7 ms·

It is fascinating to me how quickly people reject the premise of Mighty, even after PG lists all the good reasons for replacing judgment with curiosity. I’ll a

It is fascinating to me how quickly people reject the premise of Mighty, even after PG lists all the good reasons for replacing judgment with curiosity. I’ll admit, my initial reaction to Mighty was “I can’t imagine it ever being faster than my behemoth PC.” But then I stopped for a second and got curious. What is Mighty, really? It is a thin-client. That’s it. And are thin-clients a bad thing? Well, if latency is an issue, sure. But “high-ping” is a somewhat solved issue, whether in multiplayer gaming, in terminal utilities like Mosh, or even in optimistic GraphQL mutation updates. Where’s the use-case where near-zero latency is vital? The only cases I can think of are games like Rocket League: fast-twitch games where even the latency between my controller and my PC is something I happily spend hours optimizing — where latency prevents the necessary feedback loop for learning (akin to trying to learn how to hit a baseball while drunk). But beyond near-zero latency use-cases, why would a thick client ever be better than a thin client? At the edge of performance, this question is easily answered: I would never attempt to train a PyTorch model on my admittedly powerful GPU. That’s what the cloud is for. So when it comes to my browser, why am I content to eat up memory and cpu-time with hundreds of tabs open that almost always include one or two that are broken, soaking up my resources, and have to be hunted down and killed off so that IntelliJ can return to its normal lightning-fast speed? Might goes even further and asks why I would want to run IntelliJ on my machine at all. Wouldn’t I rather run IntelliJ like I used to run Vim over Mosh, where I never have to worry about storage space, about download/upload bandwidth, or about my computer becoming sluggish? And that’s the killer idea here: that thin-clients almost always beat thick-clients. One could even argue that the entire internet is premised upon this reality. I’d happily pay Mighty to try it out for a bit. Even if it doesn’t work, I’ve dropped more money of less fascinating ideas. At the very least, I’m rooting for their success, because it would change a lot more than how you consume content over the internet.

gargs 5y ago
You just described a Netbook
Secretmapper 5y ago
What's mosh? Edit: Managed to find the right string of words to google: mosh = Mobile Shell
- detaro 5y ago
  https://mosh.org/ https://mosh.org/
sjg007 5y ago
This is the standard network computing paradigm. In the past you had terminals connected to a mainframe etc.. Same thing. X11/RDP and what not. The biggest gain here is security, you won't ever run code natively. The caveat is you also have to trust the host.
- jgalt212 5y ago
  > The biggest gain here is security, you won't ever run code natively. The caveat is you also have to trust the host. That assumes any malware cannot bust out out of mightyapp's sandbox to the host's host.
- cmiles74 5y ago
  The biggest gain is the security of all of the code running on someone else's hardware. The biggest loss is also security: all of your keystrokes, passwords, traffic, etc. are stored on someone else's hardware.
twobitshifter 5y ago
I think the key difference between a pytorch model and your browser is that you’re actively using and manipulating your browser and a pytorch model is a long running process without the need for second by second interaction. And as you note the need for more power goes beyond a browser, IDEs, gaming, rendering, Bitcoin mining. Why just do it for the browser? You can, and some people do, remote into a more powerful machine for all of their work. This was the norm when terminals were true terminals. We could go back to this, but having your own computer historically had much larger benefits for people.
dmitriid 5y ago
> why would a thick client ever be better than a thin client? Because thick client distribute the load among many computer. With a thin client all that load is way more centralised. > I would never attempt to train a PyTorch No one is asking you to do that. We're talking about Mighty, a thin client that: - runs a beefy VM in the cloud - runs a single app, Chrome, in that VM - streams video to the client If a million people run Chrome on their laptops and keep it open for the entire day (and your browser is usually open throughout the day), that's... just a million people with their laptops. If a million people run Chrome through Mighty, Mighty needs a million VMs always open, and a million video streams, also always open. See how a thick client is better than a thin client?
xfer 5y ago
The post is about crazy-new idea. This app is a RDP solution that is not a crazy or a new idea. It all depends on economic viability and marketing to the right people.
PragmaticPulp 5y ago
> It is fascinating to me how quickly people reject the premise of Mighty, even after PG lists all the good reasons for replacing judgment with curiosity. I was curious about Mighty and looking forward to their technology. That’s not the problem. The judgment came largely when they announced that it cost up to $600/year. It costs so much that it’s actually cheaper to buy a whole new computer if you might need it for a year or more. Once you put a price tag on something and ask people to pay for it, judgment is fair game. This whole dismissal of people saying that they couldn’t justify the product for the price as some sort of anti-curiosity thing feels disingenuous.
- chubot 5y ago
  If people will pay 2x or 10x the price of self-hosting for AWS, then I have no doubt that some people will pay 2x or 10x the price of a laptop for Mighty. (I wouldn't, but I also don't use AWS :) )
mbesto 5y ago
> It is fascinating to me how quickly people reject the premise of Mighty, even after PG lists all the good reasons for replacing judgment with curiosity. His arguments are too vague to specifically defend Mighty. You can insert any technology and his arguments are neither valid or invalid.
tsimionescu 5y ago
> Where’s the use-case where near-zero latency is vital? I would say typing is a pretty big one. It is extraordinarily unpleasant when typing lag is anywhere above maybe 50ms,and even worse when it is variable, like it would inevitably be if going over the Internet. It's even worse with mouse movements, where occasional spikes in lag can ruin your day.
- wildermuthn 5y ago
  This is a solved problem though, with optimistic updates. See Mosh, which does this for the terminal. If you've ever had to run Vim on a high-latency remote connection, it feels like having superpowers.
  - tsimionescu 5y ago
    Sure, but it's not solved in Chrome, is it? Mighty is just a VM running Chrome as far as I understand. Not to mention, the problem is fundamentally simpler in a terminal, with a very limited range of outputs. The problem of optimistically drawing the result of your input on the screen is much harder when that input could affect any portion of the screen in any way, like it can on a JS-powered web page.
nearbuy 5y ago
> I’ll admit, my initial reaction to Mighty was “I can’t imagine it ever being faster than my behemoth PC.” I haven't found Chrome slow on my mid-range PC (i5-7500). Or my phone (Pixel 3). I feel like I must be in some parallel universe to everyone else talking about how slow Chrome is. Just now, I loaded the first 6 links on Hacker News. 1 didn't load at all due to a server error. The other 5 all loaded in under a second (measured by DOMContentLoaded). I have uBlock Origin enabled (only in Firefox on the phone). Maybe that helps. I can have 100+ tabs open without slow down if I want to. The bigger problem is I'm less productive with 100 tabs open because... there are 100 tabs open. It's just too cluttered. I'm willing to admit Mighty might be a good product if people have this problem with slow browsers. I just never found this was an issue. Maybe it's a problem on low-end machines, but how many people have a low-end computer but are willing to pay $50/month for Mighty?
- chubot 5y ago
  Do you use AdBlock? That makes a huge difference. I just set up a new computer and I was wondering why Chrome and Firefox was so slow. Then I remembered I hadn't set up AdBlock yet. I wonder if Mighty has ad blocking? It's an interesting question of which will be better for their business -- blocking or no blocking. (I won't try it because of the obvious privacy problems, because I use a fast computer, and avoid slow web sites. But it probably has an audience.)
  - nearbuy 5y ago
    I use uBlock Origin for ad blocking. Maybe ad blockers are the secret.
finnthehuman 5y ago
>even after PG lists all the good reasons for replacing judgment with curiosity PG wants you to think that they're mutually exclusive. They're not, but he has a product to hustle. The problem (for PG) is that curiosity is not uncritical. Curiosity poses more questions than those in the category "how well does this product work?" Anyone actually curious is going to wonder about the problem space, not just one proposed solution. Pretty obvious questions include: "how did software get us to the point we're exploring this as a design?" and "could the problem it seeks to solve be addressed in a way that eliminates assumptions about the solution space?", "What are peripheral ramifications of design decisions, and how much do I care?" "Would other approaches solve the same issues, have the same ramifications?" or "Are they synergies to leverage by trying multiple things in concert?"

gargs 5y ago

You just described a Netbook

Secretmapper 5y ago

What's mosh? Edit: Managed to find the right string of words to google: mosh = Mobile Shell

detaro 5y ago
https://mosh.org/ https://mosh.org/

sjg007 5y ago

This is the standard network computing paradigm. In the past you had terminals connected to a mainframe etc.. Same thing. X11/RDP and what not. The biggest gain here is security, you won't ever run code natively. The caveat is you also have to trust the host.

jgalt212 5y ago
> The biggest gain here is security, you won't ever run code natively. The caveat is you also have to trust the host. That assumes any malware cannot bust out out of mightyapp's sandbox to the host's host.
cmiles74 5y ago
The biggest gain is the security of all of the code running on someone else's hardware. The biggest loss is also security: all of your keystrokes, passwords, traffic, etc. are stored on someone else's hardware.

twobitshifter 5y ago

I think the key difference between a pytorch model and your browser is that you’re actively using and manipulating your browser and a pytorch model is a long running process without the need for second by second interaction. And as you note the need for more power goes beyond a browser, IDEs, gaming, rendering, Bitcoin mining. Why just do it for the browser? You can, and some people do, remote into a more powerful machine for all of their work. This was the norm when terminals were true terminals. We could go back to this, but having your own computer historically had much larger benefits for people.

dmitriid 5y ago

> why would a thick client ever be better than a thin client? Because thick client distribute the load among many computer. With a thin client all that load is way more centralised. > I would never attempt to train a PyTorch No one is asking you to do that. We're talking about Mighty, a thin client that: - runs a beefy VM in the cloud - runs a single app, Chrome, in that VM - streams video to the client If a million people run Chrome on their laptops and keep it open for the entire day (and your browser is usually open throughout the day), that's... just a million people with their laptops. If a million people run Chrome through Mighty, Mighty needs a million VMs always open, and a million video streams, also always open. See how a thick client is better than a thin client?

xfer 5y ago

The post is about crazy-new idea. This app is a RDP solution that is not a crazy or a new idea. It all depends on economic viability and marketing to the right people.

PragmaticPulp 5y ago

> It is fascinating to me how quickly people reject the premise of Mighty, even after PG lists all the good reasons for replacing judgment with curiosity. I was curious about Mighty and looking forward to their technology. That’s not the problem. The judgment came largely when they announced that it cost up to $600/year. It costs so much that it’s actually cheaper to buy a whole new computer if you might need it for a year or more. Once you put a price tag on something and ask people to pay for it, judgment is fair game. This whole dismissal of people saying that they couldn’t justify the product for the price as some sort of anti-curiosity thing feels disingenuous.

chubot 5y ago
If people will pay 2x or 10x the price of self-hosting for AWS, then I have no doubt that some people will pay 2x or 10x the price of a laptop for Mighty. (I wouldn't, but I also don't use AWS :) )

mbesto 5y ago

> It is fascinating to me how quickly people reject the premise of Mighty, even after PG lists all the good reasons for replacing judgment with curiosity. His arguments are too vague to specifically defend Mighty. You can insert any technology and his arguments are neither valid or invalid.

tsimionescu 5y ago

> Where’s the use-case where near-zero latency is vital? I would say typing is a pretty big one. It is extraordinarily unpleasant when typing lag is anywhere above maybe 50ms,and even worse when it is variable, like it would inevitably be if going over the Internet. It's even worse with mouse movements, where occasional spikes in lag can ruin your day.

wildermuthn 5y ago
This is a solved problem though, with optimistic updates. See Mosh, which does this for the terminal. If you've ever had to run Vim on a high-latency remote connection, it feels like having superpowers.
- tsimionescu 5y ago
  Sure, but it's not solved in Chrome, is it? Mighty is just a VM running Chrome as far as I understand. Not to mention, the problem is fundamentally simpler in a terminal, with a very limited range of outputs. The problem of optimistically drawing the result of your input on the screen is much harder when that input could affect any portion of the screen in any way, like it can on a JS-powered web page.

nearbuy 5y ago

> I’ll admit, my initial reaction to Mighty was “I can’t imagine it ever being faster than my behemoth PC.” I haven't found Chrome slow on my mid-range PC (i5-7500). Or my phone (Pixel 3). I feel like I must be in some parallel universe to everyone else talking about how slow Chrome is. Just now, I loaded the first 6 links on Hacker News. 1 didn't load at all due to a server error. The other 5 all loaded in under a second (measured by DOMContentLoaded). I have uBlock Origin enabled (only in Firefox on the phone). Maybe that helps. I can have 100+ tabs open without slow down if I want to. The bigger problem is I'm less productive with 100 tabs open because... there are 100 tabs open. It's just too cluttered. I'm willing to admit Mighty might be a good product if people have this problem with slow browsers. I just never found this was an issue. Maybe it's a problem on low-end machines, but how many people have a low-end computer but are willing to pay $50/month for Mighty?

chubot 5y ago
Do you use AdBlock? That makes a huge difference. I just set up a new computer and I was wondering why Chrome and Firefox was so slow. Then I remembered I hadn't set up AdBlock yet. I wonder if Mighty has ad blocking? It's an interesting question of which will be better for their business -- blocking or no blocking. (I won't try it because of the obvious privacy problems, because I use a fast computer, and avoid slow web sites. But it probably has an audience.)
- nearbuy 5y ago
  I use uBlock Origin for ad blocking. Maybe ad blockers are the secret.

finnthehuman 5y ago

>even after PG lists all the good reasons for replacing judgment with curiosity PG wants you to think that they're mutually exclusive. They're not, but he has a product to hustle. The problem (for PG) is that curiosity is not uncritical. Curiosity poses more questions than those in the category "how well does this product work?" Anyone actually curious is going to wonder about the problem space, not just one proposed solution. Pretty obvious questions include: "how did software get us to the point we're exploring this as a design?" and "could the problem it seeks to solve be addressed in a way that eliminates assumptions about the solution space?", "What are peripheral ramifications of design decisions, and how much do I care?" "Would other approaches solve the same issues, have the same ramifications?" or "Are they synergies to leverage by trying multiple things in concert?"