If you strip away the hype, there are no “AI agents.”
There are just models: GPT-2 → GPT-3 → GPT-4 → GPT-5.
Each generation gets better at understanding context within its window.
What people call “AI agents” today is just:
model + instructions.
By the time we reach GPT-5 (what we now see in GPT / Claude / Gemini / Grok),
models are already capable of doing work that used to require a mid-level programmer, a secretary, or even a skilled employee.
The real shift isn’t agents — it’s capability.
With GPT-2 vs GPT-3, the difference was how long humans could stay engaged — more messages, longer interactions.
✨Thinking with Blender~
Meet VIGA: a multimodal agent that autonomously codes 3D/4D blender scenes from any image, with no human, no training!
@berkeley_ai#LLMs#Blender#Agent 🧵1/6