@tjrobertson52

2 is out, and I think it's better than we all expected it to be. The reason expectations were generally low is because this model seemed to be rushed out in response to Gemini 3. 1.

It does seem significantly more capable. So what do I mean by that? It seems it can handle a lot more context without making mistakes, and generally hallucinates less.

It also seems like they were specifically focused on improving the product for the workplace with this release. ChatGPT has always been the leader for consumers, but Claude and more recently Gemini have been eating up more and more of the enterprise use. 2 seems like a direct response to that, and I think this has some real implications about the future of work.

2 is capable of, it does seem like there's a whole new category of work that can now be done by a large language model. Most of this is the kind of work that is done in spreadsheets. Or at least it could be done in a spreadsheet.

For any task where you want to give the model a lot of information and you want the model to really consider all the information and iterate over it multiple times. 2 is the best model for that right now. And again, it's not That it's smarter.

Gemini 3 is almost definitely the smarter model right now. It's just that it's willing to spend a lot more tokens. Google's just much more focused on efficiency, which I do ultimately think is going to be the winning strategy.

But right now Gemini just feels kind of lazy compared to ChatGPT and Claude. 2 seems to be the current leader for this spreadsheet type work. And when I say it like that, it might seem like it's a small subsection of work, but I think we need to start rethinking how we do a lot of the work we currently do.

I feel like we're probably only utilizing 10 or 20% of what these large language models can do right now. 2 has done for me is shine a spotlight on this skill that we just lack the skill of restructuring our work so it's conducive for a large language model, and it's probably not very helpful in the abstract.

`n`nIt's really making me pivot away from how I've been doing things, which is creating SOPs that anyone can follow, and more towards just training and coaching my teams on how to properly use large language models. My thinking is that if I can make a step by step checklist, I'll eventually be able to automate that. But someone who really understands how to use a large language model to work efficiently and to improve the quality of their work is going to be incredibly productive and valuable going into 2026.