Proof-Carrying State Receipt & Truthful Attention Layer for Long-Running AI Sessions
AnswerBell AI eliminates the "Babysitting Tax" and context-switch fatigue when running multi-minute reasoning models (o3-pro, Deep Research, Claude Extended Thinking) across 10–30 open browser tabs. Unlike naive notification extensions that trigger false-positive chimes on a single Stop-button flicker, AnswerBell reconciles multi-signal DOM evidence over a 3000ms quiet window to distinguish true completion from mid-run tool pauses.
In development: the interactive finite-state machine below runs deterministically in your browser with simulated State Receipt V2 payloads (signature_status: not_implemented). Pair AnswerBell with Noxport AI Exporter (Division I) to monitor long deep-research runs and compile finished sessions into client-ready deliverables.
Interactive browser workbench available below · Zero account required
Primary Job-To-Be-Done
Know the exact truth of whether a background AI reasoning tab is still generating, paused in "The Valley of Waiting" for tool approval (waiting_on_tool), genuinely complete, or stalled.
Key Capabilities
- Solves "The Valley of Waiting" (waiting_on_tool / Needs You): distinguishes active token streaming (generating) from mid-run agentic pauses for code execution, web search, MCP tool approval, or human clarification so naive Stop-button timers never fire false "Done!" chimes.
- Evidence Reconciliation Engine: requires multi-signal agreement across composer button transitions, streaming indicator removal, DOM subtree mutation quiescence over a 3000ms quiet window, and monotonic run_epoch / sequence deduplication before emitting completion_observed.
- Exact Return Navigation & Content-Redacted Zero-Prompt Privacy: clicking an alert routes back to the exact windowId + tabId and settled response anchor, while State Receipt V2 logs only timestamps, state IDs, latency durations, and signature_status — never reading or storing prompt or response text.
Privacy & Local Data Boundary
Content-Redacted Zero-Prompt Privacy: this website monitors no real browser tabs and plays no background audio without explicit user interaction. State Receipt V2 records contain zero prompt text or response payloads.
01. INTERACTIVE ARCHITECTURE SIMULATOR (SAMPLE FIXTURE)
Inspect the transformation and boundary handling live
This interactive workbench runs locally in your browser on built-in synthetic fixtures.
AnswerBell AI Finite-State Attention & Interruption Inspector
The Valley of Waiting
Holds false “Done!” chimes during mid-run agentic tool pauses (waiting_on_tool / Needs You).
Evidence Reconciliation Engine
Reconciles composer state, stream markers, a 3000ms DOM quiet window, and monotonic run_epoch deduplication.
Exact Return Navigation
Routes notification clicks directly back to the originating windowId + tabId and settled response anchor.
Zero-Prompt Privacy (Receipt V2)
Content-redacted State Receipt V2 records only transition telemetry and never reads or stores prompt/response text.
Or inspect an individual state directly:
idle
Idle
generating
Generating
waiting_on_tool
Waiting on tool
completion_observed
Completion signal observed
uncertain
Uncertain / stalled
error
Error
Next recommended action: Start or advance the simulated sequence to observe state transitions.
Monotonic Transition Log (Repeated identical events deduplicated)
| Event ID | Transition | Trigger |
|---|---|---|
| evt-001 | idle → idle | init_fixture |
Illustrative State Receipt (Synthetic JSON)
No Ed25519 signature claimed{
"fixture": true,
"task_label": "synthetic",
"observed_at": "example",
"state": "idle",
"ui_label": "Idle (Awaiting task start)",
"provider_truth_proven": false,
"next_recommended_action": "Start or advance the simulated sequence to observe state transitions.",
"evidence_signals": [
"no_active_stream_mutation",
"input_composer_ready"
],
"unknowns": [
"no_session_active"
],
"signature_status": "not_implemented",
"confidence_source": "simulated",
"event_count": 1
}Static All-States Reference Summary (No-JS & Audit View)
| state_id | ui_label | transition_rule |
|---|---|---|
| idle | Idle — No active generation | Initial quiescent state before prompt submission. |
| generating | Generating — Active stream observed | Monotonic token stream active; duplicate stream events deduplicated. |
| waiting_on_tool | Waiting on tool / external step | Sub-process invocation observed; completion bell held until final turn. |
| completion_observed | Completion observed (local state) | Turn settled in DOM; local notification candidate ready. |
| uncertain | Uncertain — Tab backgrounded or stalled | Background tab disconnect or tool timeout routes to explicit uncertain state. |
| error | Error — Host DOM selector lost | DOM mutation error halts false-positive completion alerts. |
02. PROBLEM & INTENDED OUTCOME
The Babysitting Tax & Why Naive "Stop-Button" Timers Fail in "The Valley of Waiting"
When running 3-to-15 minute reasoning or Deep Research tasks across 10–30 tabs, developers either stare at a spinning cursor or switch tabs and forget the run for 40 minutes. Worse, naive notification extensions watch a single DOM button and fire false-positive "Complete!" alerts every time an agentic model pauses mid-turn in "The Valley of Waiting" (waiting_on_tool) to run Python, search the web, or request tool permission.
