Few lessons I learned in the first 8 months of using AI everyday.
1. Optimze what model to use. Don't use the best most expensive models for small and redundant tasks.
2. Don't extend the session for too long as the context becomes too large and AI starts hallucinations.
3. Set the reasoning effort based on the task complexity.
4. Use a token saving tool, caveman for example. It's not a massive token reduction but might help you save some $.
Plans for the future: run local AI models.
Glad it worked for you. I was using it on a project with 16 microservices and while the context window is indeed bigger it constantly messes up basic bash/cmd/powershell commands. Not to mention it brings up things I didn't ask for and I barely managed to do things with it on the first try.
Hello, hello π
I am looking to #connect with people who are marketing more than building.
If you build something cool and share your journey daily I'm curious how far you'll go.
I am working on a project with microservices and it's a bit difficult to manage context across all services. Two days ago I asked for a change in an api call which triggerd a job from a workflow and on the surface it did what I asked for but later I discovered some dependecy was impacted