I'm being asked (pressured?) to get all my developers to 'us AI as much as possible'... so what are the real uses that others have found truly effective, or even transformative... right now we're rolling out CoPilot lics to all our devs, but so far, it's not being used much and basically everyone seem to have the same opinion... '...it is somewhat better autocomplete, but with the added bonus of giving you bogus and / or incorrect responses... so it keeps us on our toes at least.' That's essentially what all my devs have said in one form or another.<p>Are we missing some higher level uses, or is it that simple? There seem to be a sort of fever dream in management that these tools will magically make unit tests for example 'automatic now', but so far we've not seen that pan out.<p>What are others strategies around Ai and tools like CoPilot? Are there certain uses / trends that are actually really effective?
- Co-pilot for writing unit tests, renaming variables, asking for shell cmds<p>- gptscript for commit messages, MR comment ideas, etc.<p>- mods for piping a file and asking for recommendations, could do with gptscript too.<p>- Raycast for hotkeys to gpt-4 API or model running with Ollama<p>It's still super early, so I try different ways of using it every few months/weeks and keep around what works.
We use ML and DL systems as a component to our products, but we don't use AI-based tools as part of our development process. Devs are free to use AI-based tools as part of their workflow if they wish, but after a period of experimentation, almost nobody does.
I'm independent, but I try out Gemini every time I need some dumb powershell plumbing to tie things together.<p>It's a net time saver, but Gemini still trips up on all of the wacky foibles of powershell, so it needs thorough code review.