@tjrobertson52

AI is progressing at an exponential rate right now. Most people have no idea how quickly knowledge work is about to change. I wanna talk about what I mean, the implications for 2026 and why this is worth paying attention to.

2026 has already been a crazy year for AI. Claude shipped 74 features in 54 days. Like, fully fledged features, not minor improvements.

And now we know that both OpenAI and anthropic are about to release huge new models.

You may have heard that OpenAI just shut down Sora even though they had a one billion dollar deal with Disney. We now know that's because it was eating up too much compute. To do that they need for their new model, codename spud.

They also shut down their new erotic model, which never seemed like a good idea. But the point is to putting all resources into spud.

It seems that Sam Altman is more focused than ever on getting to a G I. He recently declared that their main focus right now is on automating AI research. You wanna have an AI research intern by September, that's an AI that's doing intern level AI research work.

And by March 2028, two years from now, they want AI research to be completely handled by AI.

I think OpenAI sees how quickly Claude is shipping right now. They see how quickly knowledge work is being automated by AI, and they don't wanna miss out. So they're shutting down all the side quests and they're focusing on the main story, which of course, is their core model.

And if that wasn't enough, we now know that anthropic plans to release their next big model, code named mythos. I'm personally much more excited about this release. We do about 90% of our work with Claude.

I think Claude has always been the best model for knowledge work. And right now they are cooking.

6, blew everyone away with how capable it is. But apparently this new model, mythos, will sit in a tier higher than Opus. Internally, they're calling it copybara.

So their cheers go Haiku sonnet, Opus, Copybara. I'm guessing they'll come up with a better name.

So the models are improving rapidly. All the model makers are pretty much exclusively focused on replacing knowledge work. Yet the vast majority of knowledge workers are not currently well positioned to take advantage of these improvements.

That's because to fully take advantage of these models and features that are coming out, you have to rethink your processes. You have to return to first principles, consider what these models are capable of and what humans still do better, then take one process at a time and rebuild it from the ground up. Let AI handle everything it can do well or that we can assume it will do well in the near future.

And then double down on those areas where humans are still really important.

We're already seeing a significant growing divide between the companies that can implement these processes and those that can't. And most of the current productivity games were made possible by features or models that were released just in the last few months.

Do extrapolate out and consider all the model and feature releases we're likely to see in 2026, and then consider how much better we're going to get at integrating these models into our current processes. By the end of 2026, I predict we're going to see a massive divide in productivity between companies who are paying attention and those who have their head in the sand.