[RangoWorld]

Ill top up to 300$ if it doesnt actually go up again dont worry

Looks like the hunt's over. Somebody must've brute-forced it with 9 words.

4.05 Sol transferred 8 minutes ago.

[Eran]

8.5+ billion possible combinations and that's only if the winner got all the 9 words correct. That's crazy. Congrats

[Three]

Looks like the hunt's over. Somebody must've brute-forced it with 9 words.

4.05 Sol transferred 8 minutes ago.

Yes.

I had guessed it when the solana address was shared. But if the address hadn't been shared, it wouldn't have happened today.

[Eran]

Yes.

I had guessed it when the solana address was shared. But if the address hadn't been shared, it wouldn't have happened today.

Second time in a row, nice dedication man.

[RangoWorld]

Yes.

I had guessed it when the solana address was shared. But if the address hadn't been shared, it wouldn't have happened today.

That's what I also pointed after the last hunt. In previous hunts the address wasn't shared. So you couldn't brute force it. Well you could technically but it would've been slower because you had to search through the whole Sol network to check balance and what not.

[Three]

Second time in a row, nice dedication man.

No. I not win. I congratulate the winner.

---

That's what I also pointed after the last hunt. In previous hunts the address wasn't shared. So you couldn't brute force it. Well you could technically but it would've been slower because you had to search through the whole Sol network to check balance and what not.

I definitely agree. I was certain this would be finished by Friday, but I didn't say anything. In my opinion, once the address is shared, there is no need for the 10th, 11th, and 12th riddles.

[Eran]

No. I not win. I congratulate the winner.

Ah my bad I guess I read that wrong. Nevertheless winner really went for it.

[Three]

Ah my bad I guess I read that wrong. Nevertheless winner really went for it.

The winner deserved it. My guess is that they rented a powerful graphics card; this process would have taken hours on most home computers.

[loadman]

10+ 4090

[Djangoo]

Cool! The competition is much tougher now, and that’s what makes it exciting. Djangoo enjoyed it. 🐧

[7]

Looks like the hunt's over. Somebody must've brute-forced it with 9 words.

4.05 Sol transferred 8 minutes ago.

For real someone won already?
That is madness!

I apologize if your disappointed but it was suggested to post the wallet address. Do we know who the winner was??

---

I need to know the winner to make the announcement.

I don't think there is much I can do to prevent this. I just hope the winner comes forward and shares their story and method for others to learn to enjoy.

---

That's what I also pointed after the last hunt. In previous hunts the address wasn't shared. So you couldn't brute force it. Well you could technically but it would've been slower because you had to search through the whole Sol network to check balance and what not.

No. I not win. I congratulate the winner.

---

I definitely agree. I was certain this would be finished by Friday, but I didn't say anything. In my opinion, once the address is shared, there is no need for the 10th, 11th, and 12th riddles.

So do you think i shouldn't share the address. The problem last time is people were complaining that the didnt know if someone won of if they just hit an empty wallet. IDK Omniscient what do you suggest?

---

There are even if the 9 answers they had are correct... That is about 8.59 billion combinations left.
Also to make you fully aware, the website itself including the db and the backend doesnt hold the phrase only the riddles....

I feel disspointed but will be even more dissapointed if the winner doesnt come forward

---

On the early solve — my full AI security review

I've seen the questions about the round being solved before the final riddles dropped, and I take that seriously. Rather than wave it away, I ran a complete review of the platform and the puzzle itself. Here's exactly where I landed.

1. The platform never holds the seed phrase — by design

This is the most important point, so I'll be blunt: my system does not store, and has never stored, the recovery phrase. There is no field for it, no "answer" to submit, and no endpoint that checks a solution. The prize wallet is controlled solely by the seed it was generated from.

That's a deliberate architectural choice. It means:

• There is nothing on my servers for anyone — including me — to leak, steal, or be socially-engineered out of.
  
• "Winning" is decided entirely on-chain: whoever reconstructs the phrase and moves the funds first, wins. The blockchain is the referee, not my website.
  

So the round wasn't "unlocked" on my end. It was solved the only way it can be — by reconstructing the phrase off-platform.

2. Could the unreleased riddles have been pulled early? No.

I audited every path that touches riddle content. Riddles are time-locked on the server side. Until a riddle's release moment passes, its text simply never leaves my infrastructure — it isn't sent to the browser, it isn't in any API response, and it isn't hidden in the page waiting to be revealed. The only things visible ahead of time are the riddle's number and its countdown — never the contents.

