Lots of talk about whether AI's gains on productivity are real, but the real boost isn't from being 10%-100% more productive. In the best case you can do things you literally would not have been able to do without it.

Lots of talk about whether AI's gains on productivity are real, but the real boost is isn't whether you're 10%-100% more productive. In the best case you can do things you literally would not have been able to do without it.

putting the oss model weights on a usb is such a good merch idea

While I wait for access: I hope GPT-5 (thinking) is useful for all the same things o3 is. I really like o3 for research & use it every day because it’s so reliable.

gpt-5 reacts to how you use chatgpt itself and what models you prefer really interesting

it helps to not frame your product as a gigantic laser weapon that will blow the planet up

big video models will obviously all become world models in 2 years. it doesn't have to be game-like, but you eventually need a way to direct your outputs, control your camera, etc. steerability only gets more important

the most insane tech acronym was probably SoLoMo. Imagine seeing that as many times as you see MCP today

Gaben is one of the most lowkey rich guys for how impactful he is. Changed games forever. Launched the first app store. Made one of the first consumer VR headsets. Working on BCIs. Owns a deep sea exploration company. Now a yacht manufacturer.

You can get significantly better results if you know which AI to ask what kind of question. Here are the ones I'm using right now: o3 - Best at finding information where specificity matters. Bad at creativity. Strong at straight thinking, weak in lateral-thinking. Use it for things where you want citations, numbers, breakdowns. Trails off at long context. GPT-4.5 – My former daily chat model. Strong at complex chats with long context. Good for lateral thinking. Use it for abstract discussion about a topic. Opus 4 – My favorite model right now for daily chat. Great for complex problems, replaced GPT-4.5 for me. Grok 4 - Great for analyzing things happening in realtime. Use it to understand how people feel about markets, sentiment, context. Gemini 2.5 Pro - Good for complex problems. I like it for the kind of problems that Opus 4 solves, but Opus feels better and is willing to be more specific when asked. I use it when Opus fails me.