If you have been looking for unbounded productivity growth as a software engineer or you are a non-technical person who has found a coworker in claude code, you must have heard of prompting techniques like ralph wiggum, get shit done or superpowers. I personally think they are pretty good. They have started building what next generation of software abstraction looks like. The abstraction can be defined as software tooling plus markdown files that support agent execution all the way from ideation, implementation, testing, debugging and shipping to production. Even go as far as automated deployments with feature flags, working on the metrics and choosing if the features should stay or go.
I found that these tools are really awesome for solo developers, who don’t have coordination overheads with teams. It is different from enterprise setting where there might be multiple people are working on the same project in different shapes and forms. The shape and size of features are decided on by different people. The differentiating factor, it seems, is collaboration. Not among agents, but different teams working towards different goals but sharing the same repository. There is also the question of tooling. This next generation of software is supported by a very thin command line tool and markdown files. But enterprise processes include usage of multiple tools, where the information is distributed. For example, a company might be using Confluence to share and discuss the technical design documentation. Many orgs now have AI bots joining the meeting, transcribing and everything and creating to do’s and next steps. There is important information that is stored in Slack threads that gives context to many of the technical decisions. Teams mostly use Linear or Jira for coordination rather than some spec and having tasks in markdowns in project repositories.
That said, there are many good ideas that can be adopted from these meta-prompting tools to accelerate feature development in enterprise setting.