Are we dismissing AI spend before the 6x lands? (2025) | Hacker News Reader

Are we dismissing AI spend before the 6x lands? (2025)

martinalderson.com

22 pointsukuina7 comments

I feel like I’m being sold something. Mostly the use of contrasting-phrases to repeat points… Unsure if that’s just the smell of second-hand LLM-use or if the article was largely generated.

Tbh Im more excited for cheap tokens decent quality than tokens that are a fraction of a percentage better on some benchmark.

A lot of problems out there can be solved with volume rather than bleeding edge

I'm not sure this is the right outlook. This massive spend isn't even necessary to get huge returns. We should look at how people are using LLMs right now instead of chasing promises of superintelligence.

There is a ton of value in LLMs improving accessibility for the web, and it feels like we're on the brink of AI assistants finally delivering. The new SEO is MCP tools and it's forcing app designs to be more robust than the brittle flows of the past.

This means less friction for ecommerce and breaking ground on all new UX. How is that not a big deal? Developers can provide better apps with more accurate and flexible designs on more devices. Customers will go where the best UX is and reshape who is on top in every category of consumer tech.

"We should look at how people are using LLMs right now instead of chasing promises of superintelligence."

This. When more computing power and memory resources became available to software engineers, we've seen how that impacted the software development. Sure, there were happy stories of new class of problems being attacked, which couldn't before due to resource limits, but a lot of software just stopped being frugal and did pretty much what it did before, but somehow consuming much more. Extrapolating to the case of staggering resources poured over the AI solutions, I'd be surprised if most of it won't just be consumed to generate higher resolution (and longer) videos.

Pretty sure you always have to oversell. They're ofc doing that already

What can 6x extra compute do split between all these companies? It seems they've already exhausted all natural data + curated sets + ai generated data. What else can be added?

I don't think they've fully leveraged the data they have collected yet

Reply on news.ycombinator.com