@asmartbear Yes. It’s great for ML use cases. I use it for custom surf reports, my app uses it for cancellation risk for bookings, and it’s really good for data analysis/forecasting. It’s really good when a machine needs to communicate with another machine.
Companies are super interested in AI as a Service market, and even reporting (BAIN) a 96B opportunity for managed services by 2029. However most of these companies aren’t unlocked internally, and trying to cheaply build alternative to the frontier (which they’ll never catch).
@dhh No model/harness runs on conventional apple hardware to unlock the ecosystem. They’re waiting for that. Harbor is an example of unlocking it for non conventional hardware. No one wants someone else’s agent.
@dsirokyai@ivanfioravanti Ah.. yeah this was my same thinking. Doesn’t make me excited to get the 64gb, which stinks because I know it’s still capable.
@GergelyOrosz Completely makes sense. The rhythm of business that made waterfall/bup work was the real value all along, especially communicating with customers on what to expect.
@Boku_JM@clamanna That’s not the argument. Their proposal is one step before the debug flow you’re asking for. They are saying they can give you the right outcome, and you shouldn’t care how it gets to that outcome (ie which model).
I want control as well like you. But they are selling outcomes.
@Boku_JM@clamanna This is what you can expect. But they’re selling outcomes, not models. So do you more care if the outcome is what you expect vs the tool that got you there with the cost, perf, correctness limitations he mentioned?