[joosters]

All proof-of-work algorithms can be accelerated by specific hardware of some kind or another. Bitcoin is perhaps the extreme example, with ASIC miners. But coins using other algorithms are mined by using top-end graphics cards, which is still a form of specialized hardware. There's no algorithm that can't improve upon a standard computer.

Other forms of proof-of-work have been mooted, e.g. algorithms that require lots of memory, or lots of storage, but again, specialized equipment will accelerate this too, it's just that the coins aren't worth enough for people to invest in designing the hardware so far.

[saurik]

A top-end graphics card has other uses and other value, and is something a large number of people (myself included) already own; and most importantly: they can always be used to mine a different coin, which entirely removes them from contention as "specialized" as used in this article (as the entire argument was based around how the hardware would not be useful for a different coin).

As for "all things can be accelerated", the question here is more "how unique is the thing being accelerated?" (and so I would agree the wording in my comment was slightly off), as this argument about the grim trigger holds up if and only if the specialized hardware being used is not valuable for other uses (including other coins).

As an example of what you can conclude with this thought process: if you are a small "startup coin" you should seriously considering a proof-of-work scheme that is weirdly unique in that it forces people to not be able to reuse hardware they have from another (particularly if more valuable) coin.

[joosters]

True enough, the graphics cards certainly have alternate uses while a bitcoin miner becomes a very heavy paperweight once the difficulty increases and it turns unprofitable to run.

Your thought process leads to another interesting game theory puzzle: There's a benefit in creating a coin that has a unique proof of work, but as you say, there's also value in letting miners easily switch their existing rigs to mine your coin. I wonder which is actually the best choice?

[Klathmon]

SIA is making the bet that somewhat ASIC friendly algorithms are better off.

The argument being that because ASICs are worthless for anything but mining that one coin, you can really accurately gauge and maintain a hashrate across the whole network. It would be crazy expensive to hoard a bunch of ASICs for a specific algorithm but not run them in hopes to launch a secret 51% attack.

With GPUs that's not really an option. Because GPUs can be used for so much more, it would be easy to see where you could suddenly have a flood of GPU power hit your chain which could be massively over 51% quite easily.

And they could be gone again just as quickly, leaving the remaining users with an impossibly high difficulty and killing off the chain (without a hard-fork to fix it).

It's a fascinating read which actually flipped my opinions on ASICs in favor of them now, I highly recommend it [0].

[0] https://blog.sia.tech/choosing-asics-for-sia-b318505b5b51

[maxerickson]

As an example of what you can conclude with this thought process: if you are a small "startup coin" you should seriously considering a proof-of-work scheme that is weirdly unique in that it forces people to not be able to reuse hardware they have from another (particularly if more valuable) coin.

I imagine there is much danger here of thinking oneself smarter than any attackers.