I checked the obvious attack angles you'd expect me to: session and login handling, cross-origin request protection, privileged/admin-only views, and every endpoint that returns competition data. None of them expose an unreleased riddle to a participant. I found no path — authenticated or otherwise — to read a future riddle ahead of its timer.

I also reviewed my access logs. The solver's account only ever loaded riddles that were already public at the time. There is no record — and architecturally no way to produce one — of an unreleased riddle being viewed.

3. So how was it solved "early"? The math of brute force.

This is the part people underestimate. A recovery phrase is a sequence of words drawn from a fixed 2,048-word list. Every word a solver deduces from the riddles removes a factor of 2,048 from the remaining search space. The clues aren't just flavour — each one collapses the problem enormously.

Because the wallet's public address is, by necessity, visible to everyone, a solver can test guesses entirely offline on their own hardware, checking each candidate against that public address. No interaction with me is required or even helpful for this stage.

Here's roughly what the remaining work looks like once you're down to the last few unknown words (positions known), on commodity GPU hardware:

Unknown words left	Raw combinations	Realistic effort
2	~4 million	Seconds — a single GPU
3	~8.6 billion	Minutes to a couple of hours — one strong GPU or a small rig
4	~1.7 x 10^13	Days on one GPU; a few hours on a serious cluster
5+	10^16 and up	Impractical without a large, dedicated farm

(The phrase's built-in checksum actually helps the solver here, discarding the overwhelming majority of invalid combinations instantly before any expensive work is done.)

The takeaway: once enough riddles are public, the residual is no longer "wait for the rest" — it's a short, very winnable compute race. Someone who deduced the bulk of the words and pointed even a modest amount of GPU power at the remainder could close it out in minutes to hours. That isn't a flaw — that is the game. Cracking the tail is precisely what "crack the seed" means, and the design rewards whoever combines the best deduction with the resources to finish.

4. Bottom line

• The platform was not breached, and unreleased riddles were not accessible early — I verified this across every relevant component.
  
• I never custody the phrase; the contest is settled on-chain.
  
• The early solve is consistent with a legitimate brute-force of the final unknown words — a known, expected outcome of this style of puzzle once enough clues are out.
  

I'm confident in the integrity of the round. I'm also taking the lessons forward into how I balance future puzzles so the difficulty curve holds right to the final clue. Thanks to everyone who raised this — scrutiny like this is exactly what keeps these events fair.

---

5. "But the wallet address is public — doesn't that make it crackable?"

Short answer: no. Sharing the Solana address does not make the seed phrase any easier to crack, and it reveals none of the words.

Here's why:

• It reveals zero words. The address is the end of a one-way chain: phrase → seed → key → address. You cannot run that backwards. Knowing the address tells you nothing about a single word in the phrase.
  
• It doesn't shrink the search space. The number of possible phrases is exactly the same whether the address is public or not. The address removes none of the difficulty — every candidate still has to be derived and checked in full.
  
• It's unavoidable and intentional. Any on-chain prize must have a public address — that's how you can all see the pot is real, see the balance, and verify the eventual claim transaction. A competition where the prize wallet was hidden wouldn't be verifiable or trustworthy.
  

A common misconception is that a Solana address "is" the public key, so it must be weaker. It's true that a Solana address is the public key — but that changes nothing here. Reversing a public key back to a private key (or to the seed phrase) is not something any normal hardware can do; it would require breakthroughs that would also break essentially all of modern cryptocurrency at once. It is not a factor in this competition.

Bottom line: the public address is a reference point for honesty and verification, not a shortcut. It lets you confirm the prize and check the result — it does not hand anyone the phrase, and it does not make the phrase any faster to guess. The only thing that ever narrows the phrase down is the riddles themselves.

---

I cannot do anything further than this. The game is called crack the seed and someone cracked it. Whether they needed all the clues or not it was cracked, was that cheating? No, not really. Does that show we have some skilled people on the forum finally... yes. My advice for anyone wanting to take part in the next round.... learn that which made you behind this time. It might not of been the plan but it educational and progressive to learn how it was done.

Thank you to everyone for taking part. I cannot comment anymore about this. I hope those other than the winner take this well and make them think how to better themselves in the further rounds.. if everyone is still interested. With new members joining just for the game from X i was expecting a rise in the competition skill level. I hope they stick around on HF.

