Back to archive

Thread

4 tweets

1
@DJCordhose Yeah, and you might need to go forward/backwards at any time to revisit assumptions, work on new hypotheses, etc. In the design community there is this idea that you move through cycles of increasingly detailed prototypes (sketch, paper prototype, clickable prototype), etc.
2
@bobbruno70 @DJCordhose Something like this could also be done for software/ML systems, but the awareness isn't there, and most people have their spot on which they operate (DS: prototypes, SE: full systems).
3
@bobbruno70 @DJCordhose Another thing is that what you described would ideally all be in the hands of a few individuals without need to hand over stuff (which introduces waiting, potential for misunderstanding). But given the technical complexity we're no where near being able to do that.
4
@bobbruno70 @DJCordhose I think things used to be similar for deploying systems (dev machine => prod machine), and we've developed CI/CD tooling for that. Something like this, but for ML maybe?