1. These days I care a lot more about what the software is for, what makes it useful. So I don't need to get excited about trendy technology, I get excited about e.g. building a completely new kind of gene sequencing device that will help disease research and diagnosis. (Write about this here: https://codewithoutrules.com/2017/08/07/do-something-useful/)
2. Teaching is great, because it means all those failures weren't for nothing. Someone else can learn from them (I write a weekly newsletter of my mistakes and how to avoid them - https://codewithoutrules.com/softwareclown/), and I can learn from them (https://codewithoutrules.com/2017/10/10/lone-and-level-sands...).
3. At this point you can hopefully get better sense of _why_ project fails, and avoid things that are likely to fail, or try to prevent them from failing. So hopefully % of failures goes down.