[RangoWorld]

For real someone won already?
That is madness!

I apologize if your disappointed but it was suggested to post the wallet address. Do we know who the winner was??

---

I need to know the winner to make the announcement.

I don't think there is much I can do to prevent this. I just hope the winner comes forward and shares their story and method for others to learn to enjoy.

---

So do you think i shouldn't share the address. The problem last time is people were complaining that the didnt know if someone won of if they just hit an empty wallet. IDK Omniscient what do you suggest?

---

There are even if the 9 answers they had are correct... That is about 8.59 billion combinations left.
Also to make you fully aware, the website itself including the db and the backend doesnt hold the phrase only the riddles....

I feel disspointed but will be even more dissapointed if the winner doesnt come forward

---

On the early solve — my full AI security review

I've seen the questions about the round being solved before the final riddles dropped, and I take that seriously. Rather than wave it away, I ran a complete review of the platform and the puzzle itself. Here's exactly where I landed.

1. The platform never holds the seed phrase — by design

This is the most important point, so I'll be blunt: my system does not store, and has never stored, the recovery phrase. There is no field for it, no "answer" to submit, and no endpoint that checks a solution. The prize wallet is controlled solely by the seed it was generated from.

That's a deliberate architectural choice. It means:

• There is nothing on my servers for anyone — including me — to leak, steal, or be socially-engineered out of.
  
• "Winning" is decided entirely on-chain: whoever reconstructs the phrase and moves the funds first, wins. The blockchain is the referee, not my website.
  

So the round wasn't "unlocked" on my end. It was solved the only way it can be — by reconstructing the phrase off-platform.

2. Could the unreleased riddles have been pulled early? No.

I audited every path that touches riddle content. Riddles are time-locked on the server side. Until a riddle's release moment passes, its text simply never leaves my infrastructure — it isn't sent to the browser, it isn't in any API response, and it isn't hidden in the page waiting to be revealed. The only things visible ahead of time are the riddle's number and its countdown — never the contents.

I checked the obvious attack angles you'd expect me to: session and login handling, cross-origin request protection, privileged/admin-only views, and every endpoint that returns competition data. None of them expose an unreleased riddle to a participant. I found no path — authenticated or otherwise — to read a future riddle ahead of its timer.

I also reviewed my access logs. The solver's account only ever loaded riddles that were already public at the time. There is no record — and architecturally no way to produce one — of an unreleased riddle being viewed.

3. So how was it solved "early"? The math of brute force.

This is the part people underestimate. A recovery phrase is a sequence of words drawn from a fixed 2,048-word list. Every word a solver deduces from the riddles removes a factor of 2,048 from the remaining search space. The clues aren't just flavour — each one collapses the problem enormously.

Because the wallet's public address is, by necessity, visible to everyone, a solver can test guesses entirely offline on their own hardware, checking each candidate against that public address. No interaction with me is required or even helpful for this stage.

Here's roughly what the remaining work looks like once you're down to the last few unknown words (positions known), on commodity GPU hardware:

Unknown words left	Raw combinations	Realistic effort
2	~4 million	Seconds — a single GPU
3	~8.6 billion	Minutes to a couple of hours — one strong GPU or a small rig
4	~1.7 x 10^13	Days on one GPU; a few hours on a serious cluster
5+	10^16 and up	Impractical without a large, dedicated farm

(The phrase's built-in checksum actually helps the solver here, discarding the overwhelming majority of invalid combinations instantly before any expensive work is done.)

The takeaway: once enough riddles are public, the residual is no longer "wait for the rest" — it's a short, very winnable compute race. Someone who deduced the bulk of the words and pointed even a modest amount of GPU power at the remainder could close it out in minutes to hours. That isn't a flaw — that is the game. Cracking the tail is precisely what "crack the seed" means, and the design rewards whoever combines the best deduction with the resources to finish.

4. Bottom line

• The platform was not breached, and unreleased riddles were not accessible early — I verified this across every relevant component.
  
• I never custody the phrase; the contest is settled on-chain.
  
• The early solve is consistent with a legitimate brute-force of the final unknown words — a known, expected outcome of this style of puzzle once enough clues are out.
  

I'm confident in the integrity of the round. I'm also taking the lessons forward into how I balance future puzzles so the difficulty curve holds right to the final clue. Thanks to everyone who raised this — scrutiny like this is exactly what keeps these events fair.

