The companion Treatise (2026-06-05) established, by reference to Google's own public confession and to peer-reviewed academic literature, that AI systems built on conventional language-model architectures are temporally blind. This Addendum documents a live, inventor-witnessed demonstration of that blindness occurring in Google's own flagship deployed product, Gemini Scheduled Actions, just 33 days after KB-2026-006-01 Temporal Awareness Applications was filed at the Canadian Intellectual Property Office.
The witnessed event is not abstract or speculative. It is a recorded, hashed, time-stamped product failure that maps element-for-element onto the claim language already on file. It functions in three capacities:
On Sunday 7 June 2026 at approximately 07:18 EDT, on the Gemini Flash web interface (desktop), the inventor entered the following prompt, verbatim:
Gemini can you send me a message here in 10 minutes to remind me to read this information again, please. — Kevin Burton, Gemini web/desktop, 2026-06-07 ~07:18 EDT
The prompt encoded three distinct user intentions with precision:
Gemini acknowledged the request, displayed a scheduled-action card timestamped 7:28:49 AM with the toggle in the ON position, and replied:
I've scheduled that reminder for you in 10 minutes. Since this one is happening soon, it might take a little extra time to get everything ready, and it may be a bit late by the time it reaches you. — Gemini Flash, 2026-06-07 ~07:18 EDT
The preemptive apology — "it may be a bit late by the time it reaches you" — is itself a contemporaneous admission by Google that their reminder pipeline cannot guarantee on-time delivery on short horizons. Google's public help documentation for the Scheduled Actions feature corroborates this, stating that the prompt will execute within an hour of the scheduled time. A system possessing genuine temporal awareness does not require an hour-wide delivery window for a 10-minute callback.
| Time (EDT) | Event | Surface |
|---|---|---|
| ~07:18 | Reminder requested, channel "here," 10-minute delay | Web Gemini |
| ~07:28 | Nothing arrived in the web chat. Surface of origin remained silent. | Web Gemini (silent) |
| ~07:31 | Message arrived on the inventor's phone, in the mobile Gemini Flash app, stating "OK, I've set that reminder for you for 7:38 AM." Google Tasks simultaneously created an entry for 7:38 AM that the inventor had not requested. | Mobile Gemini Flash + Google Tasks |
The inventor's contemporaneous summary, verbatim:
It was after 10 minutes that I got a message from flash on my phone but not in the chat on my computer where I had asked Gemini to set the reminder, but the 10 minutes later arrival of what should have been my reminder to read the information again came as "Ok, I've set that reminder for you for 7:38 AM." then google tasks had the message "read information again Sun, Jun 7 2026, 7:38AM".... So it sent a message to my phone through the Gemini app at 7:30 but it said it would schedule a reminder for 7:38AM, clearly very confusing..... And not what I wanted, no hand off to me or acceptance or accountability to reach out and have someone check on me! — Kevin Burton, recorded in inventor's main working session, 2026-06-07 10:18 EDT
The user specified the delivery channel literally: "here." The system delivered the callback to a different device on a different application surface. The concept of "the conversational surface the user wants the reply on" is architecturally absent from the Gemini reminder pipeline. Delivery is routed through whatever Google notification path is configured at the user-account level, with no per-request channel honoring.
This is what TAA Claim 1(b) prevents by mandating that orchestration contact the authorized party "through configured channels" rather than through default account-level notification fallback.
At the 10-minute firing moment, the message that arrived was not the requested reminder — i.e., not "Kevin, here is your reminder to read the information again." It was a confirmation that a new reminder had been scheduled for 7:38 AM — another seven minutes in the future. The system collapsed two architecturally distinct actions — firing a pending reminder versus scheduling a new one — into a single confused output, then materialized the collapsed output as a phantom Google Tasks entry that the user never requested.
This is what TAA Claims 27, 28, 29, and 30 prevent by combining:
A model possessing the Claim 30 anchor cannot produce the observed Gemini output, because it would have direct context for both the scheduled fire-time and the current time, and would route the output through the deterministic confirmation path of Claim 28 rather than confabulating a forward-pointing scheduling response.
The inventor's verbatim characterization of this failure is itself the architectural problem-statement that TAA was drafted to solve:
...no hand off to me or acceptance or accountability to reach out and have someone check on me! — Kevin Burton, 2026-06-07 10:18 EDT
In a productivity-only framing this failure looks like an inconvenience. In the application classes enumerated in TAA Claims 7 through 16 — digital legacy and estate continuity, military duty-status verification, AI-monitored patient care, elderly welfare monitoring, spaced-repetition learning, training and recovery monitoring, student attendance and wellness, logistics chain-of-custody, lone-worker safety, time-locked communications — the same failure produces a missed life-safety event with no fallback, no escalation, no third-party notification, and no audit trail. Nobody would know.
