[jayruy]

From how you describe your experience of teaching yourself to not only create a working program via independent research, but also recognize it's flaws and refactor afterwards, I have no doubt you could find some level of success in the industry. Not being able to "execute on your ideas" is not that important in many micro-managed niches of the industry, which are not great gigs in the long run, but are good if you are starting out and need to develop perspective on "getting shit done". Be humble and look to learn - you'll be fine.

I'm a bit confused on the Columbia thing - if it's 5 years instead of 4 - go for it, education is the best investment you'll ever make. If it means 4 more years after undergrad: just go work, it's probably not worth getting out-of-sync with your age cohort.