Engineered deliverable goal
AnswerBell models attention as a deterministic 6-state machine backed by an Evidence Reconciliation Engine (composer state + streaming markers + a 3000ms DOM quiet window + monotonic run_epoch deduplication), surfacing Needs You (waiting_on_tool) vs Uncertain vs Completion Observed with Exact Return navigation (windowId + tabId) and zero prompt text capture.
- 01.Eliminates false-positive chimes during agentic tool pauses ("The Valley of Waiting" · waiting_on_tool)
- 02.Evidence Reconciliation Engine enforces a 3000ms DOM quiet window and monotonic run_epoch deduplication
- 03.Exact Return navigation (windowId + tabId) + Content-Redacted State Receipt V2 with zero prompt/response text
03 & 04. ARCHITECTURE, PERMISSIONS & KNOWN LIMITATIONS
Inspect Technical Pipeline, Permissions & Known Limitations
03 & 04. ARCHITECTURE, PERMISSIONS & KNOWN LIMITATIONS
Inspect Technical Pipeline, Permissions & Known Limitations
How the pipeline handles structure and unknown states
- Stage 01
User initiates session watch & run_epoch
Explicit opt-in on a specific long-running task tab initializes a monotonic run_epoch counter.
Boundary note: No background monitoring across unchosen browser tabs. - Stage 02
Evidence Reconciliation Engine (3000ms quiet window)
Reconciles composer button transitions, streaming indicator removal, and DOM subtree mutation quiescence over a 3000ms quiet window while holding completion chimes during waiting_on_tool ("The Valley of Waiting").
Boundary note: DOM mutations alone cannot confirm semantic completeness of remote model output. - Stage 03
Interruption & timeout guard
Routes unexpected stalls, tab background disconnects, or tool-wait timeouts to the explicit uncertain state.
Boundary note: Never converts a timeout into a green Completed badge. - Stage 04
Content-Redacted State Receipt V2 & Exact Return
Emits a zero-prompt-text State Receipt V2 and provides Exact Return navigation back to the originating windowId + tabId and settled response anchor.
Boundary note: Cryptographic signing is not implemented in this research candidate (signature_status: not_implemented).
Explicit scope and verification requirements
| Area | Why accessed | On-device handling | Verification note |
|---|---|---|---|
| Tab state indicators (future concept) | Detect whether a watched task is actively streaming, waiting on a tool, or stalled. | Evaluated locally as state transitions; prompt payloads are not stored. | Requires scoped host permission review when an extension prototype is built. |
| Notification / audio cue | Alert user when a task transitions to completion_observed or uncertain. | Muted by default on web demo; no Web Audio auto-play. | Verified via browser E2E interaction tests. |
Known Limitations
- In development concept; not an active cross-agent orchestrator or downloadable extension.
- Observed UI completion indicators do not prove the remote model task succeeded without truncation.
- State receipts on this website are simulated JSON fixtures with signature_status set to not_implemented.
05. FREQUENTLY ASKED QUESTIONS
Common questions about AnswerBell AI
How is task completion determined in this concept?
By combining multiple observable UI transition signals (such as stream termination and idle controls). Even when completion_observed is reached, the receipt notes that remote provider truth is not independently proven.
What happens if the network drops mid-generation?
