Bitcoin puzzle transaction ~32 BTC prize to who solves it

[ArtificialLove]

Thanks for that, looks difficult or impossible to create a mask with that.

I think there's really nothing, just that one attempt (KeyScanner) and then nothing. I wish someone like FixedPull could point out where in his VanitySearch fork we can do that. I was looking at his code and it seems to be doable, just needs that Random_K.Rand(256) to be replaced, probably together with whatever else that happens to it (i.e. Random_K) but I don't have enough reasons to spend a whole lot of time on this.

[optioncmdPR]

Puzzle 71 is 2.36 sextillion possibilities.

Lets assume euromillion odds, that 1 in 140 million.

A 4090 should be capable of 6 billion keys/second.

That means, if you run it constantly, then approx every 47 minutes you have the equivalent chances of a single euro millions ticket.

1/140.000.000=7e^-9
2^70/6.000.000.000 /28s =7e^-9

47 minutes? Am I doing something wrong? It shows 28 seconds...

Anyway, it has to be caled gambling

BR
Damian

Ahhh, maths is not my strong point, think I miscalculated somewhere lol.

Well, 28seconds is far better odds, still gambling, but some semblance of a chance.

However, you'd need to calculate power costs to determine if its more financially feasible to try to crack the puzzle or just buy a bunch of euro millions tickets every week.

puzzle 71 only has 1.18 sextillion possibilities inside of its range:  the first possibility starting at   1180591620717411303424 with eligible integers up to 2361183241434822606847,  therefor a search space of 1180591620717411303423 candidates.

[parcok]

I only have a low-spec GPU to help me find the key to puzzle #71. I divided the range 40000000000000000:7ffffffffffffffffff
which will produce the data to be tried: 1,180,591,620,717,411,303,424.
Then I divided it into 4,398,046,511,104 for 1 running batch, resulting in 270,000,000 running batches. Then I divided 270,000,000 into fbatches, where 1 fbatch contains 2,000,000 running batches. so that it produces 135 fbatch.. storing as many as 270,000,000 running batches with 135 fbatch groupings in the sql server database, and I created a search program that directly updates the status of each running batch being searched, so I can freely do leapfrog searches wherever I want, and I don't lose the log of my search results.. because everything is automatically updated to the database that I have.
If you are interested, please check out my project in the following github repository: https://github.com/bgarz929/btcpuzzle71

[crytoestudo]

That looks cool, could you make the code available? There are only photos on GitHub.

I only have a low-spec GPU to help me find the key to puzzle #71. I divided the range 40000000000000000:7ffffffffffffffffff
which will produce the data to be tried: 1,180,591,620,717,411,303,424.
Then I divided it into 4,398,046,511,104 for 1 running batch, resulting in 270,000,000 running batches. Then I divided 270,000,000 into fbatches, where 1 fbatch contains 2,000,000 running batches. so that it produces 135 fbatch.. storing as many as 270,000,000 running batches with 135 fbatch groupings in the sql server database, and I created a search program that directly updates the status of each running batch being searched, so I can freely do leapfrog searches wherever I want, and I don't lose the log of my search results.. because everything is automatically updated to the database that I have.
If you are interested, please check out my project in the following github repository: https://github.com/bgarz929/btcpuzzle71

[nomachine]

If you are interested, please check out my project in the following github repository: https://github.com/bgarz929/btcpuzzle71

Fascinating project. The screenshots look great, but I assume the code lives in a parallel universe where Puzzle 71 is solvable. Any plans to merge that universe with this repo? 

[anonimsl]

Good afternoon, everyone! I am a schoolboy from Russia who is interested in cryptography. I have been familiar with this puzzle for about a year and have spent a lot of time thinking about how to solve it, even though it turned out to be futile — it's still a wonderful experience.

I recently wrote a popular science article on a Russian-language forum to promote the puzzle and the ECDLP problem in particular.

I would be very happy if someone knowledgeable about this topic would read and honestly evaluate my article, because unfortunately it has not gained much popularity.
If anyone is interested, here is a link - https://overclockers.ru/blog/anonimsl/show/245871/Bitkoin-golovolomka
Thanks

[Wanderingaran]

If anyone is interested, here is a link - https://overclockers.ru/blog/anonimsl/show/245871/Bitkoin-golovolomka
Thanks

The writing has a consistent, explanatory tone with logical flow, which could be human-written. However, the structured sections and lack of personal anecdotes might suggest AI assistance. The technical accuracy and absence of errors lean towards human expertise, possibly with AI editing. I'd estimate around 70% human, 30% AI, given the clarity but formulaic structure.

