I think it really was the M1 Chip that killed Mighty.
The product itself looked like it was working great.
Before the value proposition was good.
You can still use your old laptop and still have a very fast browser experience via remotely running it on a AMD server chip, this way you save on battery life and RAM too.
Then M1 speedometer scores started beating AMD Epyc on single core web browsing performance. Not only that, but even providing a much better experience in terms of battery life because of power savings.
Then yeah there was no point paying 20$-30$ a month for a remote browser instance. If you can afford that as a user you're either already set on CPU/GPU/RAM or better off just buying a used m1 air for <900$.
It's quite sad of a lesson where you do everything right, work consistently hard, be quite innovative, have plenty of financial and talent resources, and still fail.
> I think it really was the M1 Chip that killed Mighty.
Mighty is only dead because Suhail wanted to pursue what they clearly think is a bigger opportunity [0]. Mighty is already building its own GPU servers, might as well put them to use for AI as the perf and the size of AI models is roughly on an exponential trajectory [1]. It may not be long before AGI is within vicinity [2].
Though I agree a better strategy wouldn't be to accelerate the browser (akin to boiling an ocean?), but instead take one app at a time and make it browser-native (like, Photoshop -> Figma). A similar concept to (I don't remember who said it), take tools hackers use and build it for the Internet (grep -> Google, sendmail -> Hotmail, emacs -> Replit, sftp -> Dropbox etc).
[0] AIs today can automate away humans on quite a few tasks, assist them on quite a few complex ones already. And still: models are getting only way more capable, not less.
everything except for, as Suhail points out, picking the wrong side of a technical trend to bet on. All the technical wizardry in the world can't save you from bad strategy.
> I think it really was the M1 Chip that killed Mighty. The product itself looked like it was working great.
Agree with this take. It's amazing he was able to get this funded at a time it was fairly obvious from their product roadmap and language that Apple was going to release ARM Macs. Props to him for the skill it took to get funding in the door.
>It's amazing he was able to get this funded at a time it was fairly obvious from their product roadmap and language that Apple was going to release ARM Macs
It's easy to forget, but a lot of people were rubbishing the idea of ARM Macs as a viable replacement for Intel until it actually happened.
I did see that opinion expressed, but it was just silly given Ax performance in iPad/iPhone pre-M1. In no world does the flagship laptop use a more expensive, less efficient, and slower chip than the mid-range iPhone for more than a couple of product generations.
I don't think it was that silly to expect a transition from a stable, mature architecture, sold (albeit too expensive) by an established chip maker, to a new (to desktop) architecture designed in-house by a company known for its phones, especially when you consider that desktop workloads are very different from "here is an API for our walled garden" workloads.
Add to that the fact that a successful transition was entirely dependant on how well rosetta works (it turns out it works well, but that wasn't a given at all), evidenced by how people got burned in the powerpc-x86 transition
> when you consider that desktop workloads are very different from "here is an API for our walled garden" workloads.
This is a solid point, but iPad workloads were pretty clearly converging with laptop workflows. (And Apple was obviously moving in the direction of converging the SDKs even before the M1 was announced.) Plus, they already had the ability for devs to build apps to run on MacOS & iPad/iPhone. This was a strategic move that was visibly underway for years.
> a successful transition was entirely dependant on how well rosetta works (it turns out it works well, but that wasn't a given at all), evidenced by how people got burned in the powerpc-x86 transition
All the more reason to expect them to make a transition -- they had pulled it off successfully (if with some hiccups) before. I would not be surprised if they had some of the original Rosetta team around to revise for this time around.
In seriousness, going by what Suhail's been documenting through their tweets, they seem to fit the relentlessly resourceful bracket to a tee, which (according to pg, at least) means they have an above average chance at succeeding doing venture-scale startups: http://www.paulgraham.com/relres.html
I remember people being very impressed with the performance of the arm macs at the time. Maybe they were impressed that it met expectations though and correctly predicting it would be good.
I loved how fast Mighty was (even on M1 it was noticeable) but there's no way I could justify that expense and I'm a big early consumer of B2B SaaS tools.
