RayBenefield

🌞🌞🌞 Week 2 of Claude Code It's been a wild week, and my AI sentiment is growing. It's a powerful, but dangerous tool. But like a knife can be used for bad or good. Gen AI IS bad, mainly because it allows bad capitalistic actors that don't care about other humans to keep them out of the process rather than making space for them to thrive. However in the hands of the right people, it is absolutely an accelerant. I'm an automation engineer with HIGH standards for good automation. So finding the right way to train AI and the right things to add/remove from the workflow is crucial to being good at it. Quickly realizing that several harnesses/success gates aren't as necessary as I thought. So if you aren't cleaning up things regularly and rethinking your workflow, you are wasting valuable time/tokens. Your ROI on AI starts to drop. I did hit max usage on my $100 plan for Fable, and 84% for all other models. My primary goal was teaching it everything I've been doing for the last 3 years. I needed it to understand the timeline and the goals. Both for our own record and also for its own understanding of WHY the code is very different. Took me about 3 days to fully go through with it. And after that the rest was really, how do I get it introduced to what we are doing on a small scale. So since the end of the first week was creating a good art pipeline that supports human artists, I decided to have that art pipeline output interact with the framework. And this morning was the final test of something with basic art functionally fully working landed and meeting the standards of the framework enough to work with it. Now every widget's art comes from that art pipeline, and the `Close` button at the bottom of the screen was also created as its own component for the framework and works exactly as I need it to. Close the UI. Well act as an action button really. So this was the first slice of getting AI into our own company's work. And I feel really good about what's about to start happening with our own product. The basic art styles of all the widgets now come from our art pipeline, the widget scaffolds to work with the framework are now built via AI. And that close button is all of it working together for new functionality. This next week is getting my second brain online and then teaching AI to do a larger part of the framework to add more substantial features. I'm stoked for the future ahead.

💪💪💪 Week 3 of Claude I had an existential breakdown Friday morning... not cuz my "career" was somehow obsolete. But that I have spent years trying to dream in more consumable bite sized chunks for others to consume. And I feel like I can take my ankle weights off. This week... I decided to push Claude HARD into our company work. This "framework" I've spent 3 years of my life dedicated to. The first couple days since my last post, I had a laundry list and a "managed" surface area and expected it to take a week. It was barreled through in a day. That's why I broke. I've been restraining my vision for external validation and understanding. And I finally get to just trust myself and go bigger. There was this validation of "I was right, and so few cared enough to understand or listen". I've had this "moment" before in Software Engineering too. All the blood, sweat, and tears going into this and being an outcast in various ecosystems from tech to games. Always labeled as a "purist" and told I was not grounded in reality. So after that meltdown in the shower balling my eyes out. I took the ankle weights off and I started to dream bigger and reach for things I was holding back for later cuz it may be too complicated for someone to grasp. But Claude doesn't have an ego or insecurities, so my autistic side doesn't have to stress about poor reactions and nuance. Over the past week I've probably shaved 3-6 months off my timeline... and reached partially into the 6-12 month goals and laying seeds for what were 1-3 year goals. So I'm excited. Nothing to show TODAY, maybe in the next day or two. But holy shit am I blown away. So stoked for this next month. Bout to show some cool things. 😊😊😊

I think one of the best things about AI for this ecosystem is people are starting to see how much potential UEFN has. It's been limited cuz of the heavy UX tolls of the editor. But hopefully with Epic's pivot, these can get good harnesses and improve the editor experience 😊😊😊

Interestingly I've found the opposite. 😅😅😅 Yes AI could write a tool for me to do a thing, but I'm having it opt to use existing tools instead first and leverage features. I'd rather use my tokens directly on my goals rather than having it write and manage its own tools. Libraries, utils, services, etc. ARE still valuable. A lot of people are now sharing tips of "don't let your AI do all the hard work, just do this instead". So learning what exists and having your AI leverage them well can be very valuable. And as AI moves towards more sharing, having our AIs work together to send in pull requests and having their AI with the better context and experience validate it will likely be better. So if you enjoy providing libraries, utils, and middle software that experience and work hasn't gone away. It's likely to continue to grow in value as you automate the maintenance of it all. The same reasons why a lot of products aren't going away either, because maintaining a product is a lot of wasted time, effort, money that keeps you away from your actual goals.

I'll do you one better @ImmatureGamer 😉😉😉 I've got the entire livestream of my first map on the first day of Creative 1.0 😎😎😎 youtube.com/live/STADhyzZiTw…

I actually haven't had this experience really. Just been teaching it from the ground up how to code like I do and it's been phenomenal. 😊😊😊 Worst case is UEFN's launching/loading systems not accurately representing itself 😭😭😭