@Ctrl_Alt_Muse@RomeoLupascu@Grady_Booch And, of course, tests that actually find issues you need to know about, not simply checking the bounds of an assignment.
@RomeoLupascu@Grady_Booch They "test" in production. These companies don't even have QA most of the time. It's a strategy right next to not actually having customer service.
@lhtness66060 @GergelyOrosz If you read between the lines this also sounds like a storage access issue, the main change recently would be related to using repos as training data.
@zikasak @lhtness66060 @GergelyOrosz Because testing in production saves money, except when it doesn't and they throw some engineers under the bus to save face.
@kunchenguid@GergelyOrosz The lack of CEO isn't about responsibility as far as customers are concerned, it's a signal that they are a chicken with its head cut off, being bled out before it's dismembered for dinner.
@DanFarfan@GergelyOrosz You know they just bought it to get more control over the developer ecosystem right? Look at their track record for things like this. If it's not integrated it's slowly stripped down.
@BAPxAI@Grady_Booch@sboag@IBMResearch@GaryMarcus It doesn't really provide cognition either, but explaining the difference is quite a bit more complicated. Basically this system is nothing more than cached and compounded LLM outputs. The "substrate" you're looking at isn't the model or the system, it's that cached data.