Bitcoin puzzle transaction ~32 BTC prize to who solves it

[Ykra]

If you know of any "state of the art" software that performs significantly better on this hardware, I’d be genuinely interested to learn.

RC's software.
https://github.com/RetiredC/RCKangaroo

https://bitcointalk.org/index.php?topic=5517607.0

[L_u_x]

If you know of any "state of the art" software that performs significantly better on this hardware, I’d be genuinely interested to learn.

RC's software.
https://github.com/RetiredC/RCKangaroo

https://bitcointalk.org/index.php?topic=5517607.0

Exactly. Those repositories are the foundation of modern hunting. You are absolutely right — RC’s logic is the engine under my hood.

However, simply cloning the repo isn't enough for the new generation of hardware. I’ve spent the last month optimizing the core math specifically for the Blackwell architecture (sm_120).

The result: A stable 10 GKeys/s per RTX 5090 (40 GKeys/s on a 4-card cluster). This is the baseline you get when the software is actually tuned for the silicon it runs on.

But let's move past the "whose shovel is better" debate. The desert we are digging (#135) is 2¹³⁴ bits wide. No single card, no matter how optimized, will ever solve it alone in a reasonable timeframe.

This is why I am proposing the "Shared Net" strategy:

1. The Multiplier: By selling/swapping the same 100M DP28 database to multiple hunters, we create a massive, dense web of "tame" paths.
2. Individual vs Collective: You aren't "helping" others; you are using a shared infrastructure to boost your own unique "wild" kangaroos. Your chance of hitting a collision increases exponentially with the size of the tame net you merge into your local work.
3. The Honest Race: People ask, "What if many people use the same points?" Here is the truth: if the key is found via this shared net, it becomes a pure speed and fee race in the Mempool. The one with the fastest node and the highest transaction fee takes the full 13.5 BTC.

This is the only honest way to solve #135. We share the cost of the infrastructure (the points), but we compete for the prize. It turns a 9,000-year solo grind into a high-stakes technical race that could end tomorrow.

My offer stands: 100M DP28 points (Pons-compatible). Buy the multiplier for $25 or swap your unique 50M+ base.



PM me if you're ready to play.

[kind_user]

You lost months when anyone_future_again built in 1 weekend ))
His version is for all architectures, not just Blackwell....
This guy anyone_future_again do not ask money and he doesn't care of money...he did that for pleasure to prove that this can be done...
You want to get rich? go to work...

[kTimesG]

Exactly. Those repositories are the foundation of modern hunting. You are absolutely right — RC’s logic is the engine under my hood.

First you said you use JLP's jumps (which is wrongly computed anyway!!!!), then you never heard of SOTA (state of the art) and now you claim you've got it covered. Make up your mind, these two are not even closely related, and don't even work on the same structure of the shifted range.

Dude, stop the BS. No one needs your DP database, it's dust in the eyes, and definitely NOT USABLE BETWEEN METHOD STRATEGIES. Also, 10 G/s is very very slow for 5090. Maybe spend 6 to 12 months actually writing code, you might  (if you're good at coding, not at Claude-ing or Codex-ing) see the minimum 15 G/s. But no one in their right mind would use 5090 cards right now, because the ROI is awful. Serious people already know this kind of shit.

[L_u_x]

Exactly. Those repositories are the foundation of modern hunting. You are absolutely right — RC’s logic is the engine under my hood.

First you said you use JLP's jumps (which is wrongly computed anyway!!!!), then you never heard of SOTA (state of the art) and now you claim you've got it covered. Make up your mind, these two are not even closely related, and don't even work on the same structure of the shifted range.

Dude, stop the BS. No one needs your DP database, it's dust in the eyes, and definitely NOT USABLE BETWEEN METHOD STRATEGIES. Also, 10 G/s is very very slow for 5090. Maybe spend 6 to 12 months actually writing code, you might  (if you're good at coding, not at Claude-ing or Codex-ing) see the minimum 15 G/s. But no one in their right mind would use 5090 cards right now, because the ROI is awful. Serious people already know this kind of shit.