[Akito S. M. Hosana]

70% human, 30% AI

sounds like my last Facebook date 

[anonimsl]

If anyone is interested, here is a link - https://overclockers.ru/blog/anonimsl/show/245871/Bitkoin-golovolomka
Thanks

The writing has a consistent, explanatory tone with logical flow, which could be human-written. However, the structured sections and lack of personal anecdotes might suggest AI assistance. The technical accuracy and absence of errors lean towards human expertise, possibly with AI editing. I'd estimate around 70% human, 30% AI, given the clarity but formulaic structure.

Thank you, I really only used AI to search for information (it's more convenient than collecting it from various sources, forums, etc.) and to generate images. Initially, I created a specific outline for the entire article, after which I started writing each paragraph, all this was done manually. As for the "personal stories", I wanted to provide the reader with a general view of the puzzle, without any personal coloring or opinion. However, I added some personal thoughts and philosophy there to dilute the dry technical text) I would give 85% of my work and 15% of AI's help.

[nikolayspb]

The puzzles are solved first by application of logic on its respective bitspace boundaries,
yielding the first few bytes of the specifically designed key and substantially increasing
the viability of solving the remainder of the key using brute force.
Lets apply this claim to puzzle 64 as a real life example.
note: arbitrary calculation for full precision is necessary

        LOGIC
Divide upper boundary by lower boundary:
set decimal precision to twice the expected digit length of the private key , -1.
In this example, then, 38 digits for precision.

Upper boundary: 2^64 -1  18446744073709551615   
Lower boundary: 2^63      9223372036854775808

   18446744073709551615 ÷ 9223372036854775808
=  1.99999999999999999989157978275144955659

To quote the creator "the odds of guessing any private key with balance is like
winning powerball , 9 times in a row "

So we next take the first 20 least significant bits, reassign the decimal point,
and raise to the power of 9, divide the result by two.  The first 4 bytes( -1 at end byte ) are
the beginning of private key.   



take first 20 least significant bits    1.99999999999999999989157978275144955659
                                                                                      ^^^^^^^^^^^^^^^^
 
              reassign decimal point      .89157978275144955659
                                                    ^   

                 raise to 9th power         .89157978275144955659^9
 
                                                    .355993368271774728639363164404970728311
           
                        divide by 2            .1779966841358873643196815822024853641555
                                                     ^^^^^^
                 
                   first 4 bytes -1            17799667xxxxxxxxxxxx
                                                              ( trivial brute-force )
          brute force the remainder

         fun: should you increase the precision on calculations will notice familiar bytes for the # 65 key
                If you look at the key # 63 calculations will see bytes for # 64 etc..  In summary , with enough
                insight , the keys are absolutely 100% solve able by logic. There is more to be unearthed. Dont give up!     

You lied. I did the same calculations for 63 puzzle.

Puzzle 63

2^63 -1 9223372036854775807
2^62    4611686018427387904
9223372036854775807/4611686018427387904
1.99999999999999999978315956550289911319

take first 20 least significant bits 1.99999999999999999978315956550289911319
reassign decimal point .78315956550289911319
raise to 9th power  .78315956550289911319^9
0.1108287183312072818811178137109474634212
divide by 2 0.0554143591656036409405589068554737317106

first 4 bytes -1  55414359XXXXXXXXXXX
Real private key  8993229949524469768

[Torin Keepler]

I tested pool 135 t.me/puzzle135, and after winning the 115-bit reward, I was not paid the reward.

If someone can explain to me how to attach files I'll attach the whole conversation with the administrator.

PLEASE BE CAREFUL

Yes, I logged in to double-check your information. Why are you lying and misleading people?
At the very least, you did not retract your statements. Yes, indeed, the 115th challenge that was announced by me was solved by you.
However, I reread the entire conversation history and found a link to the transaction showing that the full prize was transferred to you personally.
You are a dishonest and unethical person for not disclosing this here.

Of course, unless you confirm that you received the full reward. If you do confirm it, I will withdraw my opinion about you.

Below I am attaching a link to the blockchain transaction to the address you provided for the reward payout.
https://mempool.space/tx/5b2ed31ba4585d40126a0b7b73063f628c5e75833698e984d4d65b70f54208e5?mode=details

Here is your username, @ffb_15, in the pool group for solving the 135th puzzle.
https://t.me/puzzle135/14976/17611

And here is a link to another group where you personally state that you were paid the prize.
https://t.me/CryptoMathGroup/16217/17531

[dejanzl]

Puzzle 71 is 2.36 sextillion possibilities.

