May 15th had come and gone. The date I’d circled in my calendar—the supposed release date of the Ultrahuman Ring PRO—dissolved into just another Friday. And I watched this beacon of promise vanish without delivering the one thing I actually needed. A verdict. A scatter plot. Instead, the date slipped into the rearview mirror, leaving behind nothing but a vacuum of information. No test from The Quantified Scientist. No data to confirm whether my hope was grounded in science or merely a symptom of desperation.
Obsessively refreshing The Quantified Scientist’s channel became a ritual of hope and frustration: click, scroll, sigh, repeat. Waiting for a single upload. Now, I was also scouring Google, hunting for any scrap of independent validation. Anything about the Ultrahuman Ring PRO and sleep. Anywhere people might be talking about it. But the silence was deafening—or rather, not silence at all. It was noise. Let me spare you my growing frustration with the so-called reviews—sorry, “honest reviews”—that don’t even last a full minute of video. Seriously! Is that the new attention span in 2026 1?
Then, on June 15th—during one of these obsessive refresh cycles, this time through Google instead of YouTube—something I hadn’t asked for found me: the RingConn Gen 3.
Refined Performance, Everyday Confidence
The Gen 2 had been discarded—poor sleep tracking, same fate as the Ultrahuman Ring AIR. But if I were willing to consider the Ultrahuman Ring PRO, shouldn’t I apply the same logic to the Gen 3? Should I give the Gen 2’s successor the benefit of the doubt? The website screamed “Improved Data Stability and Accuracy.” Fair enough. Except when I dug for sleep specifics? Nothing concrete, except for touting the Sleep Apnea feature and one line worth a raised eyebrow: “>99% Accuracy Sleep Time.” Compare that to the 22.5% agreement rate for “Wake” detection The Quantified Scientist 2 recorded on the Gen 2. Either they’d cracked sleep overnight, or I was being sold magic.
Detect silent breathing interruptions and oxygen drops you might never notice, addressing potential sleep apnea risks before they impact your health.RingConn
The Quantified Scientist isn’t using wakeful-state information anymore. In the original video I shared in this series, he showed a sensitivity-versus-specificity scatter plot—or, in plainer terms, a sleep vs. awake graph. X-axis (horizontal): Sensitivity, the ability to detect sleep. Y-axis: Specificity, the ability to detect wakefulness. Later, he introduced sleep stages—light, deep, REM—with awake as the fourth, by the way. He compared them all against the Zmax 3 as his reference. But then he dropped “awake” from his four-stage analysis.
As always, we’ll ignore […] awake time […], just because the Zmax tends to detect a lot of awake moments that aren’t really awake moments. This has to do with something in the scoring algorithm.The Quantified Scientist
Yet a 22.5% agreement rate for “Wake” detection is something I can’t “ignore,” even if instructed to. So I dug deeper and found a video from LordofTheSmartRings: Best Smart Ring: SLEEP (external link). In it, he tested the RingConn Gen 2 alongside other rings—including the Ultrahuman Ring AIR and Oura 4—against polysomnography (PSG). One night only, but still.
With his own hands (literally), and crucially on the 2-stage-only analysis (sleep or awake), the RingConn Gen 2 detected only about 50 minutes of his 100 minutes of awake time—as measured with PSG. That’s bad, but not as bad as the AIR’s pathetic ~20 minutes. Both failing rings overestimated sleep accordingly, by the way. And yes, the Oura 4 hit nearly perfect accuracy. I know. Of course.
Interestingly, that night he had to go to the bathroom—blaming an extra beer, apparently. And then he struggled to fall back asleep. The PSG labeled this time as awake. The Oura 4 missed the onset but aligned quite well with the PSG data overall. Let me stress this again: he got up, walked around. These devices have accelerometers, don’t they? The AIR picked up the movement, sure. But the RingConn Gen 2? Crickets.
As for 4-stage agreement with PSG in LordofTheSmartRings’s test, the RingConn Gen 2 posted a dismal 44.3% for Awake. This poor performance can’t be pinned on the Zmax—it was real PSG in a sleep lab. For reference: AIR at 43.0%, Oura 4 at 70.0%. Again.