You are getting tripped up by terminology. Let me clarify what the logs actually show.

1. Method: My setup uses the SOTA (Equivalence Classes / Negation Map) logic. My mention of JLP was strictly regarding the Jump Table Constants to ensure compatibility for buyers who use standard Pons-based tools.

2. Software: I am running a Blackwell-optimized build of RCKangaroo (SOTA implementation).

3. Compatibility: You claim SOTA DPs are "not usable between strategies." That is a half-truth. If two hunters use SOTA (Negation Map) and the same standard jump table, their DPs DO merge perfectly. Since RCKangaroo/Kang-1 is the dominant tool for #135, most  people are already on this protocol.

4. Speed: I repeat, 10 G/s per 5090 is what I get with the current CUDA port. If you have a SASS-optimized kernel that hits 15-20 G/s on Blackwell, show it.

The 100M DP28 database is SOTA-compatible.

[kTimesG]

You are getting tripped up by terminology. Let me clarify what the logs actually show.

1. Method: My setup uses the SOTA (Equivalence Classes / Negation Map) logic. My mention of JLP was strictly regarding the Jump Table Constants to ensure compatibility for buyers who use standard Pons-based tools.

~rest of BS...~

Stop trying. There is no "compatibility" for the real useful DP sets between different jump tables, even if they pass the "DP check" and even if they use the same jump function. The DP sets are disjunct, with utmost certainty (99.99999999.....999%) they do not share even a single DP point that will ever get hit.

And since the jump tables themselves are totally different between JLP and SOTA, it's pretty much the nail in the coffin for your pathetic attempts of trying to sell some useless data.

The math is pretty simple: 2**134 search space, DP 28, this means there are 2**106 distinct DPs.

But since the problem is solved after around 2**67 steps, this means only 2**39 DPs will ever get hit to solve any given problem.

Probability for some random DPs (generated from other jump table) to be useful is therefore around 1 in 2**106/2**39 == 1 in 2**67.

Conclusion: please stop with the abberations.

[L_u_x]

You are getting tripped up by terminology. Let me clarify what the logs actually show.

1. Method: My setup uses the SOTA (Equivalence Classes / Negation Map) logic. My mention of JLP was strictly regarding the Jump Table Constants to ensure compatibility for buyers who use standard Pons-based tools.

~rest of BS...~

Stop trying. There is no "compatibility" for the real useful DP sets between different jump tables, even if they pass the "DP check" and even if they use the same jump function. The DP sets are disjunct, with utmost certainty (99.99999999.....999%) they do not share even a single DP point that will ever get hit.

And since the jump tables themselves are totally different between JLP and SOTA, it's pretty much the nail in the coffin for your pathetic attempts of trying to sell some useless data.

The math is pretty simple: 2**134 search space, DP 28, this means there are 2**106 distinct DPs.

But since the problem is solved after around 2**67 steps, this means only 2**39 DPs will ever get hit to solve any given problem.

Probability for some random DPs (generated from other jump table) to be useful is therefore around 1 in 2**106/2**39 == 1 in 2**67.

Conclusion: please stop with the abberations.

[SELL/BUY/SWAP] 100M DP28 Database for #135 — The Industrial Protocol

To the critics: You are 100% correct. Your "1 in 2⁶⁷" math is the final nail in the coffin for random DP exchanges between incompatible tools. Thank you for refining my offer.

This offer is EXCLUSIVELY for those running RCKangaroo (SOTA version) with DP 28 and the standard jump table (Seed 0).

Since we are on the same "rails," forget the probability of a lone wanderer. Let’s talk about Industrial-Scale Hunting.

1. The Economics of the Absurd (#135)
The math is inexorable: To solve #135 (Range 2¹³⁴), even if you rented Elon Musk’s xAI cluster with 100,000 H100 GPUs, it would take you about 2.2 days. But renting that compute power for 54 hours would cost you over $10,000,000.
The prize is 13.5 BTC (~$1,000,000).
Brute-forcing this puzzle with pure money is a guaranteed $9,000,000 loss.