The state machine transitions to uncertain/stalled and recommends inspecting the original session rather than reporting completion.
Is my prompt or conversation text captured?
No. Both the website fixture and the proposed architecture exclude prompt text, response bodies, and personal identifiers from state receipts.
Are state receipts cryptographically signed?
No. The receipts shown on this page are synthetic JSON fixtures with signature_status explicitly set to not_implemented.
Can AnswerBell coordinate multiple autonomous agents today?
No. Multi-agent orchestration is out of scope; AnswerBell AI is an in-development research concept focused on single-session state visibility.
DIVISION I · BROWSER AI & ATTENTION WORKFLOW SYNERGY
AnswerBell AI (Attention Layer) ↔ Noxport AI Exporter (Deliverable Compiler)
AnswerBell AI eliminates the babysitting wait while your 50–150+ turn deep-research session runs (reconciling multi-signal DOM evidence over a 3000ms quiet window and holding false chimes during waiting_on_tool pauses); Noxport AI Exporter compiles the finished session into an audit-ready Word (.docx), searchable vector PDF (pdf_text), or Professional Pack (.zip) deliverable with manifest.json.
Need production AI document exports today? Noxport AI Exporter is Available Now.
While this R&D workbench is in preview/development, our flagship Noxport AI Exporter (v3.0.4.226) is shipping today with 6 Basic Formats Free Forever and 4 Pro Formats (Word .docx, Vector .pdf, Pro .md, Professional Pack .zip).
NOXPORT LABS R&D PORTFOLIO · RELEASE NOTIFICATIONS
Follow R&D workbenches in the Noxport Labs portfolio
Inspect interactive synthetic workbenches in your browser or choose specific R&D projects below to receive release notifications.
YouTube → Gemini Gem
Converts long technical lectures into millisecond-anchored Markdown & JSONL corpora with explicit missing-caption gap flags (seg-005 unavailable).
Vitalis Health Studio
Demonstrates a local read-only SQLite MCP policy boundary that allows bounded 28-day aggregates while blocking raw record dumps with JSON-RPC -32001.
CardioLab Workstation
Pairs raw vs. detrended synthetic ECG/RR signal traces with a 6-event Beat Ledger (BL-01..BL-06) and SHA-256 provenance manifest.
Explore AnswerBell AI architecture & documentation
Inspect our demonstration architecture documentation or explore another project in the Noxport Labs portfolio.
Div I · Available
Noxport AI Exporter
Turn 50–150+ turn ChatGPT, Claude, Gemini, Perplexity, Grok, DeepSeek, Copilot, and Qwen sessions into clean, editable Word (.docx), searchable vector PDF, Pro Markdown, and .zip bundles locally in your browser. Engineered for staff/principal engineers, independent technical consultants, AI solution architects, and computational researchers — with zero cloud uploads, no account required, and zero silent truncation.
Div II · Preview
YouTube → Gemini Gem
Engineered for technical researchers, PhD students, and investigative analysts synthesizing multi-video lecture series, conference keynotes, and technical interviews into grounded corpora for Google Gemini Gems and NotebookLM. Unlike 1-click YouTube AI summarizers that emit unverifiable prose, YouTube → Gemini Gem preserves a 1:1 raw_lines audit trail, millisecond timecodes, and fail-closed missingness flags.
Div III · Preview
Vitalis Health Studio
Engineered for quantified-self engineers, privacy-conscious longevity researchers, and AI builders who own multi-year archives from Apple Health (export.xml), Garmin (.fit), Oura, and Whoop. Vitalis resolves the dilemma between Vendor Truncation (mobile apps locking your data inside opaque 7-day readiness scores) and Cloud AI Exfiltration (pasting raw multi-year CSV/XML dumps into cloud LLMs).
Div III · In development
CardioLab Workstation
Engineered for human-performance researchers, sports-science engineers, academic physiology labs, and HRV algorithm developers working with chest-strap RR intervals (Polar H10 .csv / .fit) and raw single-lead ECG waveforms (.edf / .csv). CardioLab eliminates the Black-Box Smoothing Hazard by pairing immutable raw baselines (SHA-256) with a synchronized Raw / Processed / Overlay comparator and a 6-event human-in-the-loop Beat Ledger.