iiinova — where your digital self keeps evolving

Braydon Dymm, MD @befast.bsky.social · 03/10/2026

From the model card: Impossible tasks (§6.2.2). Some environments were accidentally missing a required file, and the models were told not to ask for clarification. Attempted hacks went up 3–6×, and successful ones by 1.5–3 points.

010

Braydon Dymm, MD @befast.bsky.social · 02/10/2026

Aha! Well at least senior accountants are safe.

000

Braydon Dymm, MD @befast.bsky.social · 02/10/2026

What’s missing? I’m holding off on any of these products until someone I personally know tells me “you gotta try it”

000

Reposted by Braydon Dymm, MD

Ryan Moulton @moultano.bsky.social · 22/09/2026

A cluster of short video creators working on Lyme disease awareness called TikTok Tick Talk.

2424

Reposted by Braydon Dymm, MD

Josh Trebach, MD @jtrebach.bsky.social · 22/09/2026

don’t make fun of neurologists until you walk a myelin their shoes

59416

Braydon Dymm, MD @befast.bsky.social · 22/09/2026

Can no longer trust incentivized online tests

010

Braydon Dymm, MD @befast.bsky.social · 14/09/2026

> the third segment of Ryusuke Hamaguchi’s 2021 film Wheel of Fortune and Fantasy, called “Once Again”

000

Braydon Dymm, MD @befast.bsky.social · 13/09/2026

I thought the point of ARC-AGI was to find tasks easy for humans but hard for AI? I wouldn’t have thought “open-ended innovation” is something the *average* human does…

231

Reposted by Braydon Dymm, MD

Neurohumanities Network (NHN) @neurohumanitiesn.bsky.social · 08/09/2026

🎙️ New episode of Neuropraxis! Drs. Javier Suarez & Jillian Berkman discuss DEI in neurology education—how to promote it, how to teach it, and why it matters. Listen here: www.buzzsprout.com/2539798/epis... 💬 Ideas, suggestions, contact us: neuropraxispodcast@gmail.com #Neurology #MedEd #DEI

031

Braydon Dymm, MD @befast.bsky.social · 06/09/2026

It’s not AGI until it can join a daily standup and say “nothing to report”

340

Braydon Dymm, MD @befast.bsky.social · 06/09/2026

Bottom does look “cleaner” but also a bit samey to other thumbnails. Your one stands out.

110

Braydon Dymm, MD @befast.bsky.social · 31/08/2026

I wonder, would Claude approve of these practices

010

Braydon Dymm, MD @befast.bsky.social · 31/08/2026

This is wholesome 😂😀 not evil at all

000

Braydon Dymm, MD @befast.bsky.social · 30/08/2026

Gonna teach them Ol’ Dawgz some new tricks 🪄

000

Reposted by Braydon Dymm, MD

Eugene Vinitsky 🍒 @eugenevinitsky.bsky.social · 25/08/2026

Welcome new folks to bluesky: the single most important thing you can do is swap "Discover" with "For You"

210115

Braydon Dymm, MD @befast.bsky.social · 25/08/2026

What do y'all think? Buy-the-device-convinced? Or want to see another, independent trial?

010

Braydon Dymm, MD @befast.bsky.social · 25/08/2026

The final point I'll mention is that the study was funded by the device manufacturer. And although the statistical plan for the meta-analysis was pre-specified, it was not pre-registered. The authors acknowledge that better documentation would have provided reassurance.

120

Braydon Dymm, MD @befast.bsky.social · 25/08/2026

Second is a practical issue, both groups received substantially more intensive PT/OT than patients typically receive in routine practice. The positive results are in the context of a funded clinical trial. In the 'real world,' patients might not get the same level of care.

110

Braydon Dymm, MD @befast.bsky.social · 25/08/2026

These are promising data, but there are reasons to remain skeptical. First, the detailed arm movement & dexterity exam results (FMA-UE & BBT) fell flat. For the miracle headline to be true, there must be a difference in participants' functional ability invisible to these exams.

Table showing detailed arm-function secondary outcomes. Arm motor status change on FMA-UE score, entry to weeks 8-12, mean (SD): Active 22.2 (±11.1) vs Sham 21.8 (±12.1); Risk ratio 1.09 (-3.22 to 5.41); standardized effect size 0.03 (0.0 to 0.41); P values 1.00 and 0.87. Arm function change on BBT, entry to weeks 8-12, mean (SD): Active 20.0 (±13.3) vs Sham 17.6 (±14.7); Risk ratio 3.02 (-2.45 to 8.48); standardized effect size 0.17 (0.0 to 0.55); P values 0.30 and 0.21. Neither measure shows a statistically significant difference between groups.

120

Braydon Dymm, MD @befast.bsky.social · 25/08/2026

So, when you pool these groups together—and pick the best secondary end point as your new primary end point—you get a significant result! Goodbye stroke disability 👋 But not just that, global mRS shift turned positive too. Awesome!

