chongman99

What is the rationale for "isn't easily repurposed"?

The hardware can train LLM but also be used for vision, digital twin, signal detection, autonomous agents, etc.

Military uses seem important too.

Can the large GPU based data centers not be repurposed to that?

I can't say anything about this mask that has a ton of cloth, but we had people wearing the http://bunnypapr.org in New York City summer. They did fine.

There is a fan blowing 30 to 60 liters per minute through the bag/volume. This is like a cool breeze if you aim it at the skin.

In the winter, you have to aim it at the shell and not the skin because you'd freeze otherwise. But, mostly these are worn indoors.

Thanks for the link. I run viralhelmets.com and a few people came and visited the site based on google analytics. Looking at the log files, I found this backlink.

We tried getting visibility with bunnyPAPR.org and now with the $15 Hoodie Viral Helmet (slimmer).

The mass market appeal of these is quite low. People are tired.

Hospital market was a no go (they need certifications).

Airports... maybe. But we found it tough trying to break in to that market. Airlines aren't really supportive.

I track this stuff here: https://www.reddit.com/r/viralHelmets/

Take a look at my instructable on filters and fans. https://www.instructables.com/Viral-Helmets-9999-Hospital-Gr...

I also posted in reddit about it. https://www.reddit.com/r/cosplayprops/comments/l0ucd2/help_h...

On the viralhelmets I make, the fans are about the same loudness of laptop fans. You do have to talk louder, but you can still hear.

This is not technically true (but often true).

Water vapor is a gas and is much smaller than the "leaks" of particles. A perfectly sealed filter (like an elastomeric mask with high fit factor) will still release a ton of H20.

EVIDENCE: The exhaust from PAPRS and viralhelmets.com has a ton of moisture.

We have put out an open source* design of a $15 to make a "Hoodie Viral Helmet". It's basically a PAPR built from hospital grade filters. If you are interested, go take a look: https://www.viralhelmets.com/15-hoodie-viral-helmet

I run the non-commercial site http://viralhelmets.com.

DETAILS / WHY I'M POSTING

I saw a lot of people pwning the $199 and $300 viral helmets that have come out. (see https://news.ycombinator.com/item?id=25848412) Yeah, at those price points, it doesn't make sense.

The $15 design actually comes out of about 1000+ hours of engineering work (see bunnyPAPR.org). We tried to get interest to commercialize and get certified (NIOSH/FDA). But medical certification in the US is a nightmare and hospitals are slow. We couldn't get much interest or backers.

So, I'm releasing the key parts of our research as a $15 free-to-build plan.

THE FILTERS MATTER One big difference is that we use hospital-sourced PFT and HMEF filters. We don't trust the supply chains that are pumping out special purpose filters for "fad" mask designs. After the fake/underperforming KN95 scandals in April/May, we stuck with supply chains that existed before COVID. The Stanford pneumask and maskson.org projects took a similar approach.

UPTAKE/ MARKET FIT We tried approaching some nursing home, school districts, and hospitals about this. They are so inundated by snake oil vendors, they weren't very interested in being a beta tester. Same with CDC... inundated. <shrug>

We do have a few doctors using these, but not publicizing it.

CODA Viral Helmets could protect all wearers. Could end the pandemic. But, that's a hard sell since they are so "untested". (Though we use PFT and HMEF filters in surgery, and PAPRs in biohazard scenarios... <shrug>)

*Not technically "open source". CC-NC-SA license.

Thanks to @yonran for mentioning ViralHelmets.com in the other yc post.

That was that thinking in April, but more careful studies are showing that wearer protection is quite high. And that bandanas and neck gaiters do actually offer reasonable protection.

Here's the preprint from Linsey Marr who is considered by many a leader in aerosol risk.

https://www.medrxiv.org/content/10.1101/2020.11.18.20233353v...