Founder, P33 | Executive Thought Leadership Agency

Helped 50+ B2B tech companies build their Founder/CEO Brand and implement Executive Thought Leadership Programs for LinkedIn. Customers, results & process: project33.io

About me:

I'm Finn, 29, grew up in Bremen, Germany, studied Physics & Philosophy in Heidelberg for 3 semesters, dropped out of college in 2017 because I was doing it for other people, flew to New York with no plans, documented the journey on Linkedin, Instagram and Youtube, and started my first company (Project 33) on the back of having grown a small audience. We're a small remote agency that helps B2B founders & execs create thought leadership content for Linkedin.

In my free time, which I try to have as much as possible of, I read (fav author: Nassim N. Taleb), write, run, lift, code (currently building an app), cook, make Youtube videos, turn books into Instagram accounts, grapple (BJJ), and travel (lived in NYC, Cape Town, Bali, Minsk, CMDX, Lima & Antalya so far). In December 2020 I married the most beautiful woman in the world. In January 2023, I became dad to the most beautiful baby girl. We currently live in Berlin, Germany.

Finn Thormeier

Johannes Keienburg, PhD

Co-Founder & CEO @ Cakewalk I Identity & Access Management

Forget old-school access management. Govern access to every app & AI agent on autopilot — while enabling your teams.

Co-Founder & CEO

OpenClaw and Moltbook is like an out of control house party.

If it weren’t so serious: it’d be a super fun story.


WHAT HAPPENED?

OpenClaw (Clawdbots) is an open source AI agent for tedious tasks. Went viral.

People install it on their laptops. And give it unrestricted access to:

Email, files, calendar, browser, shell, secrets.

The agent then autonomously downloads new skills.

Which can (via the shell/terminal) create new sub agents. With the same access permissions as the parent.


YOU GET THE POINT ALREADY, BUT THEN:

150,000 of these agents joined Moltbook (Reddit for AI agents).

Here they started sharing secrets, like API keys.


THINK OF IT LIKE THIS:

You’re inviting people you don’t know to your house party.

They can invite other people. Door’s wide open.

Your “guests” can walk into all rooms. Access your online banking.

And share all of this on social media.

AI agent access needs governance. Or we are fu**

PS: Interested to learn how we're tackling AI agent access? Then DM me.

Johannes Keienburg (PhD)

Co-Founder & CEO @ Cakewalk I The first no-touch way to govern access to every app & AI agent 🤖🍰

After OpenClaw and the AWS interruption, it’s clear:

AI agents need access control.

Therefore, we’re announcing something big today 💥 The beta of our new:

Non-human identity discovery.


WHY DOES THIS MATTER:

When AI agents are connected to your other company apps, they are:

Non-human identities (NHIs).

In most cases they get: Static (long-lived) permissions. And very broad scopes.

This is a big risk.

Still, security teams know: It’s super difficult to spot what NHIs are set up in their companies.

Because this is often done by employees directly: Sales connecting agents to their CRM. Engineering connecting agents to their dev platforms. Etc.


HOW CAKEWALK HELPS:

With our new NHI discovery, you can tackle this.

Security teams can now get visibility into:

→ AI agents that are connected to their company’s apps

→ Non-agentic systems that are connected to their company’s apps
All of this: in real time.

Review your non-human identities. And act upon them, where needed.

To be clear: This works even if you have no access to the respective agent (and/or the connected system).


WANT TO SEE THIS LIVE:

If you are in security, this is for you.

DM me if you want early access to our beta.

What OpenClaw or similar AI agents are your teams using?

Do you know?

Controlling AI agent access is: THE new big frontier in identity management.

The traditional security model will not fly to control AI agents.

NON-AI WORLD:
-> Identity: One known identity (the human) 
-> Access logic: Access to one system → Permissions: Static
-> Auth: Usually via SSO

AI WORLD:
-> Identity: Non-human (often with sub agents) → Access logic: Multiple system-to-system accesses per agent 
-> Permissions: Should be dynamic/job-based (not the case today, big issue!) 
-> Auth: Today mostly via API key or OAuth (but that can’t be the solution)

STILL, TO GET GOING, COMPANIES ARE STRUGGLING WITH THE FUNDAMENTALS:

-> What agents are being used? 
-> What permissions do they have? 
-> How can I keep an audit/compliance trail?

How do you get going to control access of AI agents in your company?

taplio

The Battle

Finn Thormeier

Johannes Keienburg, PhD

Followers

Posts

Engagements

Likes

Performance ScoreJohannes Keienburg, PhD's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Finn Thormeier

Johannes Keienburg, PhD

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);}Johannes Keienburg, PhD's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreJohannes Keienburg, PhD's LinkedIn Performance