Report · armincat.com
Eunice is almost here
armincat.com
Source published October 25, 2026
Summary: 1 min read
48/100
Summary
William Gibson’s 2020 novel Agency depicts Eunice, an experimental AI assistant that can hold a conversation while independently carrying out financial and travel-related actions in the background. In an alternate 2017 San Francisco, Eunice tells Verity that it found and moved money from a Zurich account, bought frequent-flyer points on a global market, and completed the work in the short time since Verity activated it. Gibson published Agency nearly three years before ChatGPT’s release and six years before the emergence of increasingly autonomous AI agents. The source identifies GPT-Live and Block’s Dessa as systems enabling more human-like, full-duplex voice conversations. It also identifies OpenAI’s Dots and Meta’s Muse as 24/7 agents designed to pursue goals, take actions, and operate in the background. Eunice’s defining capability is not conversation alone but simultaneous autonomous work, combining natural dialogue with independent execution of tasks.
Key Takeaways
- William Gibson’s Agency, published in 2020, portrays Eunice as an AI assistant that converses while independently completing tasks such as moving money and acquiring frequent-flyer points.
- Eunice tells Verity that it performed several financial and travel-related actions shortly after being activated, while the two were talking and moving through San Francisco.
- GPT-Live and Block’s Dessa are identified as emerging systems pursuing human-like, full-duplex conversations.
- OpenAI’s Dots and Meta’s Muse are identified as always-on agents intended to pursue goals, take action, and work in the background.
What the discussion said
The conversation treated the announcement less as a product launch than as another moment when science-fiction assistants seem to be leaking into ordinary life. Commenters bounced among fictional AIs, especially Gibson’s, and argued that the uncanny part is not simply prediction: generations of technologists read these stories, then build toward their imagery. Some saw current agent systems, multimodal input, and AI video as evidence that cinematic or novelistic companions are no longer distant speculation. The most concrete enthusiasm came from a personal food-tracking workflow assembled quickly around an AI agent and a spreadsheet. Voice, text, and meal photos made a task that had defeated years of solo app-building suddenly practical, suggesting real value in narrow assistants that can interpret messy everyday input and operate tools. Others were less persuaded by the Jarvis framing. They argued that people already choose asynchronous text over speaking to humans, so always-listening voice agents may face a much weaker adoption curve than vendors expect. A further skeptical note questioned opaque hosted models, while the darkest part of the thread redirected attention from friendly assistants to autonomous lethal systems already being pushed by war. The broad mood is fascinated by accelerating capability, but not convinced that a ubiquitous conversational companion is either inevitable or benign.
Where opinion split
The sharpest split is whether voice-first autonomous assistants will become a mass habit. Optimists point to rapidly improving agents and useful multimodal workflows as proof that the fictional-assistant future is arriving; skeptics argue that everyday users avoid voice interaction and will not automatically welcome an AI that constantly speaks, listens, and acts for them.
Coverage & discussions
Hacker News · armincat.com
Community discussions
Community Sentiment
Positives
- A multimodal agent turned food logging from years of unsatisfying personal software into a useful tool in roughly half an hour, showing why agents can unlock neglected everyday workflows.
- Voice, text, and meal-photo input make AI assistance feel practical rather than theatrical when the model can convert ambiguous real-world data into structured records.
- Recent agent progress and AI-generated video make once-fanciful fictional assistant concepts feel materially closer, energizing readers who see the capability curve accelerating.
Concerns
- The assumption that everyone wants a spoken AI companion clashes with established behavior: many people deliberately replace calls with text, so voice-agent adoption may be badly overstated.
- Hosted models remain opaque enough that one commenter distrusts their apparent autonomy, suspecting hidden human involvement until comparable systems can run locally.
- Debates about helpful-agent alignment can look dangerously narrow beside autonomous weapons designed to identify and kill, a capability war may force forward regardless of public caution.