2. The "Shared Net" Philosophy (Anarcho-Pool)
How do we build a grid where a traditional pool is a financial disaster? We use game theory and individual greed.

Why am I selling a 100M point database (which cost me ~$300 in Blackwell server time) for only $25?
Because my unique "wild" kangaroos have already checked these paths. For me, they are empty. But for YOUR unique "wild" kangaroos, these 100M paths are a massive N² multiplier. I am selling you the infrastructure to recoup my costs and fund the next 100M points. You are buying time compression: 200 days of RTX 4090 work packed into one file.

Why am I eager to BUY your 100M database for $25?
1. Because a key might be found during the merge of our databases.
2. Because by doubling my database from 100M to 200M, I quadruple (4x) my total collision probability for all future searches. Spending $25 to get a 400% probability boost is the most rational trade in this game.

3. The Mempool Mexican Standoff
Let’s be honest about why people are afraid to swap. If a key is found on the intersection of our databases, our nodes will trigger `FOUND` at the exact same time. We instantly turn from partners into enemies.
This is a Speed and Fee Race. The winner is the one with the fastest node and the balls to set a 0.5 BTC transaction fee to get confirmed first. This is the only honest way: we share the net, but we fight for the fish.My Offer:

• SELL: 100M DP28 (RCKangaroo compatible) — $25.
  
• BUY: Your unique 100M DP28 (RCKangaroo compatible) — $25.
  
• SWAP: 1:1 with hunters.
  

Stop digging the desert with a spoon. Let's build an industrial net.
We can verify compatibility via 10k samples and handle payments via USDT/BTC.

PM me if you are on SOTA and ready to play for real.

[kind_user]

Sorry to say but 25$???wtf? You want to get rich? Go on ebay, maybe you have a chance...
Why to sell instead of giving free?
This is a forum, not a market...
25$ = 1 min of renting GPU online
This is a SCAM and i recommend to admins to ban this user.

[L_u_x]

Sorry to say but 25$???wtf? You want to get rich? Go on ebay, maybe you have a chance...
Why to sell instead of giving free?
This is a forum, not a market...
25$ = 1 min of renting GPU online
This is a SCAM and i recommend to admins to ban this user.

You clearly didn't read my post carefully. Let me make a special offer just for you, since you are so concerned about the "market" and "altruism."

I don't want your money. In fact, I want to GIVE you $25.

The Deal:
You provide me with a unique, verified 100M DP28 database specifically for Puzzle #135 (Range 2¹³⁴ to 2¹³⁵), generated with RCKangaroo/SOTA using the standard Seed 0.You give it to me "for free" as you suggested others should do, and in return, I will pay you $25 as a gift.

• If $25 is "getting rich" — here is your chance.
• If $25 is "1 minute of renting" — this should be the easiest profit of your life.

[kTimesG]

Stop trying.

This offer is EXCLUSIVELY for those running RCKangaroo (SOTA version) with DP 28 and the standard jump table (Seed 0).

Why am I selling a 100M point database (which cost me ~$300 in Blackwell server time) for only $25?
Because my unique "wild" kangaroos have already checked these paths. For me, they are empty. But for YOUR unique "wild" kangaroos, these 100M paths are a massive N² multiplier. I am selling you the infrastructure to recoup my costs and fund the next 100M points. You are buying time compression: 200 days of RTX 4090 work packed into one file.

We can verify compatibility via 10k samples and handle payments via USDT/BTC.

Well since the last nail of the coffin was hammered let's carefully throw it 6 feet under and pave the hole.

Even if someone uses RCKang (which is a proof of concept) and uses the "Seed 0" or whatever, and doesn't touch the jump logic, and the DP logic, and plays exactly the rules of the game, THERE ARE ZERO GUARANTEES YOUR DPs ARE LEGIT.

Why? Because anyone can simply scan through the range, discard any sort of jump logic, and just filter out points that pass the DP check.

Result: The speed is around two times faster, but the results (DP points) are useless.

