Screenless living: one voice agent orchestrating all frontier models end to end
EricBuess · x · 2026-10-05
Eric Buess describes his "screenless" experiment: he never touches a computer, tablet or phone, talking instead to a single voice assistant, Titus, which orchestrates all frontier models as one distributed system to autonomously complete complex, long-running tasks.
Key points:
- Titus runs all his digital surfaces — computers, websites, phones, services — entirely via two-way voice through one app.
- He measures the orchestration by time from task given to task landed, and tasks landed per hour/week.
- His motivation: using AI to keep people aligned with the version of themselves they aspire to be, simulating an end-of-life retrospective to minimize regret.
- The post is his reply to Andrej Karpathy on just-in-time understanding, with a six-minute film clip and full essay.
An extreme personal experiment, but a concrete demo of voice-first multi-model agent orchestration.
More from coding & agent
- Perplexity Computer builds a GeoGuessr agent that pinpoints photo locations with 3D globe — AravSrinivas · 2026-10-05
- Addy Osmani: give your coding agent ways to check its own work — addyosmani · 2026-10-05
- Grok's honest review of FLUJO: an ambitious local-first MCP platform with bus factor 1 — Ambitious-Prompt-975 · 2026-10-05
- CMU ships cua-speedrun: standardized benchmark finally measures computer-use agent speed and cost — rsalakhu · 2026-10-05
- Ex-Bing chief Mikhail Parakhin: LLMs inherit mediocre code taste from pretraining — MParakhin · 2026-10-05
- exe ships a human-friendly API explorer, betting humans still matter in the agent era — davidcrawshaw · 2026-10-05