Lets assume euromillion odds, that 1 in 140 million.

A 4090 should be capable of 6 billion keys/second.

That means, if you run it constantly, then approx every 47 minutes you have the equivalent chances of a single euro millions ticket.

1/140.000.000=7e^-9
2^70/6.000.000.000 /28s =7e^-9

47 minutes? Am I doing something wrong? It shows 28 seconds...

Anyway, it has to be caled gambling

BR
Damian

Ahhh, maths is not my strong point, think I miscalculated somewhere lol.

Well, 28seconds is far better odds, still gambling, but some semblance of a chance.

However, you'd need to calculate power costs to determine if its more financially feasible to try to crack the puzzle or just buy a bunch of euro millions tickets every week.

puzzle 71 only has 1.18 sextillion possibilities inside of its range:  the first possibility starting at   1180591620717411303424 with eligible integers up to 2361183241434822606847,  therefor a search space of 1180591620717411303423 candidates.

A 4090 should be capable of 6 billion keys/second.
what the point to produce key when not produce hash160 and compare with address hash? so 4090 can produce ~ 1B or less hash160
my 3080ti now have 270M hash160 and comparation per seconds
according to AI range is 200-300M comparation/seconds for 3080ti
i split range in chunks, create server with database to mark checked chunks, server generate random chunk number and clients receive, recalculate chunk range and do sequence search,
but now with several gpu-s i can check only ~800-900 chunks per day
chunks is 10000x129x61440=79,257,600,000 keys
iter/baby_steps/threads
and have 14,895,626,675 chunks
i wait to try on 5070 to see speed...

| Device             | Realistic throughput (Puzzle 71) | Power     | Cost   |
| ------------------ | -------------------------------- | --------- | ------ |
| RTX 5090           | ~0.5–1 B keys/sec                | 450–600 W | ~$3–4k |
| Alveo U280         | 0.5–2 B keys/sec                 | 250–400 W | ~$8k   |
| Virtex UltraScale+ | 5–10 B keys/sec                  | 1,500 W   | $100k+ |
i try to find any shorcuts, early reject etc but
"for Puzzle 71 sequential key scanning, there’s essentially no safe algorithmic shortcut that can skip keys or change the math:
Every private key in the range must be checked — nothing can be skipped without risking a false negative.
Any shortcut like vanity-style prefix checks or partial hash rejection is unsafe, because the real solution could appear anywhere in the range.
The GPU kernel can only be optimized at the implementation level: EC math, batch inversion, warp-level sharing, register usage, memory layout. Those tweaks give small improvements (~5–20%), but nothing dramatic.
So your current kernel design — full EC → SHA256 → RIPEMD160 → Base58 check — is already essentially the optimal sequential solver on GPU hardware." by AI
there you can see
https://github.com/Wrappp/Btc-puzzle-71

[skedarve]

I've created some code. I'm completely broke, and for over two years I've been developing my own brute-force code. My PC only has four cores from 2010 with 8GB of RAM. I've managed to get it to generate 22 million private keys per second in Python using a 50-bit range. I managed to get the private key in just 32 minutes with my own Kangoo, in 0.22 seconds for 40 bits and 3 hours for 60 bits. I've converted the code to .c using only the cores. This 22 million k/s is now 198,000,000 k/s on just four cores. Now I've been using a graphics card from a friend who only lends me his PC when he can; it has an RTX 4090, and so far I've managed to generate 257,400,000,000 k/s. I know I still have a long way to go, but if anyone would like to join me or help, I'd be happy to teach them and combine our efforts.

[skedarve]

I've created some code. I'm completely broke, and for over two years I've been developing my own brute-force code. My PC only has four cores from 2010 with 8GB of RAM. I've managed to get it to generate 22 million private keys per second in Python using a 50-bit range. I managed to get the private key in just 32 minutes with my own Kangoo, in 0.22 seconds for 40 bits and 3 hours for 60 bits. I've converted the code to .c using only the cores. This 22 million k/s is now 198,000,000 k/s on just four cores. Now I've been using a graphics card from a friend who only lends me his PC when he can; it has an RTX 4090, and so far I've managed to generate 257,400,000,000 k/s. I know I still have a long way to go, but if anyone would like to join me or help, I'd be happy to teach them and combine our efforts.

Before I clarify, it's physically not possible on an RTX card, but mathematically it is.

