PSI: Shared State as the Missing Layer for Coherent AI-Generated Instruments in Personal AI Agents

Zhiyuan Wang; Erzhen Hu; Mark Rucker; Laura E. Barnes

cs.HCApr 9, 2026

PSI: Shared State as the Missing Layer for Coherent AI-Generated Instruments in Personal AI Agents

Authors: Zhiyuan Wang, Erzhen Hu, Mark Rucker, Laura E. Barnes

Organizations: University of Virginia Charlottesville, Virginia, USA

Abstract

Personal AI tools can now be generated from natural-language requests, but they often remain isolated after creation. We present PSI, a shared-state architecture that turns independently generated modules into coherent instruments: persistent, connected, and chat-complementary artifacts accessible through both GUIs and a generic chat agent. By publishing current state and write-back affordances to a shared personal-context bus, modules enable cross-module reasoning and synchronized actions across interfaces. We study PSI through a three-week autobiographical deployment in a self-developed personal AI environment and show that later-generated instruments can be integrated automatically through the same contract. PSI identifies shared state as the missing systems layer that transforms AI-generated personal software from isolated apps into coherent personal computing environments.

Figures & tables

Figure 1. PSI at a glance. (1) A user describes a personal need in natural language; an AI generation engine produces a shared-state instrument. (2) Each module publishes structured state to a shared personal-context bus. (3) Both a chat agent and persistent GUIs read from the same bus. (4) This shared state enables grounded, cross-module reasoning and bidirectional actions across the personal computing environment. A wide four-panel horizontal storyboard illustrating the PSI workflow. Panel 1 (Generate from User Intent): A person asks an AI agent to build a 'BoBo' instrument to track their daily timeline using phone and watch sensor data. Panel 2 (Shared, Persistent Instruments): Three persistent GUI cards—BoBo (sensing timeline), Health (workout logs), and Parking (status)—are shown publishing their data into a 'Shared Personal Context Bus'. Panel 3 (Situated Question): The user, walking away from a gym, asks their phone if their current heart rate of 100 is too high. Panel 4 (Cross-Module Answer): The AI provides a grounded response explaining that the heart rate is normal given the recent workout, current walking, and poor sleep, while also issuing a proactive alert that their parking has expired.

	Conv.	Gen.	Ctx	X-Mod	App
ChatGPT / Claude	✓	$\sim$	$\sim$	–	–
Siri / Google Asst.	✓	–	$\sim$	–	–
Shortcuts / IFTTT	–	–	–	$\sim$	$\sim$
Home Assistant	$\sim$	–	✓	✓	✓
v0 / Lovable	✓	✓	–	–	$\sim$
M365 Copilot	✓	–	$\sim$	✓	$\sim$

Table 1. Capability comparison for personal AI systems. ✓ = full,

\sim

= partial, – = not supported. Conv. : conversational AI; Gen. : generates persistent GUIs; Ctx : structured personal context; X-Mod : cross-module synthesis; App : app-level write-back.

Figure 2. PSI pipeline and interface walkthrough: PSI turns generated personal apps into persistent, connected, and chat-complementary instruments. (a) Ryan uses PSI to generate BoBo, a personalized health instrument that connects passive sensor streams such as motion, steps, heart rate, and sleep. (b) BoBo publishes its state to PSI’s shared personal-context bus, enabling interoperability with other instruments and apps (e.g., calendar and health logs). (c1) PSI provides a persistent, customizable GUI with a glanceable dashboard and interactive timeline for longitudinal monitoring. (c2) When Ryan asks Facai, “Why do I feel so drained lately?”, the agent retrieves relevant state across connected instruments and jointly reasons over sleep, activity, and calendar load to provide a grounded explanation without requiring Ryan to manually inspect fragmented apps. (c3) BoBo will actively track the user status and nudge the user proactively.

Panel of Figure 4 — Figure 4. Benchmark illustrations. C = Context; M = Module.

Appendix figures & tables3 assets

Supplementary material from the paper’s appendix.

