I spent 50 hours over one long weekend building an app I can't sell, running a prediction model I can't explain, losing $250 on a market I don't fully understand. Building is fun. That's the whole point.

There is a big gap between vibe coding something that kind of works and pushing a production-grade app into the world. I've tried to cross that gap myself with varying levels of success. The learning curve is real and it doesn't care about your enthusiasm.

Who the hell is Ralph Loops?

nategemberling.substack.com/p/ai-stockholm…

Shouldn’t Claude just replace Slack? Today we have a chat for humans and a chat for machines. Claude has a better chance of making a chat for both.

The floor of what's possible with AI keeps rising. The gap between a team running the current best toolset and a team running what was best eight months ago is not trivial. It compounds quietly. Then shows up suddenly in the work.

Most companies measure AI adoption by breadth. How many people are using it. How many use cases identified. What percentage went through training. None of that tells you whether the tooling is current, the workflows are still right, or early adopters are operating at the

😂

Most AI adoption conversations focus on the hesitant. The more interesting problem is the people who adopted early and stopped updating. They have workflows. They talk fluently about it in meetings. The tools they're running are just eight months old in a market where eight

Non-technical people who can think in systems have always been blocked by one thing: they couldn't build the thing in their head. That blocker is gone now. The question is what you do with that.

I'm down $250 from my Polymarket experiment. I am not the world's next great quant. But I proved to myself I can take an idea from my head and make it functional. That's worth way more than $250.

Find your shamans. 3-10 people you actually trust on AI. Not the ones with the loudest megaphone. The quieter ones doing the work. Unsubscribe from everyone else. The noise is the problem, not the solution.

The best AI follows I have are people with day jobs. They share what broke, what worked, what surprised them. Nobody is selling a course. That is the signal.

People are fed up with Salesforce. 'Salesforce is the new Oracle' — heard it from multiple enterprise leaders this quarter. Not from us. From them. The once all-powerful seems to be grasping at market share in ways that actively hurt their customers.

When I was an SDR my edge was three things: finding emails, mail merges, and personalizing at scale. All three are now fully commoditized. The things that made someone a great SDR a decade ago have been pushed into the tooling layer. What's left is relevance.

The challenge in outbound is no longer personalization in the old sense. Not Tom's college. Not his favorite restaurant. It comes down to timing, leaning into signals, and generating a message compelling enough for an AI assistant to prioritize it for their human.

The real value of AI in the enterprise isn't one killer use case. It's in the aggregate. A PM converting scattered notes into a crisp product draft in 20 minutes. A finance analyst drafting a variance explanation in 10 minutes instead of 2 hours. Multiplied across every

Agent sprawl is the new SaaS sprawl. Every team building their own AI tools with no central strategy. Fragmented, narrow, ungoverned. We watched this movie with SaaS in 2015. We know how it ends.

More AI content does not mean more AI clarity. Usually the opposite. Find the person who says "I tried this and it failed" more than "here is the future." Follow that person.

Nobody spends their weekends selling. Developers spend their weekends building things because building is inherently enjoyable. AI just gave that feeling to everyone who ever had an idea but couldn't make it real. That's why this is spreading so fast.