For 25$:

RTX 4090: around 24 Gk/s scan speed
Price: 7 cents/hour
DP: 28

Total found DPs: 115 million

Your "offer": 100 million

So you're in a profit. End of story.

[L_u_x]

Stop trying.

This offer is EXCLUSIVELY for those running RCKangaroo (SOTA version) with DP 28 and the standard jump table (Seed 0).

Why am I selling a 100M point database (which cost me ~$300 in Blackwell server time) for only $25?
Because my unique "wild" kangaroos have already checked these paths. For me, they are empty. But for YOUR unique "wild" kangaroos, these 100M paths are a massive N² multiplier. I am selling you the infrastructure to recoup my costs and fund the next 100M points. You are buying time compression: 200 days of RTX 4090 work packed into one file.

We can verify compatibility via 10k samples and handle payments via USDT/BTC.

Well since the last nail of the coffin was hammered let's carefully throw it 6 feet under and pave the hole.

Even if someone uses RCKang (which is a proof of concept) and uses the "Seed 0" or whatever, and doesn't touch the jump logic, and the DP logic, and plays exactly the rules of the game, THERE ARE ZERO GUARANTEES YOUR DPs ARE LEGIT.

Why? Because anyone can simply scan through the range, discard any sort of jump logic, and just filter out points that pass the DP check.

Result: The speed is around two times faster, but the results (DP points) are useless.

For 25$:

RTX 4090: around 24 Gk/s scan speed
Price: 7 cents/hour
DP: 28

Total found DPs: 115 million

Your "offer": 100 million

So you're in a profit. End of story.

I appreciate the skepticism. It really does look like a coffin from that perspective. But I’d like to try and pull some of those nails out, because my understanding of the Kangaroo logic leads me to a slightly different conclusion.

It seems to me that we might be confusing a "Flag" with a "Road."

If I were just "scanning for zeros" (Sequential Scan), you would be 100% right. A scanned point is just a single coordinate in a 2¹³⁴ desert. The chance of a wild kangaroo landing exactly on that grain of sand is indeed 1 in 2⁶⁷ — practically zero.

But here is why I think these DPs are different:
In a Kangaroo walk, a DP is not just a point; it is the endpoint of a deterministic path.

• In my DP28 setup, each point is the result of a walk that is, on average, 268 million steps long.
  
• If a wild kangaroo lands on ANY of those 268 million points along that "road," it is mathematically forced to follow the path and land on the exact same DP.
  

My reasoning:
So, it seems to me that 100M Kangaroo DPs don't represent 100M "points," but rather a net covering 26.8 Quadrillion points.
By merging this into a local search that uses the same "rails" (the standard RC seed), a hunter isn't just looking for a collision at the endpoint — they are looking for a collision anywhere along the 268-million-step path.

Regarding the legitimacy:
I understand the fear of fake data. But a simple scan for zeros cannot produce a valid distance d that satisfies the d*G = X check for a synchronized walk. That is why I provide the sample. I just want to show that these "roads" are real and were actually walked.

Maybe my ROI math is off, and maybe $25 is too much to ask for a "Shared Net" experiment. But I’m just trying to find a way to pull those nails out and see if we can solve #135 through infrastructure instead of waiting 9,000 years alone.

If the math of path coalescence (Birthday Paradox on a graph) holds true, then these roads are not disjoint — they are traps.

What do you think? Am I missing something in the way paths converge?

[nomachine]

https://quantumai.google/static/site-assets/downloads/cryptocurrency-whitepaper.pdf

Google Research released a paper today on quantum computing with implications for modern cryptography. Their new estimate suggests that breaking current cryptographic systems might require around 500,000 qubits, significantly less than the previously assumed ~1 million. This reduction is substantial because earlier models generally assumed that a quantum attack capable of recovering a private key would take months. According to Google’s updated analysis, however, a sufficiently powerful quantum computer could potentially derive a private key from public key in about 9 minutes, which is shorter than Bitcoin’s average block time. One mitigation approach mentioned is MARA Slipstream, which sends transactions privately to miners so they can be included in a block without first appearing in the public mempool, reducing the window in which an attacker could attempt a quantum key recovery. The good news is that practical quantum computers capable of this are likely still around a decade away, leaving time for the ecosystem to deploy post-quantum cryptography and other protocol upgrades.

