Future of Coding is an online community with a welcoming, cooperative, and revolutionary spirit. We are unified in the belief that the common practice of programming is tragically less humane than it could be.

Future of Coding

Recently wrote about my own personal agentic engineering patterns taking software from 0 to 1:

<https://shreyasprakash.com/my-agentic-engineering-workflow/|https://shreyasprakash.com/my-agentic-engineering-workflow/> 

I keep referring to this <https://www.bassimeledath.com/blog/levels-of-agentic-engineering|bassimeledath.com/blog/levels-of-agentic-engineering> as a good taxonomy of agentic development patterns.

<@U02E4DAQGSZ> Right now the threshold I haven't crossed is LM-based PR review. It's plausible that someone has written a skill that checks for regressions in, say, React components that refresh too eagerly and clobber user input. I think I am going to seek out those repositories of open skills.
&gt; MCPs and custom skills give your LLM access to your database, your APIs, your CI pipeline, your design system, Playwright for browser testing, [...] my team shares a PR review skill that we've all iterated on (and still do) that conditionally launches subagents depending on the nature of the PR. [...] Skills get the same treatment as code: pull requests, reviews, version history.

Feels a lot like shopping for soundfonts, though. Nobody really does the vetting on those apart from 'try it and see'.

Putting two and two together from the Atelier share-out. Very interested in mucking with <@U0249H3TD46>'s grounding efforts under Verify.
<https://atelier.dev/guides/task-verification/|atelier.dev/guides/task-verification>

We use cursor bot for review which is very good. But no-way we are taking the human out the loop any time soon. Still, cursor bot does find a lot of little niggly race conditions or mistakes so it definitely helps