I stumbled upon this YouTube video the other day — and this guy gave probably the most honest, no-nonsense advice I’ve heard for anyone just starting out in the AI space. He cut straight through the noise, called out what’s really going on, and named exactly who’s selling you false hope.

The video is from Max Brodeur-Urbas, founder and CEO of Gumloop — an automation platform running over 4 million workflows a day for companies like Shopify, Instacart, and DoorDash. His team is 15 people. The platform powers enterprise-scale automation. And he built it after being banned from the US for 5 years, coding alone in his bedroom in Vancouver.

Worth listening to.

The hype is real — and it’s designed to fool you

You’ve seen it. Twitter is full of people claiming they automated everything, work one hour a week, and make $10 million from a SaaS app over the weekend.

Max’s take: most of that is marketing. They’re lying to you.

What sells easily isn’t a product — it’s hope. Specifically the hope that you can skip the hard work, shortcut directly to the value, and buy your way to success with a course or a workflow template. Max calls these people “course bros.” They sell you a vision of effortless productivity without offering anything genuinely new.

The painful truth he laid out: if there was a magic workflow that made you $30,000 this weekend, they wouldn’t be giving it away on Twitter. And the only person who reliably makes money from that course — is the person selling it to you.

This isn’t new. Max drew a clear line between AI hype and what came before it — crypto, NFTs. Every time a bubble forms, there’s a section of the community that’s vulnerable. Easily persuaded. Desperate to believe that something will save them from their current situation. And every time, people line up to exploit exactly that.

The real anti-pattern: 50 AI agents running your company

Max named something specific he called an “anti-pattern” in how people are using AI right now.

It goes like this: you automate everything, hand over decision-making to a suite of AI tools, and build a system so complex it becomes a slot machine. You press a button, something comes out, and you have no idea why.

His rule is simple: only automate what you actually understand.

If you’re using AI to code something you don’t understand, you’re not building software — you’re creating malware that will bite you later. If you’re automating a workflow you could never do yourself, what you end up with is noise, not efficiency. Vibe coding only goes so far.

The people generating real value with AI are doing something different. They’re taking something they understand deeply, applying AI to speed it up, and scaling from there. They’re AI-enabled, not AI-replaced.

How he actually built Gumloop

The origin story matters here because it wasn’t a plan — it was a series of accidents.

Max studied software engineering at McGill, landed a job at Microsoft, hated it almost immediately, and quit. He moved back to Vancouver with a plan to build something from his bedroom. One weekend trip to visit old roommates in Seattle ended with him being turned away at the border, banned from the US for 5 years on suspicion he’d overstay his visit.

He drove back in shock. Took a couple of days to settle. Then just worked.

For the next 6 months he tried everything — video game moderation in VR, trust and safety tooling, bot detection, anti-spam platforms. He’d build an MVP, try to sell it, learn it was a bad idea, and start over. Sometimes once a week.

What he learned from failing that many times: in startups, you’re not trying to prove yourself right. You’re trying to find the reason something won’t work. The faster you find it, the more time you save. He wasted 3 months early on hoping someone would validate his ideas. The better approach is hunting for someone to tell you exactly why it won’t work — and if you can’t find that reason, you might actually have something.

Gumloop grew out of him noticing something specific in the AutoGPT Discord server. People were asking questions like “what is GitHub?” and “how do I use a terminal?” They wanted to use the technology but couldn’t get past the setup. So he built a simple UI to solve that problem. He didn’t think it would become anything significant.

The insight that changed everything: the agents weren’t actually useful. People had the platform but were frustrated because AI kept failing them. What they really wanted was reliability and predictability — not intelligence, just consistent execution. So he gave them that. A simple framework that automated steps one after another. And it turned out the audience that went crazy for it was 80% non-technical: operations people, HR teams, business admins who just wanted something to actually work.

That’s Gumloop. Built from observation, not strategy. From failure, not a plan.

The split that’s coming

One of the most interesting things Max said was about the future of engineering.

He floated the idea that the last generation of truly great engineers might already have been born — people who had to understand what was happening first, then got accelerated by AI. The new generation can skip the understanding entirely and just accelerate.

That creates a split. A smaller group of people who use AI as a learning tool, who understand the fundamentals and use it to go deeper. And a much larger group who just accept that the thing works without ever asking why — until it doesn’t.

If you have the discipline to pause, actually try to understand the problem, and use AI as a teacher rather than a shortcut, you’ll become genuinely exceptional faster than any previous generation could. But if you don’t — you fall to the slop with everyone else.

One thing he said that stayed with me

There’s a million reasons to not build any startup. Easy to ask what the moat is, why a bigger company won’t just copy it, why it’ll never work.

Max’s answer: the people obsessing over those questions never build anything. They end up as pawns in someone else’s game.

The people who do build something — they just tried. When it didn’t work, they tried again. The confidence wasn’t earned in advance. It was blind. And it was the only way in.

Watch the full video: youtube.com/watch?v=CxFQykWiJqY