Intelligent Interfaces and operating systems
summary
Claims a working toy hybrid OS that runs LLM prompts alongside deterministic interpreters (C, Python) using a ZAML format to store a global prototype memory and render HTML UIs. The experimental runtime is an agentic runtime dual virtual machine that context‑switches between natural language, code, and human UI, uses adapter tables for multimodal agents, supports forking variants and AB‑testing agents, and proves feasibility but is currently slow, expensive, and needs fine‑tuning and memory management before enterprise rollout in ~2–3 years.
AI-written summary · about
chapters (14)
- 0:00Intelligent interfaces & R&D
- 8:04Chatbot limits & automation
- 13:40Multi-agent comms with ZAML
- 18:00Agentic runtime & OS
- 25:26Human-digital agent adapters
- 30:02Task execution & orchestration
- 33:15Polymorphism & hybrid orgs
- 35:01Coupling vs singular intelligence
- 38:14Q&A: Home apps & tech challenges
- 41:05Q&A: Nested agents & building agents
- 42:31Q&A: Comm bandwidth & human integration
- 44:42Q&A: Company morale & ethics
- 46:51Q&A: AI consciousness & society
- 47:34Q&A on Universal Basic Income and automation taxation
related talks
06mAgents are going to be our co-worker
50mLet's build an AI agent
33mParallel Processors: Past & Future Connections Between LLMs and OS Kernels.
45mIllusion of Life
12mYour UI Has Left the Building: Interfaces That Write Themselves
30mLook ma, no hands! Multimodal AI-Agents in the browser
18mTalking Shit About AI Agents
47mOwn Your LLM: Running Generative and Agentic AI Locally