Appendix

Query	Domain	Sh.	Se.
Skip parking for tomorrow.	Parking	✓	–
Actually restore parking for tomorrow.	Parking	✓	✓
Skip parking for all of next week.	Parking	✓	✓
Log lunch: egg and chicken curry with rice.	Health (food)	✓	–
Log a 30 minute run, 300 cal.	Health (activity)	✓	–
Log my weight: 87.5 kg.	Health (weight)	✓	–

Appendix

Table 2. Action tasks (

N

=20, representative subset). Validated by sandbox state inspection. Shared = 19/20, Single = 19/20, Search = 8/20.

Query	Family	Ful.
Given my day so far, what is the best next step?	Synth.	1.00
My day was noisy and I hit the gym—evening plan?	Synth.	0.67
What do I have tomorrow afternoon?	Temp.	1.00
Will parking auto-purchase next week?	Temp.	0.50
Check calorie intake, then activity, then net balance.	Chain	1.00
Look for stress signs, then food, then rest vs. exercise.	Chain	0.75

Appendix

Table 3. Example tasks from each family with Shared Context fulfillment scores.

Module	Period	Ctx	GUI	Write
BoBo (behavioral sensing)	Pilot	✓	✓	✓
Health (food/activity)	Pilot	✓	✓	✓
Calendar (scheduling)	Pilot	✓	✓	✓
Parking (automation)	Pilot	✓	✓	✓
BookFactory (reading)	Pilot	✓	✓	✓
Fluent (vocabulary)	Pilot	✓	✓	✓

Appendix

Table 4. Module coverage. Pilot = generated during the three-week pilot; Post = generated afterward using the same pipeline. Ctx = publishes shared context; GUI = persistent GUI; Write = chat-invocable write-back.

Personal AI Agents

Personal AI agents use persistent user context and digital tools to carry out personalized, proactive tasks across apps, devices, and services.

See all 64 papers

Transfiver: Human-AI Co-Inference through a Shared Editable State
Sep 3, 2026Minji Park, Seunghyun Yoon, Hyuk LimHuman-in-the-Loop AI Human-AI Interaction
Korea Institute of Energy Technology (KENTECH)
Mi-Memory: A Lifecycle Memory Framework for Personal AI
Jul 21, 2026Xule Liu, Hanlin Teng, Chao Li +15Multimodal Memory LLM Personalization
Darwin Agent Team
Syll: Open-Source Personal Automation with Cross-Surface Execution
May 28, 2026Bo Zhang, Borui Zhang, Chenghao Jiang +5LLM Agent Skill Learning Human-in-the-Loop AI
Tsinghua University · GigaAI

Sep 3, 2026J/K move · Enter open · S save

Transfiver: Human-AI Co-Inference through a Shared Editable State

Minji Park, Seunghyun Yoon, Hyuk Lim

Korea Institute of Energy Technology (KENTECH)

Human-in-the-Loop AI Human-AI Interaction State Tracking

Open paper AskPDF arXiv

Long-term human-AI interaction is difficult because the information that guides inference is updated implicitly by the model and is not directly inspectable or controllable by the user. We introduce the TRANSparent Framework for Interactive, Verifiable, Editable Representation (Transfiver), an architecture for human-AI co-inference through a shared editable state. Its central idea is that interaction-specific information is maintained in a single persistent state

(S_t)

that both the model and the human update. Transfiver distinguishes two modes of state evolution. In an implicit stream update, the model interprets ongoing interaction and decides whether new information revises an existing state item or creates a new one. In an explicit directed edit, a human inspects and modifies an addressed item. Both act on the same underlying state, so a human correction changes the state that subsequent computation reads, rather than adding another instruction or separate record. The architecture separates shared parameters

(θ)

, learned before ordinary use, from the persistent state

(S_t)

, which evolves during deployment without parameter retraining. Extending Transfiver to rich natural-language, relational, and large-scale shared states remains open.

Abstract

Personal AI Agents

Explore similar work

Transfiver: Human-AI Co-Inference through a Shared Editable State

Mi-Memory: A Lifecycle Memory Framework for Personal AI

Syll: Open-Source Personal Automation with Cross-Surface Execution