Lotfullah Andishmand

Imagine being this early

If I were 18 today, I would focus on a few simple things not shortcuts, not hype. First, I would master one programming language deeply. Not just syntax. But thinking. Problem solving. Patterns. Second, I would work on communication. Because your ideas are only as powerful as your ability to explain them. Third, I would become hands-on with every relevant AI tool. Not just using them for answers but understanding how they work, where they fail, and how to build with them. I would write daily. Code daily. Learn daily. At 18, You need skills. You don’t need funding. You need fundamentals. The world is moving fast. But the winners will still be those who build strong basics early. Focus on language. Focus on communication. Focus on AI literacy. Do this for 3–4 years silently… And the doors will open loudly.

There are at least 2 big but subtle factors contributing to the sense of overwork due to agents right now. 1. The leverage on incremental effort has gone up substantially due to AI, and anyone using these tools tend to feel it first. We’re so used to everything taking so long to get done, that spending that much extra time on something didn’t have the same value. But now you can sense the compounding leverage you have with agents much more acutely. It feels more like when you’re a people manager and not maximizing what your team is working on. The worst thing you can do is waste your team’s time, point them in the wrong direction, or have them be idle. Now ICs get a similar sense of this via managing agents. Prioritization and thinking through how to break up tasks and maximize work becomes a key skill. 2. The other rested factor is we’ve just made it so much easier to start incremental tasks that end up taking much longer than we realize. Very easy to get to the 90% solution but the final 10% takes the vast majority of the time. So we are starting far more projects because of the lower barrier, and spending a ton of time to finish the work. I regularly start a project at 9PM that I think will be quick, and find myself at midnight still completing the work. The interesting thing about this fact is that we’ll use AI to test lots of new ideas an hypothesis, and quickly figure out which areas to continue and sustain. This will ultimately lead to a lot of the job creation with AI, because teams and companies will decide that the experiment needs to get promoted into a production process. But they wouldn’t have even started that “experiment” without AI in the first place. This overwork obviously can’t sustain across the economy, but it’s interesting to see how this is shaping things right now.

Introducing Claude Opus 4.7, our most capable Opus model yet. It handles long-running tasks with more rigor, follows instructions more precisely, and verifies its own outputs before reporting back. You can hand off your hardest work with less supervision.