[kTimesG]

The good news is that practical quantum computers capable of this are likely still around a decade away, leaving time for the ecosystem to deploy post-quantum cryptography and other protocol upgrades.

Judging by the exponential rate in QC advancements, 10 years may be an over-estimate.

The bad news is, once 160 goes down (if I'd be Google, it would be the first thing to test my new shiny QC on), all we're left with are the daily prefix catalog updates Most likely, still of Puzzle 71, by that day.

And since 160 is way easier than a full 256-bit ECDLP, you do the math. Can they break it today if they want to? Maybe?

I understand the fear of fake data. But a simple scan for zeros cannot produce a valid distance d that satisfies the d*G = X check for a synchronized walk. That is why I provide the sample. I just want to show that these "roads" are real and were actually walked.

Yes it can. Have you ever heard of VanitySearch? Exclude the H160 step and there you have it: d*G = Q

You cannot prove that a DP belongs to a valid walk unless the walk is traversed from scratch, starting from some known point. Can we end this LLM-assisted non-sense please?

[kind_user]

Please someone ban and delete his messages with non-sense...
I rwad this forum because some guys are posting some interesting things.
Your posts are marketing and you fill this forum with a lot of marketing stuff...
If you have so much money, rent GPU's online and you habe your data...

But this is life sometimes...you don't know how to do it, so you want to get rich based on others work...

Maybe play the lottery...

[L_u_x]

I understand the fear of fake data. But a simple scan for zeros cannot produce a valid distance d that satisfies the d*G = X check for a synchronized walk. That is why I provide the sample. I just want to show that these "roads" are real and were actually walked.

Yes it can. Have you ever heard of VanitySearch? Exclude the H160 step and there you have it: d*G = Q

You cannot prove that a DP belongs to a valid walk unless the walk is traversed from scratch, starting from some known point. Can we end this LLM-assisted non-sense please?







Stop the BS. Any developer who actually understands the math of the Kangaroo algorithm knows this is false. You don't need to recompute the whole walk; you only need a Pre-step Proof.

The "Step-Back" Logic:
For any DP (X, d) in a legitimate database, I can provide the state exactly one jump before it reached the DP: P(n-1) and the JumpTable Index[/i].

The Math:
If P(n-1) + JumpTable[index] = DP, the walk is verified.

Why a "Scanner" fails here:
A sequential scanner (VanitySearch style) finds a point with 28 zeros by pure luck or brute force. It cannot produce a predecessor P(n-1) that satisfies the deterministic jump table. Finding such a predecessor without actually walking the path is as hard as solving the ECDLP itself.

On Legitimacy:
My database is not a collection of "random rocks." It is a verified chain of deterministic steps. If you were as "serious" as you claim, you would know that spot-checking 5 random points with their predecessors is a 100% proof of a valid walk.

Regarding the LLM:
 If you want to argue about "nonsense," look at your claim that a walk can only be proven "from scratch."

The roads are real. The math is verifiable. Stop digging your own grave with flawed logic.

[kind_user]

Lux user why are you still here for begging and not rich with your theory?

You try here to convince people with your theory that is reinventing the wheel....
Serious?
Try stopping use ChatGPT to convince people...
If you show me that you have at least 1 PHD in math or physics i believe you...

[pscamillo]

[PSCKangaroo] Fork of RCKangaroo — optimized for high-RAM + single GPU setups

Hi everyone,

I've been working on modifying RCKangaroo to get the most out of my specific hardware: an RTX 5070 paired with 128 GB of RAM. The original RCKangaroo is a brilliant piece of software — all credit to RetiredCoder for the SOTA algorithm — but I wanted to squeeze every bit of performance from my particular setup, where RAM is abundant but GPU is a single mid-range card.

