I'm still relatively new to Claude Code, but I've had it working pretty hard on coding. I am running just one agent and I have rarely seen it make mistakes. Usually they are self caught in testing.
To catch the rare miss, I have a custom skill, that I simply load in a fresh session, that scrubs the code for errors, inefficiencies, or bad code hygiene, passing a report to me for fix delegation.
I like the idea mentioned here, but it feels overpowered and designed to eat through tokens. Not surprising for the CEO of a company that sells tokens to suggest a way you can triple your usage of said tokens....
Has anyone, who actually uses Claude Code regularly, run into a job so complex that it would warrant this level of overkill?
@WlkrSeattleRngr@CSI_Starbase New heat shield. They were actually trying to get this shield to survive. The rotation was to let the cameras see more of the shield.
@ResoluteTitan88@RedyAu_@Truthful_ast Definitely an area that needs to be developed. Ironically, the AI satellites that will soon add to the space clutter will likely enable AI to manage all of that without the possibility of humans error.
As for moving satellites, many don't have propulsion at all. Most of those that do have propulsion use hypergolic fuels for small adjustments. Conversely, SpaceX's Starlink satellites all utilize ion engines to give them years of fuel for orbital adjustments and, when their lives end, they use those engines to deorbit in a planned way. This should become the standard.
Everything, all at once. One starship can carry 20x the mass of a Falcon 9. Every space mission is limited by the size of the vehicle that sends it to space. That size limit just went WAY up. The cost per lb is going to plummet as well. You could send a giant satellite, or dozens of small ones for the cost of a single launch today.
@SFWsalesgirl@elonmusk Pretty sure you could always get a Her to the Empire trailer from AI. Grok could do it with a lot of hand holding. Veo could probably fool you into thinking it was real
AI is unfortunately not capable of accurately analyzing photos yet. Frontier AIs, like Grok, are Language Learning Models (LLMs). They seem smart because they specialize in mimicking human speech, and analyzing written word. But they don't have the physical context, or the computing power, needed to consistently understand images.
Skills is kinda in the backend on Grok Android but definitely need more work. Grok often hallucinates while using skills, in my experience. If I tell it to store info in a file for future use, it will pretend to do so, but.ssve nothing. If you ask it about the info in the very next message it looks at you funny and asks "what info?" Lol, quite infuriating.
They'll never do it. The image uploading is what got them slammed for deep fakes. Basically any edits you do to an uploaded image have to be approved by the Pope or it gets moderated. Now, if we're lucky, they'll finally learn how to tell the difference between an uploaded photo and one they've created themselves. Then they can relax a bit as long as you are only animating characters generated by Grok.
@Lil__Alien_@elonmusk iPhone has the lowest restrictions. Web is oddly second lowest. Android is highest because Google has aggressive rules and considers kissing to be foreplay.
@elonmusk Just don't try to make any of your characters kiss.... Remember, it only does stuff that would be allowed in a rated R movie. Any contact, of any object, with a face is Triple X material, every time.