This week, a Reddit developer showed off his personal project Neco: a local LLM running on his own computer (using the open-source Ollama and Open WebUI stack), which not only answers questions but also actively "daydreams" and talks to itself, knowing how long its host machine has been running, how much memory remains, and so on. We notice: the project itself is small, but it points to a judgment worth paying attention to — AI may be shifting from an "open and use, close when done" tool into a "tenant living inside your machine."
What this is
Neco has two distinctive features: first, "system awareness" — the model can see its computer's uptime, memory, load, battery, and temperature, but is not authorized to control the machine; second, an "idle daemon process" — a background service that occasionally pipes up with a sentence even when you are not chatting with it. The project is still in the experimentation stage. Next up on the roadmap is long-term memory — letting the AI remember across weeks and months what was discussed and thought about before.
Industry view
Supporters will say this is the right direction for on-device AI (running locally on the device, with nothing uploaded to the cloud) — better privacy, lower latency, and over time real savings on API call costs. But the objections also stand up — first, read-only system awareness is just a handful of preset metrics, nowhere near real "consciousness"; it looks more like a carefully staged demo effect. Second, for most people, "AI speaking to you unprompted" is more likely to feel like an interruption than a delight — your Mac's notification bar is noisy enough already. Third, the open-source community currently has no accepted engineering solution for getting "long-term memory" right; similar attempts on Reddit are mostly half-finished.
Impact on regular people
For enterprise IT: No need to panic in the short term — this remains geek-circle self-driven exploration, at least 18-24 months away from any enterprise procurement list.
For individual professionals: The trend worth watching is — AI assistants are moving from "you ask, it answers" to "it proactively nudges you." Within the next two years, your AI tools will very likely nag you to file that weekly report.
For the consumer market: Apple and Huawei's system-level AI assistants are already heading in this direction. Going forward, your phone and computer may grow increasingly chatty — whether to accept that proactivity is a real user-level question.