Raw LLM Responses

Everyone's own in-house models usually do well or excel in one or two types of requests but will perform less favorably in others. My work provides us both Gemini Pro and ChatGPT Business plans and I can definitely say that whenever I wanted to vibe-code a quick and rudimentary tool to help get a tedious job done faster that ChatGPT's coding ability is far and above Gemini's, but if you want to actively learn about such things then Gemini's is much better at guiding you through things and explaining what things do and why you're doing it a certain way. These deals will typically allow some sort of access to the data that the models use or interact with, sometimes how exactly they process them. This means Google can look at Anthropic's models some and see how they're achieving better results in areas Google's Gemini model isn't doing as well in. It could be the model weights themselves, how the model parses and processes the request from a user to determine what the user is saying vs what they're actually wanting, etc. Google then can refine their own model with it. China did something very similar with EV's. At first they were very open with allowing foreign EV's into their country, particularly Tesla, which they then watched how their citizens shopped for EV's, what they looked for in them, etc. China's own EV market then started to boom as they brought out model after model of EV's that competed very closely with Tesla and now Tesla is a much smaller share of the Chinese EV market. Sometimes, in order to improve your own product you have to let a competitor flourish a bit and make a deal where you learn from their successes and failures during that period. On a personal level, if you ever go for a promotion and they choose someone over you then you should try and reach out to the person they chose and speak with them about how you each managed the interviews and the questions asked in them to see what they did better than you and use that in the