The main idea behind PSCKangaroo is a TRAP/HUNT strategy:

• Phase 1 (TRAP): Fill the entire 128 GB of RAM with TAME distinguished points.
• Phase 2 (HUNT): Switch the GPU to 100% WILD kangaroos that check against the massive TAME table. WILDs are never stored — they are checked and discarded.

This effectively doubles the number of TAMEs compared to a balanced TAME/WILD split, which directly increases T-W collision probability per step.

To fit even more TAMEs into the same RAM, I implemented an ultra-compact 16-byte DP format (down from 25 bytes in the original). This gives +56% more entries at the cost of truncating 32 bits from the distance field. The truncated bits are recovered on collision via an async BSGS resolver running on CPU (4 threads, ~150ms per resolution). Yes, this introduces hash false positives that fail verification, but real collisions are resolved correctly by the BSGS step. The FP count is tracked in the stats — it's a known trade-off, not a bug.

Other features added:

• 3x endomorphism using secp256k1's β/λ constants (verified against bitcoin-core/secp256k1)
• XDP 8x (threshold-based DP detection, accepts 8 patterns instead of 1)
• Table freeze (no rotation when full, prevents FP explosion on long runs)
• Checkpoint system with auto-save and Ctrl+C safe exit

Validated by solving Puzzle #80 (known answer).

I'm sharing the code as-is for anyone who might find it useful. Feel free to use, modify, or improve it.

GitHub: https://github.com/pscamillo/PSCKangaroo
License: GPLv3 (same as the original RCKangaroo)

Feedback and contributions are welcome. I'm not claiming this is better than other approaches — it's just what works for my specific hardware profile (single GPU + lots of RAM). If you run a similar setup, it might help you too.

[nomachine]

And since 160 is way easier than a full 256-bit ECDLP, you do the math. Can they break it today if they want to? Maybe?

What worries me more is what it’ll break.

Banks.
SWIFT.
The NYSE.
Military networks.
Nuclear weapon systems.

Every encrypted website on the planet.

That’s straight-up ultimate weapon territory. 

[kTimesG]

Stop the BS. Any developer who actually understands the math of the Kangaroo algorithm knows this is false. You don't need to recompute the whole walk; you only need a Pre-step Proof.

<some math BS...~

Why a "Scanner" fails here:
A sequential scanner (VanitySearch style) finds a point with 28 zeros by pure luck or brute force. It cannot produce a predecessor P(n-1) that satisfies the deterministic jump table. Finding such a predecessor without actually walking the path is as hard as solving the ECDLP itself.

Thanks for proving that you are indeed as retarded as the LLM that generates the crap you're posting.

Finding some whatever predecessor is so simple that it can be done on paper. You only need to find some valid point with a satisfactory X. However, it doesn't prove that such predecessor was part of a longer walk. You'll simply say that's where the walk started. What the fuck is wrong with you, comparing this to ECDLP hardness?

You should indeed get banned, not for the selloff, but for the BS.

[kind_user]

I do not know if anybody observed the patterns gave by vanitysearch:

1PWo3JeB7g2ABtAGRkVRaYgZGDDKrvZZRB
1PWo3JeBQ7rA8A5YdP1paevYDxNnRv6cyv
1PWo3JeBQj1utAWiTv4YKVWK6KFLEPD2NE

Always is giving the next character after B one after another as is presented. This is not similar or coincidence because i receive almost all the time in a 64 bits search.

Also this search:

1PWo3JeB9jrGMLiH83vD775NRqHZMR2hHB
1PWo3JeBjrG9NriYKvHamQknJEydb7su5y
1PWo3JeBDVjrG9Y88qM5n7AwxxgP6hdozP
1PWo3JeBW8TqBSBjrGNEBmbxbRwCxWSTCK
1PWo3JeBSUCyQwhLQtGYc2LLzPmugjrGKT
1PWo3JeBVHrGwJJrMLkFCk4bZRPB9aRj6K--->This is very close because is existing only one 1PWo3JeB9jrGw for puzzle 71