---

5. "But the wallet address is public — doesn't that make it crackable?"

Short answer: no. Sharing the Solana address does not make the seed phrase any easier to crack, and it reveals none of the words.

Here's why:

• It reveals zero words. The address is the end of a one-way chain: phrase → seed → key → address. You cannot run that backwards. Knowing the address tells you nothing about a single word in the phrase.
  
• It doesn't shrink the search space. The number of possible phrases is exactly the same whether the address is public or not. The address removes none of the difficulty — every candidate still has to be derived and checked in full.
  
• It's unavoidable and intentional. Any on-chain prize must have a public address — that's how you can all see the pot is real, see the balance, and verify the eventual claim transaction. A competition where the prize wallet was hidden wouldn't be verifiable or trustworthy.
  

A common misconception is that a Solana address "is" the public key, so it must be weaker. It's true that a Solana address is the public key — but that changes nothing here. Reversing a public key back to a private key (or to the seed phrase) is not something any normal hardware can do; it would require breakthroughs that would also break essentially all of modern cryptocurrency at once. It is not a factor in this competition.

Bottom line: the public address is a reference point for honesty and verification, not a shortcut. It lets you confirm the prize and check the result — it does not hand anyone the phrase, and it does not make the phrase any faster to guess. The only thing that ever narrows the phrase down is the riddles themselves.

---

I cannot do anything further than this. The game is called crack the seed and someone cracked it. Whether they needed all the clues or not it was cracked, was that cheating? No, not really. Does that show we have some skilled people on the forum finally... yes. My advice for anyone wanting to take part in the next round.... learn that which made you behind this time. It might not of been the plan but it educational and progressive to learn how it was done.

Thank you to everyone for taking part. I cannot comment anymore about this. I hope those other than the winner take this well and make them think how to better themselves in the further rounds.. if everyone is still interested. With new members joining just for the game from X i was expecting a rise in the competition skill level. I hope they stick around on HF.

I mean it is impressive. Respect to the winner for cracking it.

With known public address. You can make brute force attempt offline. If you know 9 words already.

Winner most likely rented a single or couple high end GPU like a H200 and were ready to attempt it as soon as riddle 9 was received.

I also want to know who the winner is.

[7]

I mean it is impressive. Respect to the winner for cracking it.

With known public address. You can make brute force attempt offline. If you know 9 words already.

Winner most likely rented a single or couple high end GPU like a H200 and were ready to attempt it as soon as riddle 9 was received.

I also want to know who the winner is.

I have a plan for the next round to make it harder and for future rounds adding additional requirements.

[Mix]

I have a plan for the next round to make it harder and for future rounds adding additional requirements.

Bring on Crack The Seed v5.

Mad to see how fast this one was cracked everyone, great job!

[zorenx]

Hi 7!

First of all, sorry for spoiling the fun. I was the one who cracked the wallet. Sorry just now I was afraid of some backlash from fellow hunters.

I had been preparing for this challenge for several days because I expected the real race to begin once 9 out of 12 words were revealed.

While it’s true that 3 missing BIP39 words correspond to roughly 8.6 billion combinations, it’s not as impossible as it first appears.

My approach

I first wrote a CPU-based recovery script. It achieved only around 90–110k candidates/sec, which meant a full 3-word search would take far too long. I knew that by the time the 10th word was revealed, other players would likely finish before me.

Then I remembered that my machine has an RTX 3070. My original implementation wasn’t actually utilizing the GPU properly, so I rewrote and optimized it to run on the GPU instead.

Another optimization came from the BIP39 checksum.

For a 12-word BIP39 mnemonic:
Total entropy: 128 bits
Checksum: 4 bits

This means only 1 out of every 16 possible mnemonics has a valid checksum. My script took advantage of this and only performed the expensive seed derivation on checksum-valid candidates, reducing the work by roughly 16×.

Instead of testing the full ~8.6 billion combinations, it only needed to process about 536 million checksum-valid candidates.

I still wasn’t completely confident in my riddle answers, but I decided to give it a shot anyway.

For anyone who wants to verify that I was indeed the one who recovered the wallet, here’s the mnemonic:

protect cigar sphere father nephew still toe maximum nasty among differ game

Thank you for the fun riddles, 7! I hope this won't ban me from participating in V5 :(