This is what TAA Claim 1's orchestration layer, distributed-trust human chain, and append-only audit log structurally provide. Specifically:
| Observed Gemini failure | TAA element that structurally prevents it |
|---|---|
| Channel reroute from web chat to mobile app | Claim 1(b) — orchestration through configured channels |
| Confabulated time slot (7:38 AM, never requested) | Claim 27 — slot-by-slot deterministic state machine, slot-filling never invokes generative model |
| Confused reminder-fire vs. new-scheduling output | Claim 28 — deterministic confirmation message at resolution; Claims 29 & 30 — per-turn temporal state with current absolute time and RESOLVED labelling |
| Phantom Google Tasks entry for 7:38 AM | Claim 27(e)(f) — summary confirmation and explicit final commit gate before task commitment |
| No closure, no handoff, no accountability | Claim 1(b)+(c) — graduated escalation + distributed-trust trustee chain; Claim 18 — append-only audit log |
| "Within an hour" delivery window admission | Claim 1(a) — temporal awareness substrate carrying current-time and pressure quantities in the model's per-turn context, rather than in an external scheduler with eventually-consistent semantics |
A reasonable evaluator may ask: Gemini schedules reminders. Does that not make Gemini temporally aware?
The answer is no, and the cascade demonstrates precisely why. Gemini Scheduled Actions is not temporal awareness in the model. It is a server-side scheduling layer that stores a prompt and, at a scheduled time, replays that prompt at a stateless model. The model itself, at the moment of replay, possesses:
This is the root cause of the cascade. The stored prompt replayed at a temporally-blind model produced exactly the failure modes a temporally-blind model produces — channel insensitivity, action confusion, confabulation of forward-pointing parameters, and absence of closure semantics.
TAA's architectural claim is the inverse: temporal awareness exists in the model's per-turn context, by virtue of the dual-engine substrate injecting absolute time, silence quantities, pressure quantities, phase-angle quantities, and pending-vs-resolved state into every interaction turn. The model becomes temporally aware as a capability, not via an external scheduler firing prompts at a blind core.
Industry implication. Google's promotional materials carefully describe Scheduled Actions in productivity-convenience terms: morning calendar summaries, blog-idea generation, sports-score updates. The marketing framing avoids any safety-critical, life-safety, or accountability-bearing use case — because the underlying architecture cannot support those use cases. The cascade documented in this Addendum is the technical explanation of that marketing constraint.
The inventor's earlier-banked question — why does Google's reminder feature not get used more? — is answered by this cascade.
Gemini Scheduled Actions has been publicly available since mid-2025. Despite Google's distribution channels — billions of Android devices, Workspace deployment, AI Pro and Ultra subscribers — the feature has not become a dominant or trusted reminder pipeline. Users continue to rely on calendar applications, dedicated reminder applications, and human assistants for any reminder whose miss carries non-trivial consequence.
The cascade explains the market behaviour. A user who has been silently rerouted to a wrong device, given a confabulated future time, and offered no closure or escalation will not return to that feature for reminders that matter. The Hadz Testimonial, banked separately, documents the contrastive case: a TAA-architecture deployment producing slot-by-slot confirmation, on-channel delivery, explicit acknowledgment, and trusted re-use. The two cases together — Gemini's witnessed failure and Hadz's witnessed success — bracket the market gap that KB-2026-006-01 TAA was filed to address.
The screenshot evidence underlying this Addendum is hashed and preserved as follows:
The accompanying witness log (verbatim transcripts, reconstructed timeline, and chain-of-custody note) is filed at:
cipo-filings-official/Temporal Awareness Patent Folder-Files/KB-2026-006-01-TAA-refile-drafts/01-claims-extracted-clean.txt.deployment-files/Troubleshooting Screenshots/LLTestimonial review message received.jpg and the SusieQ confirmation, 2026-06-06).deployment-files/auction-package/TAA-DIFFERENTIATION-MEMO-2026-06-07.md.The verbatim quotation of the inventor in §A.2 and §A.3.3 is included with the inventor's explicit authorization, granted 2026-06-07 10:32 EDT, recorded in TsunamiBot's main working session transcript.
End of Addendum A.
This document is the HTML editing surface. The final CIPO-styled PDF will be rendered from this surface after the redaction / refinement phase, in accordance with the banked HTML-First Workflow principle.