In French, we’d conclude: CQFD 4.
Nothing else about sleep, because everything was about the new feature: Vascular Health. And sleep apnea detection wasn’t new—Gen 2 already had it. Actually, this felt like a cruel joke. If the Gen 2 couldn’t even track sleep properly, how could it possibly detect sleep apnea? That’s the “deuxième effet Kiss Cool” at work. Their FAQ section spelled it out clearly:
Which RingConn model should I choose — Gen 3, Gen 2, or Gen 2 Air?
Gen 3 — Best for advanced health insights, including vascular health trends and proactive alerts.
Gen 2 — Ideal if sleep and snoring risk monitoring are your main focus.
Gen 2 Air — A budget-friendly choice for essential daily health tracking.
How can they praise the Gen 2 for sleep when it is actually one of the worst-performing devices out there
?
Of course, without cold, hard data from a validation study, I was reduced to mere speculation. Like one of those regulars at a French village bar, rebuilding the world over a succession of cheap red wines, I sifted through the fragments available and tried to construct a coherent theory. The Gen 2 had been terrible. The Gen 3 promised improvement. The company praised both devices. Therefore, perhaps the Gen 3 was better. Or perhaps it was simply newer. Or perhaps “better” meant better according to a marketing department that had never been forced to compare its sleep stages against an EEG.
As my patience eroded, I began to forage for any available information on the Gen 3. Since it had been in the wild slightly longer than the PRO, I hoped to find a few actual reviews. About sleep, of course! I didn’t care about anything else. The experience was disheartening. It’s a strange modern tragedy: how empty “reviews” have become—polished, superficial echoes of the manufacturer’s press release, devoid of critical thought. But with a nice affiliate link!
But then, I found a thread in a Facebook group (external link)—a place I haven’t ventured into since 2007—a raw, unfiltered scream into the void that shattered my hope of using the Gen 3 as a backup:
“Sleep tracking has been off anywhere from 45 minutes to 1.5 hours (total sleep, REM, light sleep, and deep sleep).”
“It failed to detect two of my naps, which honestly annoyed me.”
This is a single anecdote, not a validated study. And yet—for someone who doesn’t care about “readiness scores” or “wellness tips,” but cares deeply about the binary truth: was I asleep, or was I awake?—this was a devastating blow. A margin of error so vast it might as well have been random guessing. And naps? Undetected? That wasn’t just poor tracking. That was a complete failure. Did RingConn even improve anything sleep-related compared to the Gen 2 (see box 1)?
So there it was: a would-be backup plan—in case the PRO turned out to be vaporware or a disappointment—collapsing under its own anecdotal weight before it even had the decency to be properly tested.
To be continued…
1 Actually, it is. According to research from Dr. Gloria Mark at the University of California, smartphones have eroded our attention spans down to a staggering 47 seconds. Forty-seven! Back in my student days, instructors kept harping on the 90-minute rule—that was the ceiling before adults needed a physical or mental reset. So did our brains evolve, or devolve? ^
2 In RingConn Gen 2 Smart Ring – Initial Scientific Review (external link). ^
3 The ZMax headband is a consumer-grade EEG recorder for home sleep monitoring. It captures frontal electroencephalography (EEG) and actigraphy, but skips submental electromyography (EMG). Easy to slap on and comfortable enough to wear, it delivers signal quality good enough for reliable—though far from perfect—sleep metrics. REM detection was its primary design goal, and while the ZMax holds up against gold-standard PSG, it doesn’t meet the full American Academy of Sleep Medicine electrode requirements. So yes, it’s legit, but don’t treat it like a hospital bed. ^
4 CQFD—”ce qu’il fallait démontrer,” or “that which was to be demonstrated”—the French cousin of Latin’s QED (Quod erat demonstrandum). Math nerds tack it onto the end of proofs to seal the deal. In everyday French, you slap it down the same way English speakers mutter “case closed” or “thus proving my point.” Consider the matter settled. ^