X in a while." Lead with the outcome they signed up for.
Remind them why those actions matter for their goal, not yours.

This won't catch everything. But it catches the obvious ones most teams miss completely.

What does your churn early-warning system look like today?

I turned screen recordings into scheduled posts.

Every word is mine. The workflow is AI.

I record screen walkthroughs and talking head videos on Tella. That's the content. My words, my takes, my experience.

Used to stop there. The recording would sit on my drive and... that was it.

Now Claude Code pulls the transcript via MCP. Reads what I actually said. Suggests LinkedIn posts based on my words, not some generated template.

Then it schedules them as drafts in Typefully. Also via MCP.

Recording to transcript to draft to scheduled. I don't copy-paste anything.

Honestly didn't expect it to work this seamlessly.

The trick? AI never writes from scratch. It only works with what I already said. That's the difference between content that sounds like you and content that sounds like... everyone else.

If you're already making videos... you're doing the hard part. The rest is workflow.

What's the one part of your content workflow you wish you could automate?

P.S.
Hey, I'm Niko.
I share how I use AI workflows and product analytics to build better products.
I also make walkthroughs and tutorials on YouTube.
Follow along if that's your jam.

Conversion signals for SaaS | Co-founder @ Juttu

People in tech keep using AI to "build me another agent." Boring.
AI is way more fun when you weaponize it.

I spent 47 minutes trying to cancel a subscription this week. 
In 2026.
I won't name names.
But I turned my frustration into a game.

Introducing The Cancel Quest.

There's a leaderboard*. 
Public shame and glory.

Current high score to beat: 10 sec.

Your move.
Come play: https://lnkd.in/e_CfPKTF

*yes it means I vibe coded an app AND hooked a database. 💅🏼 <hello world> 💅🏼

2026 is the year we stop talking about agents and start using them.

Most products say “agent”
And have a sparkle emoji.

The only metric that matters is jobs completed. Outcomes for the user.

My biggest learning, the harness beats the model.

What’s a harness?
It’s everything but the model.

It’s the framework enabling them to perform complex, multi-step tasks by handling critical functions like memory management, tool integration, context handling, and workflow orchestration.

Claude Code Harness is the smartest I have used.
It plans sub steps, grabs grep and other CLI tools to fetch data, and calls MCP tools when needed. It behaves more like an orchestrator, not a chat window.

For SaaS, that means pick harnesses that close loops.
Get a job done for you.

opencode is a strong open path, still very code first. I want non coding harnesses for support and analytics!

What end to end job did an agent finish for you without babysitting? If you could get a non coding harness tomorrow, which workflow gets it?

P.S.  

Hey, I'm Niko.  
I'm building a retention tooling SaaS in public. Talking about AI, B2B subscriptions, pricing, and all that good stuff.  
Follow along if that's your jam.

taplio

The Battle

Gleb Gordeev

Niko Noll

Followers

Posts

Engagements

Likes

Performance ScoreNiko Noll's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Gleb Gordeev

Niko Noll

Followers

Posts

Engagements

Likes

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Niko Noll's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreNiko Noll's LinkedIn Performance