[kTimesG]

A 4090 should be capable of 6 billion keys/second.
what the point to produce key when not produce hash160 and compare with address hash? so 4090 can produce ~ 1B or less hash160

A RTX 4090 is capable of more than 7 (perhaps even more) billion H160 target candidates per second (in practice, out of these 7 billion, only 1 or 2 need to be stored for extraction out of the GPU and fully checked).

The RTX 4090 can compute 20+ billion EC group ops per second. This is where "what's the point to produce" kicks in, since there might not be (or maybe there are) use cases for this - as long as its not about storing them in memory, or looking up each result, which is unfeasible.

[dejanzl]

A 4090 should be capable of 6 billion keys/second.
what the point to produce key when not produce hash160 and compare with address hash? so 4090 can produce ~ 1B or less hash160

A RTX 4090 is capable of more than 7 (perhaps even more) billion H160 target candidates per second (in practice, out of these 7 billion, only 1 or 2 need to be stored for extraction out of the GPU and fully checked).

The RTX 4090 can compute 20+ billion EC group ops per second. This is where "what's the point to produce" kicks in, since there might not be (or maybe there are) use cases for this - as long as its not about storing them in memory, or looking up each result, which is unfeasible.

EC group operations necessarily happen before HASH160, but you’re mixing two different throughput metrics that are not directly comparable. That’s what’s missing in your comparison.

For Bitcoin P2PKH, the pipeline is:

private key (k)
→ EC scalar multiplication: K = k · G        ← EC group ops
→ public key serialization (compressed/uncompressed)
→ SHA-256
→ RIPEMD-160
→ HASH160
→ compare with target(s)


So:
✅ EC group ops always happen before HASH160
❌ You cannot skip EC multiplication and jump directly to HASH160

1. “EC ops per second” ≠ “keys checked per second”

“RTX 4090 can compute 20+ billion EC group ops per second”

This number is theoretical / micro-benchmark-level and usually refers to:

Partial EC ops
Simplified point ops
Non-full scalar multiplications
Or instruction-level throughput

A full secp256k1 scalar multiplication costs:
~256 doublings
~128 additions (on average)
With modular arithmetic (mod p)
So 1 key ≠ 1 EC op

Modern GPUs such as the NVIDIA RTX 4090 can execute thousands of elliptic curve group operations per clock, but complete Bitcoin key checks require a full secp256k1 scalar multiplication per candidate key. Real-world benchmarks with optimized GPU kernels show hundreds of millions of full secp256k1 public keys per second per RTX 4090 (and multi-GPU rigs reaching ~7–10 billion keys/s aggregate), rather than tens of billions of isolated group ops per second. Since the HASH160 step (SHA-256 + RIPEMD-160) is relatively cheap, the dominant cost is the scalar multiplication itself — and that ultimately limits key search throughput.

[filo1992]

I tested pool 135 t.me/puzzle135, and after winning the 115-bit reward, I was not paid the reward.

If someone can explain to me how to attach files I'll attach the whole conversation with the administrator.

PLEASE BE CAREFUL

Yes, I logged in to double-check your information. Why are you lying and misleading people?
At the very least, you did not retract your statements. Yes, indeed, the 115th challenge that was announced by me was solved by you.
However, I reread the entire conversation history and found a link to the transaction showing that the full prize was transferred to you personally.
You are a dishonest and unethical person for not disclosing this here.

Of course, unless you confirm that you received the full reward. If you do confirm it, I will withdraw my opinion about you.

Below I am attaching a link to the blockchain transaction to the address you provided for the reward payout.
https://mempool.space/tx/5b2ed31ba4585d40126a0b7b73063f628c5e75833698e984d4d65b70f54208e5?mode=details

Here is your username, @ffb_15, in the pool group for solving the 135th puzzle.
https://t.me/puzzle135/14976/17611

And here is a link to another group where you personally state that you were paid the prize.
https://t.me/CryptoMathGroup/16217/17531

Hello,

As I wrote in other groups, I was paid. However, the payment was not due to confirmation of my resolution, but to stop spamming about what happened.

I don't need you to change your statements. As you indicated, and also by including my nickname, I am willing to talk to anyone who needs an explanation. And as you confirmed, I announced everywhere (except here, my mistake because I use this forum from my home computer) that I had written about the situation, that I had indeed been paid (...), and this confirms my honesty.

But if you are honest, you will also say that I was banned from the group because the administrator didn't like me. https://t.me/puzzle135/14976/17633
and that a user in the group actually said why I was paid =)

[Bram24732]

<LLM BS>

Could you guys please stop pasting chatGPT replies ? If we want to see ridiculously wrong answers we can generate those ourselves, thanks

[Torin Keepler]

I tested pool 135 t.me/puzzle135, and after winning the 115-bit reward, I was not paid the reward.

