All the example use cases of AI personal assistants remind me of the use cases for google glass a la 2013: "now you won't have to look down when a new message arrives! wow!" I expect mainstream adoption of such AI "advances" to be similar.

SOTA LLMS keep improving but if you use them enough, their strengths and weaknesses are pretty clear. If you want to vibe code at max velocity, you should be aware of them too.

Vibe coding is prompt engineering. Write better prompts.

Contrarians like to think they are independent thinkers but they aren't. Their thoughts are still dependent on what the herd thinks. They are just negatively instead of positively correlated. If you are truly an independent thinker, there should be no correlation.

Just dropped Episode 2 of "The Forward Pass", our new show about the cutting edge of AI. Join me, Dan Schwarz Ray Fernando and Chris Scoville to get the latest alpha on how to use these tools to build your products. This episode we cover the Ghibli Craze and autoregressive

Mentla.ai is going to dramatically expand access to therapy with AI. We believe that AI therapy will be, in some ways, dramatically better than human therapy but I believe, in the end, it will be a different tool and each - human and AI therapy - will serve their

Windsurf is amazing but with OpenAI's acquisition (surely a data play), can't help but think that by using it, I am just training AI to replace me and all the other SWE/DSs out there. It is a deal with the devil?

Why AI therapy needs to be built from the ground up for that function.

Researching Superthrive for a stressed tree. gpt4.5 said it is a wonder and told me how to use it. o3 told me it barely a step above snake oil and not to use it. I wonder how often these models are in stark disagreement with each other. We need some LLM battles.

Congrats to da new pope!

Often fiction is better for self help than a self help book

The more time you spend talking to o3 about a problem, the worse it gets. The more time a human spends on a problem, the better they get. AI won't be replace all white collar workers until this pattern is fixed. And IMO it isn't is solvable just with longer context windows.

Can someone please ship a SOTA model that is a bit more restrained? 🙏

You are spending too much time with an LLM if this joke hits the spot!

We should all have growing concern about anti-semitism in the USA.

I have a PhD in neuroscience. And here are 3 key weaknesses of SOTA LLMs compared to the human brain. AI’s intelligence is currently jagged. It has become super-intelligent in some ways. In others, not so much. These weaknesses are blockers on the path to AGI. 1. Continual