Hacker Timesnew | past | comments | ask | show | jobs | submit | eterm's commentslogin

There's a soft failure-mode for bitcoin where due to the alternating difficulty adjustment, you could end up with people only mining every other 2016-block adjustment.

Let's call this cycle A and cycle B.

If A is too hard, miners drop out, cycle B gets easier, miners flood back, cycle A gets harder.

This results in the hard cycle getting longer and the easy cycle getting shorter.

This isn't completely critical as there is I believe a small damping effect, so it isn't completely lethal to bitcoin, but a key thing about bitcoin mining is that whether other people are mining or not doesn't actually affect your own profitiability.

Other people dropping out doesn't actually mean you get more bitcoins per hour/watt, it only affects the next difficulty adjustment as a secondary effect.


The damping effect is that part of your costs are the hardware, space, depreciation etc. leaving that stuff idle costs money - so it makes sense to mine in the less profitable periods too.

That depends on each miner's energy costs, so long as (variable cost of energy - revenue from coins) < fixed costs. It's still negative cashflow either way, but the monthly losses have to be weighed against the cost of going insolvent and losing the hardware.

Yes though AFAIK electricity is a large %

Crypto-miners are switching to AI token farming when bitcoin is low. They have compute that's both installed and powered, so why not do what pays better?

For bitcoin at least, you need totally different silicon.

I guess you could share the power supply and cooling infra, but I am dubious the savings are enough to have half your silicon idle all the time.


What the hell is AI token farming?

I think they mean serving inference workloads

How does that work? Isn't most bitcoin mining done on custom ASICs? I didn't think that the ASIC could be repurposed for inference.

Training ASICs (like Google’s TPUs) can generally run inference too, since inference is a subset of training computations. TPUs are widely used for both.

Mining ASICs (Bitcoin, etc.) cannot be repurposed…they’re hardwired for a single hash algorithm and lack matrix math needed for neural networks.


The biggest cost is the power which is often on multi year contracts. The hardware is comparatively cheap

That's wildly inaccurate. The cost in enormous both on the inference side and the mining side and has short lifetimes if you want SOTA.

The difficulty can only adjust by a factor of 4 which also limits the incentive change. You'd need more than 90% of miners to disappear to start seeing actual problems.

I think you're right, it's counterintuitive but less competition means less rewards to share for those who keep mining. Though transaction fees / hour shouldn't decrease, maybe your share of that is bigger.

I thought the rate of mining was tied to the maximum transaction rate the network can support?

It's the other way around, and there's no obligation to even carry transactions when mining, although it's incentivised through fees.

Your mining rate is simply your hash rate vs the hash difficulty.

Conceptually, it's analoglous to rolling random numbers in (0,1) until you get to a number smaller than 1/X, where X is large.

How long it takes you to do that, isn't dependent on how many other people are also trying to do that, if you get 1 hit per hour, then lots of other people getting hits doesn't actually stop you getting your 1 hit per hour.

Now, that's not quite the whole truth, as there's a small amount of time needed for propagation of the previous chain, but with an average hit globally of ~10 minutes, that's not actually a big factor.

What could happen to incentivise people is increased fees if blocks get less common due to dropped miners, there'd be more competition to get into blocks if they start filling up.

That combined with the fixed costs such as depreciation as othes mentioned, keeps the risk of this form of failure to a minimum.


It does seem ridiculous that over 20 years ago, gmail was advertised with a real-time allowance ticking away increasing, which started at an incredibly generous 1GB allowance and you could watch it tick up in real time faster than you could fill it with mail.

People designed "gmail-as-storage" apps to take advantage of this.

20 years later and we get a pathetic 15GB for mail, photos and everything else combined.


1GB that grew to 7GB over about 4 years and then 15GB over another 5 years. And has been stuck at 15GB for about 13 years. https://lifetourer.com/gmail-and-storage-capacity-cmon-googl...

The limit used to cost a whole dollar of hard drive space (plus redundancy), sometimes more than that. If they kept that up with adjustment for inflation then 100GB would be the free tier today, not a $20/year tier.


TBF that's a little bit apples-to-orchards, since publicly routed e-mails have certain expectable size/frequency characteristics compared to, say, all the videos someone possesses.

Ai detectors are bullshit.

That said, the second paragraph has the distinctive stocatto tone of AI

But AI is shaping how we write, so this could well all be hand written just by someone who spends time with AI output.


AI is also based on how we write. Some people are bound to write in a similar vein to LLMs naturally. See this person’s blog about it [0].

[0]: https://marcusolang.substack.com/p/im-kenyan-i-dont-write-li...


> stocatto

Staccato, which is Italian for "detached, separated".

When I see simple Italian words used as technical terms in music or art, I think "oh, this must be what English speakers feel when they work in tech - a lot of common words becoming specific concepts in that particular field".


Also, he is dangerously close to "stocazzo", which is similar to a very offensive way to say "fuck no!".

Thanks for the correction, I felt I hadn't got the word quite right.

Consider what it'd mean if there were parts of the Earth that could not be seen from the moon, it would also mean those locations could never themselves see the moon.

Ignoring the orbital period implications, I think it'd be bigger news if either US or Europe, or Asia couldn't ever actually see the moon.


This is the best ELI5 explanation I have heard. Thanks !

I listen to post-rock.

There are usually no lyrics, there's an absolute ton out there, and something about the music gets my brain flowing better than other instrumental music.


I adore this genre, and if you enjoy this submission I'd recommend the following games:

    NANDgame ( Free! https://nandgame.com/ )
    Silicon Zeroes ( https://store.steampowered.com/app/684270/Silicon_Zeroes/ )
    Turing Complete ( https://store.steampowered.com/app/1444480/Turing_Complete/ )
    Human Resource Machine / 7 Billion Humans ( https://store.steampowered.com/app/792100/7_Billion_Humans/ )
    MHRD ( https://store.steampowered.com/app/576030/MHRD/ )
They're all slightly different in terms how the construction of a computer is pitched, none of them are perfect, they all have quirks and flaws, but they're all fun.

Some like Human Resource Machine take the approachof

I wish Turing Complete wasn't quite so buggy or awkward, for a while it was by far the most promising of the bunch, but it's never quite polished and it's ended up in a bit of frustrating state.

Notable mention also to The Signal State, Shenzhen I/O, and TIS100 which are higher level than this, but scratch a similar itch.

there's ones like TIS100 which I keep meaning to revisit, but I find it very difficult to get back into these games without starting from scratch, and resetting my TIS100 progress is too intimidating.


2.7 is confusing because you can wire up the bitline and the word line the wrong way around and the tests still pass.

Your bug submission endpoint is getting a 429, so I'll report a bug here:

I see a difficulty pop up after I click "run tests" but it then gets hidden and doesn't do anything.

This was after selecting intermediate on the truth tables level, then clicking "next level" from there.


I refuse to believe any are as bad as the Azure Portal one.

It feels like pre-GPT levels of smart.


They no doubt predate .editorconfig, but the problem as described is now better solved by .editorconfig, which can be used to configure directory and file specific configuration and works cross-editor too.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: