@theo Of course, you also have to consider where LLMs will be by the time your app is built. That’s the reality now. It doesn’t mean building less, but building differently and exploring new categories of products.
@theo I go with Karpathy’s idea that, before building an app, it’s worth asking yourself: “Could [my app] just be a prompt?” For me this is part of the innovation process.
https://t.co/VhhNXzqQce
@MattiasMagito@Bruno_Ortolland@emmanuel_2m@kiaran_ritchie@Blender Yeah, I think so too. I was just inspired by all those extreme examples people keep posting, where it’s not entirely clear what was actually modeled by Astra and what was added from a model library.
@AIHumanTeam@Bruno_Ortolland@emmanuel_2m@kiaran_ritchie@Blender The prompt was simple: create a highly detailed, game-ready 3D character of Joan of Arc in a new scene, with the quality and visual style of a AAA game character.
Next, I want to try asking it to use polygonal modeling.
"Selected model is at capacity. Please try a different model" because people let Astra find and download hundreds of 3D models from the internet to proof it's 3D modeling capacity.
@Bruno_Ortolland@emmanuel_2m@kiaran_ritchie@Blender This is Fable 5.1 with the exact same prompt. Just a bunch of deformed primitives stacked together.
People posting mind-blowing examples should also share the prompt and the log. It's not like people can't just ask the model how it made it.
@Bruno_Ortolland@emmanuel_2m@kiaran_ritchie@Blender I asked Astra to create an AAA grade character of Joan d'Arc. Then I watched Astra talking about downloading a character base model from Make Human. It did nothing useable with it.
I believe most mindblowing examples are either the model "cheating" or someone else.
The result
@theo I suspect a lot of ai-fluencers talking about agentic coding really just mean copy-pasting between different harnesses and AI models, especially since they rarely talk about what their actual setup looks like.
@NousResearch@cyrilXBT@bot I think people who need an agent for this also need tutorials on how to do it. The information that it’s possible might not be enough.
@OlivioSarikas@Itsfoss Money. Bug reports. Docs. Translations. Spreading the word. From each according to their ability. And you're already making tutorials featuring open-source software. I would assume that counts.
@DavidOndrej1@BasilTheGreat It’s pretty bizarre that you present yourself as a serious AI influencer, yet fail to recognize obvious AI slop the moment it validates your own xenophobia.
@thsottiaux I’d love to see audio file transcription in Codex, and definitely in ChatGPT Work as well. Ideally, you could select a folder containing audio files and have them automatically transcribed via Whisper through a transcription plugin built directly into the app.