maybe you know what inference time compute is and just didn't bring it up because it's hard to explain, but it's an absurd omission in any conversation about scaling and upcoming models. if you're going to review rumors you need to actually understand the industry and the context.

what's meh and compared to what?

nothing in that conversation was a product review about something currently shipping. It was gloating over an anticipated lack of ability in a product that's never been announced, has no release date and certainly hasn't shipped. No one is pissing their pants.

and just to close out my manifesto - you kind of made the point yourself in the episode - if these things are going to be god, why waste any resources building netscape now?

LLMs are less than a decade old and we've seen remarkable and accelerating progress in the past 3 years. We've seen constantly shifting goal posts on what "AGI" is, but to act like the technology clearly doesn't work when it's still constantly improving just doesn't make much sense.

Use o1-preview, use advanced voice, use gpt-4o with canvas individually and it can be difficult to see the overall value, but the only reason they aren't already combined is cost. Imagine them as a single coherent system and that should clarify what's actually possible with the current tech.

and really consider incentives - OpenAI can't charge $2000 a month for a chatbot. Consumers don't have the ability to turn the output into revenue, but an agent that can approximate a jr dev is worth 100x that to a company like Google. Chatbots are not the product - it's always been about agents.

think of them more like darpanet than the 2005 internet. There's no incentive to build the UX around the current models because the tech improves more quickly than the you can possibly build the killer app. Even if we've hit a wall on scaling, the current tech is world-changing with better ux.

@reckless.bsky.social as a tech platform that bleeds into a normie audience, you have a responsibility to help those people prepare imminent unprecedented change. Consider how much your contempt for experts here reeks of the same dismissal of academics we saw with covid vaccines. pure vibes.

@reckless.bsky.social you're a smart guy - normally when something doesn't make sense to, you calling bullshit is the safe bet. Occasionally, though, when something doesn't make sense to you, there might actually be something you're missing.

@reckless.bsky.social when you're suggesting that the biggest, most well-run companies in the world are wasting the largest cap-ex investments we've seen since the pyramids - it's probably a good sign that it's time to reevaluate your priors.

these aren't your typical sv tech bros - it's a field that is flush with literally the smartest people in the world. When you're laughing at the stupid things Stanford cs and Princeton physics PhDs are saying, it's probably time to reevaluate your priors.