Why `D20` cannot stay a delay-only anchor
| `Phase-targeting field` | What the packet now has to say explicitly | What goes wrong if omitted |
|---|---|---|
| `Oscillation gate` | Which band, channel / spatial filter, power / SNR threshold, phase-reset rejection, and no-stim or missed-trigger rate defined an eligible REM epoch. | A fast trigger is overread as if the target oscillation had actually been estimable throughout the session. |
| `Targeting metric` | Which circular phase-error metric, off-target comparator, and causal-versus-post-hoc benchmark were used under the disclosed REM protocol. | Low mean latency is mistaken for correct phase targeting. |
| `Downstream effect` | Which outcome changed: power, frequency, physiology, behavior, or none, and on what time window. | Targeting success is silently promoted to physiological or behavioral efficacy. |
| `Stability policy` | Whether the preferred phase stayed fixed, drifted, or was updated within or across sessions, and how abstention was handled when the target weakened. | One locally useful phase is overread as a stable controller across the whole route. |
## What the first artifact can honestly close now
| Artifact piece | What it closes now | What still stays outside the claim |
|---|---|---|
| `Closed-loop safety KPI dashboard` | A bounded dashboard that fixes `stability rate`, `stop-intervention rate`, `recovery time`, and declared end-to-end timing percentiles for one named loop family under one disclosed protocol. | Any statement that one dashboard already gives a universal closed-loop threshold across loop types. |
| `Phase-targeting log` | A bounded log that fixes `oscillation gate`, `no-stim / missed-trigger rate`, `circular phase error`, `off-target comparator`, and `effect readout` for the present REM targeting protocol. | Any statement that a fast loop or a low average phase error already proves reliable phase-specific control or stable downstream benefit. |
| `Delay-tolerance band note` | A narrow rule for which timing deterioration begins to break the present loop class under the present protocol, oscillation gate, and stop logic. | Any statement that a single ms threshold generalizes across neurofeedback, P300 control, speech BCI, phase-targeting, and chronic adaptive stimulation. |
| `Downgrade-rule note` | A public-safe rule that keeps the row at a local-controller result whenever stability drops below the predeclared floor or recovery time drifts beyond the operating band. | Any interpretation that a fast or stable local loop is automatically boundary-complete, fixed-decoder durable, chronically deployable, or identity-relevant. |
## Funding wording that still fits
| Route | Submission wording that stays honest | Why it still fits this row |
|---|---|---|
| `Japan Industrial Science Research Institute research grant` (`watch`) | `Operational delay-tolerance and safety-stop dashboard for closed-loop EEG control in the mind-upload route` | The row reads naturally as an engineering-style operations and safety package rather than as a stronger neuroscience or mind-preservation claim. |
| `Nakatani Foundation research grant` (`next-cycle watch`) | `Control-theoretic delay-tolerance benchmark for closed-loop EEG safety and retraining boundaries` | It still fits as a bounded BME-facing validation package, but only as a watch route because the current cycle is closed. |
## External dependency tasks that stay outside this session
- Generalization across loop classes still remains outside the present packet.
- Phase-targeting estimability, downstream effect, and preferred-phase stability still remain separate evidence objects even inside the current REM family.
- Fixed-decoder durability, rescue-mode recalibration burden, and clinic/home transfer still remain separate evidence objects rather than completed claims of this row.
- Any interpretation that one closed-loop timing result proves a boundary-complete controller, a stable preserved self, or a deployable WBE loop remains outside the present route.
## What changed in this run
- The public batch did not change.
- The main change is that `U8-1` now has a public packet that records the cleanest public-open anchor, the first safety artifact, and the watch-level funding bridge after the closed-loop timing page was strengthened.
- This remains a route-setting update, not a solved-claim update.
`D20` is not merely a generic fast-loop dataset. The linked REM paper, Jaramillo et al. (2024), delivered auditory stimulation phase-locked to alpha or theta oscillations during REM sleep and reported phase-dependent power and frequency changes. That makes the route scientifically narrower and stricter than a generic latency packet. Zrenner et al. (2020) showed that meaningful real-time phase estimation itself degrades when oscillatory amplitude and SNR are low. Gordon et al. (2021) showed that prefrontal theta-triggered stimulation needed additional constraints to avoid low-amplitude and phase-reset epochs. Vigué-Guix et al. (2022) showed that reliable trial-to-trial alpha phase locking still did not yield a consistent behavioral gain. Kim et al. (2023) showed across public datasets that prediction quality improves primarily with higher power and SNR and with waiting for eligible epochs. Hougland et al. (2025) then showed within-session fluctuation and low test-retest reliability of the optimal mu-phase. Therefore, the first public-safe artifact for `U8-1` cannot remain a delay-only dashboard. It also needs a phase-targeting log that keeps estimability, targeting accuracy, downstream effect, and phase-stability separate.