Tobi Lütke: recommendation

Transcript context

…Final question: What is it you hope to learn next? I have been on a wonderful reconnecting-with-engineering kick for the last little while. I’ve really, really had a great time coming back. I think Copilot and AI have allowed me to mitigate all the downsides of not spending any weekends in work on computing projects. Again, I’m incredibly interested in LLMs, transformer, the machine learning world. It feels like an almost unending well of information and progress. I’m just fairly tapped on this. It’s hard to see this as a field because it just changes every couple of days. I am most thinking about — and I really feel like I can contribute a little bit — just figuring out how to have higher-trust companies at scale. People describe as “I’m a small-company person,” “I’m a big-company person.” and I just don’t think what actually the difference is. I think people are using labels around something they feel, and they haven’t got the right words. It’s probably many things that contribute to this, but certainly, parts of these things are a sense of agency and ability to impact. A lot of what happens in companies is that policies and processes — they’re well-meaning. They bring up the floor, so no one actually does something really, really wrong. But what people don’t see is, they also bring down the ceiling. You end up in these places where it doesn’t matter who you are, you’re going to do 7-out-of-10 work. Just the whole concept of entrepreneurship is about going for world class, and people need to leave in many places after a company gets to a certain size. I really think this is simply a path-dependent, unacknowledged situation that just comes from the way we have . . . the tools that we had to coordinate in people, and everyone holds their dismissal of trust as a part of this. I think the best areas of a company like Shopify keep this ceiling open so that everyone can reach as high as they have ambition for. Sometimes teams come together and just do absolute world class. That means you have to be willing to accept underperformance as well. The floor can’t be quite so high, so sometimes you get somebody who can’t ship. Sometimes you get something other direction, then the wrong way. It all comes with the territory. Businesses tend to think about these things as disastrous events, and they’ll do everything to not experience this, and therefore stymie all creativity. I think we are now gaining tools and approaches that can do this at scale. Just like what the engineers experience with writing code, and then you have a Copilot that helps you write code well, given what you’re working on, and quickly gain the insights that you need in the task that they vary understanding. Then after you are done with it, there’s automated systems that test. There’s automated linting, automated unit tests, and so on. I know this sounds incredibly nerdy, but what this basically is, is trust plus Copilot and automated verify. It’s a take on the trust-but-verify thing. I think that creates a wonderfully fun environment. now this sounds incredibly nerdy, but what this basically is, is trust plus Copilot and automated verify. It’s a take on the trust-but-verify thing. I think that creates a wonderfully fun environment. It turns working on areas into almost video game-ish. I think we know how to build these systems now, and I think we can build enormously better companies this way. They’re just more fun for everyone, and they also lead to better products. This is clearly possible because I’ve seen it be possible, and we just have to come up with a couple more ideas along those ways, and we have to figure out the particular downsides because, again, nothing is perfect. It’s just different sets of tradeoffs. I think the tradeoffs of building a company this way, rather than just reducing it to zero trust and mechanize everything, is enormous for society and productivity, and then just fun at work. I’m excited about that.…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.