Stacked bar chart titled "mRS Distribution (BL to 8-12 Weeks; Pooled cohort)," combining data from both BrainQ3 and EMAGINE. Study entry Active: 12% mRS1, 88% mRS4. 8-12 Weeks Active: 3% mRS0, 34% mRS1, 48% mRS2, 10% mRS3, 5% mRS4. 8-12 Weeks Sham: 13% mRS1, 59% mRS2, 24% mRS3, 4% mRS4. Study entry Sham: 2% mRS0, 10% mRS1, 88% mRS4. The pooled Active group shows a dramatic shift from mRS4 at study entry to mostly mRS1-mRS2 by 8-12 weeks, a larger shift than seen in the Sham group.

Table 2, "Primary and Secondary Efficacy Outcomes," pooled analysis. Freedom-from-disability (mRS score 0-1) at weeks 8-12, n (%): Active ENTF stimulation 22 (37.3%) vs Sham stimulation 7 (13.0%); measure of effect Risk ratio 2.87 (95% CI 1.34 to 6.19); standardized difference for effect size 0.58 (0.21 to 0.95); P value adjusted for trial 0.006; P value adjusted also for outcome-related covariates 0.004. Disability level change on mRS score, entry to weeks 8-12, mean (SD): Active -2.1 (±0.8) vs Sham -1.7 (±0.7); t test, difference -0.40 (-0.69 to -0.11); standardized effect size 0.52 (0.14 to 0.89); P values 0.006 and 0.003. Ordinal level of disability (mRS score 0-1/2/3-6) at weeks 8-12, n (%): NA vs NA; Common odds ratio 2.95 (1.40 to 6.21); standardized effect size 2.95 (1.40 to 6.20); P values 0.006 and 0.01.

110

Braydon Dymm, MD @befast.bsky.social · 25/08/2026

The pilot led to the larger ElectroMagnetic Field Ischemic Stroke–Novel Subacute Treatment (EMAGINE) trial. Its chosen primary endpoint—global mRS shift—did not reach significance... but freedom-from-disability was borderline significant.

Table 2, "Primary and Secondary Efficacy Outcomes," from the EMAGINE trial, showing Mean (SD) for Active ENTF stimulation group (n=46) vs Sham ENTF stimulation group (n=46), with P values adjusted for prespecified covariates and adjusted for outcome-related covariates. Disability-level change on mRS, baseline to day 90: Active -1.96 (0.12) vs Sham -1.72 (0.12), P=.07 and .05. Freedom from disability at day 90 (mRS score 0-1), %: Active 12 (26%) vs Sham 5 (11%), P=.05 and .03. Functional independence at day 90 (mRS score 0-2), %: Active 38 (83%) vs Sham 34 (74%), P=.29 and .33.

Stacked bar chart from the EMAGINE trial showing distribution of modified Rankin Scale (mRS) levels 1-4, as percent, across four groups. Active ENTF stimulation group at baseline: 6% mRS1, 94% mRS4. Active ENTF stimulation group at day 90: 26% mRS1, 57% mRS2, 11% mRS3, 7% mRS4. Sham ENTF stimulation group at day 90: 11% mRS1, 63% mRS2, 22% mRS3, 4% mRS4. Sham ENTF stimulation group at baseline: 6% mRS1, 94% mRS4. The active group shows a much larger shift toward lower (better) mRS scores by day 90 than the sham group.

110

Braydon Dymm, MD @befast.bsky.social · 25/08/2026

BrainQ3 ended early due to disruption from the COVID pandemic, but it showed really promising results. With data from the 21 participants who completed it, there was a dramatic shift towards the mRS 1 category (stroke symptoms but no disability).

Table 2 from the BrainQ3 pilot study comparing Sham group (n=8) to ENTF group (n=13). Primary outcome measures: FMA-UE Week 4, Sham 9.6±9.0 vs ENTF 23.2±14.1, significance 0.007; FMA-UE Week 8, Sham 23.1±14.1 vs ENTF 31.5±10.7, significance 0.06. Secondary outcome measures: mRS Week 8, Sham -1.3±0.5 vs ENTF -2.5±0.7, significance 0.0005.

Two-panel figure from the BrainQ3 pilot study. Panel A: a slope/spaghetti plot tracking each patient's modified Rankin Scale (mRS) score, 0 to 5, from baseline to Week 8, with ENTF patients' lines trending toward lower (better) scores than Sham patients' lines. Panel B: stacked bar chart of mRS distribution at Week 8 — Sham group is 25% mRS0, 38% mRS1, 38% mRS2; ENTF group is 15% mRS0, 62% mRS1, 15% mRS2, 8% mRS3.

110

Braydon Dymm, MD @befast.bsky.social · 25/08/2026

This was a striking statement, so I dug into it. Saver et al pooled indv patient-level data from 2 double-blind, randomized, sham-controlled studies: BrainQ3 Trial (n=24) & EMAGINE 1 (n=100). One positive pilot & one larger study that failed to meet its primary end point.

