Why We Wait

Please exit the AI waiting room.

When uncertainty rises, there’s a natural reaction to wait and see. How far can AI scaling take us? We’re only 2.5 years out from ChatGPT’s release (it feels like a lifetime, the narrative has twisted and turned so much!) Will knowledge work exist in 2028? Software engineers? Be skeptical of the AI prophets (social media bros), and instead focus on what can be done now. Waiting is the worst choice we can make.

Besides, maybe we won’t need to wait long to find out the answers if you believe in the “this decade or bust” hypothesis, the idea that AI systems scale with compute, and we can only sustain this level of compute scaling for the next few years.

I’ve seen glimpses of an AI promised land along its jagged frontier, the wide-range in ability between tasks it nails and fails. I’ve also been frustrated by all of the easy things these systems cannot do. Claude Code can now one-shot entire working applications with a frontend, backend, and photo uploads but sometimes fails to consistently perform simple style edits throughout a large app. This is the nature of working with an alien intelligence, constant surprise by what it can and cannot do.

Why Now

At this moment, it seems clear that humans and AI systems can achieve far more than humans or AI alone. The jagged frontier has progressed from the land of novelty towards the truly useful. Now is the best time in history to build things. The time between idea and creation has never been shorter. Perhaps in the future, that time will be shorter still, but we shouldn’t discount the current moment. There has never been a better time to create.

Yesterday, Claude Code and I built a game called deskguessr in a few hours. This would have taken me days to weeks before.

What to Do

If the future were certain, we wouldn’t be writing and discussing it so much. So leave the AI waiting room, and find all the ways in which the AI systems can help you. Rather than focus on what they cannot do, find what they can. The best way to predict your future is to create it.

Step out of the waiting room and into your future.

Resources:

• Simon Willison - probably the best single resource to keep up to date with the latest advances in AI as they apply to practical software engineering
• Dwarkesh Podcast - great for interesting discussions on AI futures and keeping up with the latest words and phrases that will make it seem like you know what you’re talking about
• Epoch - more ways to understand the future of AI
• SemiAnalysis - deep dives into semiconductors. Since AI involves an industrial scaling process, understanding the hardware is crucial.
• And most importantly… Claude/ChatGPT/Google Gemini - the best way to learn about AI is to use it to do something you’ve always wanted to do!