But what is wrong in the meta situation is that while Meta officially projects its 2026 AI capital expenditures to be between $130 billion and $145 billion, investigative reports indicate its true future liabilities exceed $690 billion once unlisted back-end deals and lease commitments are factored in. They have used shadow accounting in many places to hide these things. As it's a publicly traded company so if that goes south it hits a lot of 401ks. So that in a nutshell is why its kinda a bad thing. But you are still kinda right in that its not an existential threat and it's not going to wipe Meta out even though those numbers could totally destroy the vast majority of public companies.

But from a fellow engineer's perspective we don't need another commercial solution in this space. Every major tech vendor is working on harnesses, and we will get ones that will run circles around yours for free. The open source ones that get popular will have mit/freebsd/etc permissive licenses. (or no one will adopt them) But coming at it from the angle you are indicates you haven't reasoned well about this. Please don't hurt yourself and those around you by starting a startup on this idea.

I'd be happy to pay a small monthly fee to license the model to run locally. I'm already paying for Claude, GPT, Gemini,etc.

The main thing I wonder is how well supported is it in cloud environements? AKS/EKS/etc?

Otherwise this is cool, we need more competition here.

You would hope that apple would open source it, but they are one of the worst companies in the world for open sourcing things. Shame on all their engineers.

I can see an overlap with people who want smaller computers who also want max power, but I just would not believe that is a significant group. (again, all personal observations)

The US is polarized with the republicans not wanting any socialized health care because they don't want to care for people not like them. You can't organize and correctly run a health care systems when half you politicians actively sabotage and burn it down.

Why do you say this? I thought the p40 only had a memory bandwidth of 346 Gbytes/sec. The m4 is 546 GB/s. So the macbook should kick the crap out of the p40.

First most companies thought they needed to do containers before serverless, and frankly it took them a while to get good at that.

Second the programming model was crap. It's really hard to debug across a bunch of function calls that are completely seperate apps. It's just a lot of work, and it made you want to go monolith and containers.

Third, the spin up time was a deal killer in that most people would not let that go, and wanted something always running so there was no latency. Sure workload exist that do not require that, but they are niche, and serverless stayed niche.

We mostly email links to spreadsheets running in cloud. So it really doesn't matter what your OS is any more from an excel perspective, as long as your computer can run a modern browser you are good.