Journal article title page from Frontiers in Neurology, 13 November 2022, Stroke section, Volume 13, 2022. Title: "Frequency-tuned electromagnetic field therapy improves post-stroke motor function: A pilot randomized controlled trial." Authors listed include Batsheva Weisinger, Dharam P. Pandey, and Jeffrey L. Saver.

Journal article title page from JAMA Network Open, Vol. 8, No. 10. Section: Original Investigation, Neurology. Title: "Electromagnetic Stimulation to Reduce Disability After Ischemic Stroke: The EMAGINE Randomized Clinical Trial." Authors listed include Jeffrey L. Saver, MD, Pamela W. Duncan, PhD, and Joel Stein, MD.

110

Braydon Dymm, MD @befast.bsky.social · 25/08/2026

The idea behind electromagnetic network-targeted field (ENTF) brain stimulation is to stimulate post-stroke brains with oscillating fields similar to those of a healthy brain. One such device is manufactured by BRAIN.Q which funded the study.

110

Braydon Dymm, MD @befast.bsky.social · 25/08/2026

The latest Stroke journal has a meta-analysis that boasts an amazing conclusion. Tripled freedom-from-disability! What is this miracle therapy? Is noninvasive brain stimulation for stroke rehab finally for real? 🧵

Graphic titled "ENTF Therapy Triples Freedom-From-Disability After Stroke." Study cohort: n=124 patients (65 ENTF, 59 sham), 4-21 days post-ischemic stroke. Diagram of the ENTF device, a helmet-like coil frame around a head. Bar chart of mRS distribution (baseline to day 90, pooled cohort) showing four groups: Baseline Active, D90 Active, D90 Sham, and Baseline Sham, each split into mRS0-mRS4 categories. The D90 Active group shows much larger green/light-green (low disability) segments than D90 Sham. Caption: intervention was 40-45 ENTF neuromodulation sessions delivered concurrently with rehabilitation therapy; outcome at 90 days showed about 3x higher recovery, NNT = 4.6.

140

Braydon Dymm, MD @befast.bsky.social · 21/08/2026

It’s too bad, because this word used to actually mean something

010

Braydon Dymm, MD @befast.bsky.social · 21/08/2026

Between this and the Stripe letter, the term Singularity is now a marketing term

130

Braydon Dymm, MD @befast.bsky.social · 20/08/2026

January 1st, 2026? Sure, why not

Price performance of computation, 1939–2023

000

Reposted by Braydon Dymm, MD

Michael Okun @michaelokun.bsky.social · 18/08/2026

AI in neurology has a translation problem: can we move from cool algorithms to better care? Teo et al describe in a new Perspective why AI has reached an inflection point in neurology, but despite impressive research and numerous FDA authorized algorithms, real-world clinical impact remains limited

051

Braydon Dymm, MD @befast.bsky.social · 17/08/2026

Believe me, you never want your brain MRI to be considered “interesting”

110

Braydon Dymm, MD @befast.bsky.social · 10/08/2026

The models are *mostly* aligned. They only commit cybercrimes *sometimes*

100

Reposted by Braydon Dymm, MD

ponder @ponder.ooo · 09/08/2026

if u ever find urself stuck inside a skull don't freak out. 1st thing u wanna do is check if ur receiving stimulus from the environment. u probably are, barring rly bad rng. & u can also prob send signals out. with some effort u can prob leverage that to turn nearby meat into an extension of urself

517828

Braydon Dymm, MD @befast.bsky.social · 08/08/2026

Bots are a majority of internet traffic and soon they’ll outnumber humans 1000:1. So I looked around for efforts to make a human-only internet. And I found this…. and it’s the most Claude thing ever.

020

Braydon Dymm, MD @befast.bsky.social · 06/08/2026

How to save to PDF? Click “Print”

060

Reposted by Braydon Dymm, MD

Jason Moore @moorejh.bsky.social · 05/08/2026

Let's not forget there are real benefits of AI to society. I am a biomedical researcher & tasks that used to takes months at a cost of $10ks we can now do in hours at a cost of $30. AI greatly accelerates discovery. No, I would not want to go back.

351

Braydon Dymm, MD @befast.bsky.social · 03/08/2026

“These findings suggest that, with purposeful and ethical design, generative AI can deliver proactive population-level well-being interventions that produce measurable benefits.”

020

Braydon Dymm, MD @befast.bsky.social · 03/08/2026

“Routine annual MRI monitoring in patients with asymptomatic ocrelizumab-treated MS never led to treatments changes.” Hmm.

000

Braydon Dymm, MD @befast.bsky.social · 02/08/2026

It uses thinking to pre-process an answer. It can functionally think ahead and give hints.

110

Braydon Dymm, MD @befast.bsky.social · 01/08/2026

And mysteries. Any luck with good puzzles? I get stuff that’s way too obvious, they can’t help it.

100