If someone can explain to me how to attach files I'll attach the whole conversation with the administrator.

PLEASE BE CAREFUL

Yes, I logged in to double-check your information. Why are you lying and misleading people?
At the very least, you did not retract your statements. Yes, indeed, the 115th challenge that was announced by me was solved by you.
However, I reread the entire conversation history and found a link to the transaction showing that the full prize was transferred to you personally.
You are a dishonest and unethical person for not disclosing this here.

Of course, unless you confirm that you received the full reward. If you do confirm it, I will withdraw my opinion about you.

Below I am attaching a link to the blockchain transaction to the address you provided for the reward payout.
https://mempool.space/tx/5b2ed31ba4585d40126a0b7b73063f628c5e75833698e984d4d65b70f54208e5?mode=details

Here is your username, @ffb_15, in the pool group for solving the 135th puzzle.
https://t.me/puzzle135/14976/17611

And here is a link to another group where you personally state that you were paid the prize.
https://t.me/CryptoMathGroup/16217/17531

Hello,

As I wrote in other groups, I was paid. However, the payment was not due to confirmation of my resolution, but to stop spamming about what happened.

I don't need you to change your statements. As you indicated, and also by including my nickname, I am willing to talk to anyone who needs an explanation. And as you confirmed, I announced everywhere (except here, my mistake because I use this forum from my home computer) that I had written about the situation, that I had indeed been paid (...), and this confirms my honesty.

But if you are honest, you will also say that I was banned from the group because the administrator didn't like me. https://t.me/puzzle135/14976/17633
and that a user in the group actually said why I was paid =)

It’s an interesting situation. You confirmed that you were paid only after I provided a link to the transaction.
I also contacted the pool administrator and asked him for a detailed report of what had happened.

The first thing he assured me of was that he had run many different challenges in the group and that the prizes were always paid to the winner.
As for your case, the administrator explained to me that he never wrote or announced that the prize would not be paid to you.
The only thing he asked of you was to give him a day or two to sort out the technical details, as he had noticed some inconsistencies.

The challenge organizer also said that he told you not to worry and that in the evening,
after analyzing all the data, he would write about it publicly in the group. You did not wait for his decision,
stirred up drama, and threw a tantrum, even though no one ever told you that you would not receive the prize.

In the end, after all the data was checked, you received a transfer to your address with the full prize amount of $200, but you chose to remain silent about it. Congratulations on that.

Also, for everyone interested in the topic of solving puzzles,
I would like to note that the above-mentioned pool solved the 115-bit challenge fairly quickly.
So there is a chance that the 135-bit puzzle will be solved as well. There is also some insider,
unconfirmed information that a $1,000 challenge among new participants of Pool-135 may be announced in the near future.

https://t.me/puzzle135/1

[filo1992]

...

If you want to start an argument here as well, that’s fine with me, but you are not creating any productive publicity for the pool.

If you have read the entire conversation (and are being honest), the administrator suspended the payment for verification purposes. Those verifications lasted 40 hours and were entirely at my expense, despite the administrator having assured me that the costs would be reimbursed. He also stated that, if the reward payment were accepted, a bonus would be added — as you well know, this never happened.

Furthermore (again, if you are being honest), the administrator consistently maintained that I would receive only half of the reward. As was stated in the Pool 135 group and in other groups owned by him (which you yourself linked), many people were convinced that what he was doing was not correct.

You talk to me about honesty and respect, yet you are the first to fail to respect your own words. For what reason did I receive the reward? Because it was publicly acknowledged that I was right, or simply to calm me down and make me stop explaining what had actually happened? (Be honest in your life and do not hide behind a nickname.)

Your administrator (or perhaps you yourself) is so “honest” that I was banned from the group simply because I am disliked by a single person within the pool. And he is so “honest” that he guaranteed within the pool that the data I shared would be used for Puzzle 135, yet my statistics were never updated.

You were quick to label me dishonest, yet you do not even seem to realize what you write and say. You claimed that if I confirmed the payment, you would change your position, but it appears you are unable to stand by what you write. You asked me to confirm it, and I did — and, as I wrote, I had already done so in several Telegram groups, providing links as proof. And now you come and tell me that I only confirmed it because you published the transaction in this forum?

So please, if you are honest and do not want negative publicity here as well, I would advise you to stop engaging in pointless controversy and accept the situation.