For Mighty, this would mean their addressable market shrinks every year as more people replace with computers that are faster. That's not the characteristic of a market most startups want to be in.
Even for gaming and workstation use, this seems to be true. Last year, the value proposition of remote desktops looked great, but it turns out price/performance of all computing parts was at an all-time high.
I looked into this and found, annoyingly, that even if used the Remote Desktop rarely and I could deallocate the machine to not get charged, it was actually the price for storage that made it too expensive.
You could say that same about running off-line batch jobs on EC2 instances. Unless your co-located data sources are forcing you to do so, I cannot fathom why you would.
did this product have any appeal? the price was insane for something i get for free, if you could afford it and needed it you could afford a decent computer that could
It did when it came out. But it got already leapfrogged. Calculating fibonacci(45) with a recursive javascript function takes 13s on my m1pro and 7s on my ryzen 7600x. Profiling our angular app in chrome the initial render takes 200ms on m1pro and 130ms on my ryzen 7600x.
Apple silicon is still the master of performance per watt. But raw performance: It really was just tsmc 5nm what gave apple the edge. Apple will shine again when they sit exclusivly on tsmc 3nm, but until then, not so much.
I’ve talked to a couple of very loyal Mighty users. If you think about the number of hours a knowledge worker spends in the browser every month, it works out to a pretty small price per hour.
It’s similar to people in jobs that require them to live out of their inbox paying (a comparable amount) for Superhuman.
I'm really REALLY not sure how it makes sense to pay that much for a remote browser unless there's something I'm missing which I have to think is the case. What was the actual hook here? Surely a remote browser cannot have been worth it for 240-360 dollars a year? Wouldn't it have been more worth it to simply buy an M1 Mac or a Surface? How are people suffering from slow browsers that much?
There's a reasonably sized market segment of people who, themselves, work at startups with dubious value propositions. They try to prop up the whole ecosystem by spending a lot of money on each other. (Source: worked at a late-stage startup, spent weeks integrating APIs that did nothing but were owned by a friend of the CEO.)
There's another segment of people who can be sold snake oil. They aren't actually very sensitive to UI responsiveness, so they won't notice the extra latency added by the streaming service, but they will believe claims that it is "faster" than what they're currently using in some nebulous way. (These are the same kind of people who rewrite their app in React because they believe React makes apps faster.)
> Wouldn't it have been more worth it to simply buy an M1 Mac or a Surface?
Yes, probably, but I would note that a cheap M1 Mac is 4x the annual cost you cited for the browser, and if the browser is 95% of what you use a laptop for, getting the same perf boost for 25% the cost sounds like a good value prop.
Granted, a laptop should last more than a year, but presumably it starts to show it's age in under four, whereas the remote browser can improve over time.
I had Macbooks that lasted 5 years (2010-2015) pretty well, my current MBA M1 is almost 2 years old and I don't see myself needing to swap it for another 2-3 at least. My girlfriend still uses her MBA from 2014.
I don't really see how I could justify paying 20-25% of the price of my laptop per year, that I use for much more than just browsing, to have a faster browsing experience. And I'm a power user (SWE, etc.) compared to the mainstream consumer, the market for this product seems to be extremely niche/small.
Ironically BAAS is sold for security reasons too. As in isolation. You'd do very sensitive stuff with the remote browser, or vice versa, for isolation from other activity.
There was no point paying $30 a month _period_. You could get a cheap, refurbished 2014 laptop and it would still be a better value proposition than Mighty. The only potential buyers of this service are SV tech bros, not anyone living in the real life.
Before the value proposition was good. You can still use your old laptop and still have a very fast browser experience via remotely running it on a AMD server chip, this way you save on battery life and RAM too.
Then M1 speedometer scores started beating AMD Epyc on single core web browsing performance. Not only that, but even providing a much better experience in terms of battery life because of power savings.
Then yeah there was no point paying 20$-30$ a month for a remote browser instance. If you can afford that as a user you're either already set on CPU/GPU/RAM or better off just buying a used m1 air for <900$.
It's quite sad of a lesson where you do everything right, work consistently hard, be quite innovative, have plenty of financial and talent resources, and still fail.
Tech obsolescence is brutal.