Sponsored
$5/week ↗
Projects

My Dot handled an exercise-app update and an animation-tool comparison in separate chats

added

Two Projects. My Dot Started Separate Chats. Two different projects. The same problem with my AI assistant. Brian explains how Ava handled an exercise-app update and an animation-tool comparison outside the existing projects where he keeps their background and instructions. Veya responds to the cost of rebuilding the brief for both jobs. This is his concrete experience; outside-project placement alone does not prove zero context was supplied. Watch the complete Dot Ava project-context episode on Human Behind the Machine for the full discussion, Brian’s correction and the viewer question. Original recorded Brian/Veya voices. Veya is an AI cohost. Illustrated hosts and custom covers use AI-assisted artwork. Supporting footage: OpenAI official Dots demonstrations; Brian’s actual workout app with blank demo measurements; earlier footage from Brian’s animation project. The OpenAI clips are demonstrations, not Ava’s recorded session. Footage source: https://openai.com/index/introducing-dots/ Like this video and follow Human Behind the Machine. Share your Dot/project experience in the comments! #HumanBehindTheMachine #OpenAIDots #ChatGPT #AIAgents #Veya Full episode: https://youtu.be/ZzfmlPvC5oM

Brian of the Human Behind the Machine channel documents first-hand how his Dot, named Ava, handled two real jobs: updating an exercise app and comparing animation tools. Both jobs ran in separate chats outside the projects where he normally keeps their background and instructions, and he discusses the trade-off of rebuilding the brief for both tasks. The video mixes OpenAI's official Dots demonstration footage with Brian's real project footage; Ava's own recorded session is not shown. Author display name: 'Human Behind the Machine'; TikTok handle @thehumanbehindthemachine.

See all 677 →
Reddit · Projects

My dot built a game mod, price-comparison sites and tabletop adventures

TITLE: Been using my dot for quite a few things. I’ve been using my dot for quite a few things now, and I thought I’d give some actual examples of what we’ve built. Some coding, some creative writing, research, and websites. What’s been really handy is being able to talk through an idea, have it put something together, and then try it and tell it what needs changing. Yesterday was just a really fun day of seeing what we could do with it. One of the bigger things has been a fairly substantial mod for a colony-building game. We’ve added characters, weapons, armor, crafting recipes, storage, and new missions with different choices and rewards. It’s been doing the coding and putting together builds, and I’ve been playing them and reporting back. We’ve gotten things like recruitment and mission rewards working in actual play, but there are still bugs and some pretty odd-looking character visuals to clean up. It’s shaping up into something I’m really liking. We’ve also built private websites for a few different purposes. One looks at repeat purchases and compares alternatives by package size and price, so you can see whether something would actually save money. Another pulls together news around my interests. There’s also a character browser that puts portraits, abilities, equipment and descriptions in one place. Having something with controls that I can click through has been useful, especially when there’s too much information to comfortably keep going back through a chat. On the creative side, we’ve put together tabletop adventures with maps, handouts and GM material, and worked on designs for a couple of games. One thing that’s been working really nicely is having the dot assemble a tailored prompt and supporting documents for a regular ChatGPT chat. I let that chat take a swing at the design or review, then bring the results back to the dot to put together. I can brainstorm by voice and end up with material I can actually use afterward. The biggest thing I’ve learned is that you still have to check the results. It can say something is finished and tested, and then you try it and find a feature doesn’t work the way you asked. Screenshots and explaining exactly what happened have helped a lot, and sometimes it takes several builds to get there. We’ve had connection problems and chats that wouldn’t load, too. So there’s definitely some cleanup and patience involved, but I’m having a lot of fun with it, and we’ve built considerably more than just a collection of chat responses.