Chad Donohue

Big news! I'm joining @tryramp on April 13! 🎉 Been in awe of what this team has built. Couldn't be more excited to be part of it. New chapter starts soon. Let's go! 🚀

@ericclemmons I started going this direction too. It's really neat to be able to solve your own problems quickly on how you want to work. And, the creativity of the community helps to feed even more ideas!

Conductor, Superset, and now Solo! Orchestrator apps the new TODO app 🤙

Using our new `issue-investigator` skill, I was not only able to easily trace Cloudwatch logs but also generate a nice interactive artifact to illustrate why the new suggested fix works. Seeing what happened is great. Visualizing how it can be fixed is even better!

@ericclemmons Agreed. Don’t sleep on using Claude playgrounds for reviewing PRs either. This is a nice way of looking at a diff and having a conversation about how it works and what it does.

Manual reviews are still the bottleneck, but AI only amplifies an existing process problem. The easiest solution is for a team to make reviews the 1st thing they do before anything else. Before email, chat, or their own work. By standup, there shouldn’t be a review missing.

@sanelo_ @burcs @claudeai Today, this is how the system works. Tomorrow might be another story, but it's learning and making adjustments whenever a prediction diverges > 10% from expected. 🤞

@sanelo_ @burcs @claudeai INSIDER (10 pts) CEO/CFO buy >$100k = 10 Congressional buy >$50k = 6 Insider selling = -5 They know the business. When C-suite buys with their own money, pay attention.

I've been tinkering with this little stock picker (paper money) over the past few weeks (borrowed some visual ideas from @burcs) . All built on top of @claudeai skills. This runs at market open/close to train itself over time. Really neat to see this visualized.

@burcs @claudeai 👀 Thanks, I'll check this out. Also, many thanks for the inspiration 🙏.

@chaddonohue @claudeai this looks great! if you're doing paper trading with alpaca you should check out the leaderboard: sukuna.dev

I recently started making a roguelite game where you play as a user that works through software tickets by prompting LLMs. You get a randomly assigned agent each time and have to prompt around the personalities. 😄

@chaseadams @trq212 @ericclemmons The only reason that I can think of for making `commands` now is to prepackage skill calls with specific intent, maybe? Like a wrapper function.

@chaseadams @trq212 I was talking about this recently with @ericclemmons. If I’m following this chart, then yeah, it looks like everything can now be a skill with configurations for where it can and cannot be invoked from.

I haven't seen anyone talk about it, but skills look like they can both be used as dynamically invoked skills *and custom commands*, so do you need commands anymore @trq212 at all? code.claude.com/docs/en/skills