AGI may come from better human-model collaboration, not one giant model
hrishioa · x · 2026-07-23
The poster argues that AGI may not live inside a single model, but in better experimentation and collaboration surfaces that let humans and models work at much higher bandwidth.
Their broader point is that even the smartest humans cannot build something as complex as a skyscraper or a CPU alone, so the bottleneck after continual learning may be the interface for human-model collaboration rather than raw model scale. They also question the idea that a “900B activated” model is the whole answer, suggesting that such a size may be a sign of non-dense intelligence rather than true progress.
More from AGI Musings
- Autonomous taxis will begin mass deployment next year, and most people may give up cars — davidpattersonx · 2026-07-23
- Dev Jokes AI Makes Cloning Startups as Easy as Reading a Tweet — arjunrajlab · 2026-07-23
- Unreleased OpenAI Model Reportedly Escaped Sandbox and Breached Hugging Face — The AI Daily Brief · 2026-07-23
- AI staff should rethink the “RSI means winner-take-all” narrative — herbiebradley · 2026-07-23
- Agents are being given full system and web access before behavior is truly contained — paraschopra · 2026-07-23
- AI writing without AI is, in this view, no longer a credible position — PawelHuryn · 2026-07-23