kTimesG's Phy Challenge
June 5, 2026. Bitcoin had its worst week since 2022, and kTimesG answered with a puzzle. Same kTimesG as the 80-bit challenge, different game. This time there's no range to search. The post promises "no loose ends, no hashing."
kTimesG's Phy Challenge
bitcoin
A signed forum post, a signature the author's cat cut one character short, and grey emojis spiralling around a gold coin. The signature gives up the address. The key behind it, and 800,000 sats, nobody has found yet.
- Author
- kTimesG
- Prize recorded
- 0.008 BTC
- Still unsolved
- 1 · 0.008 BTC
Census
- unsolved
- 1100%
- public keys
- 1100%
- private keys
- 00%
Access
- Load
getCollection("phy") - Showphy in the playground
What's in the first post
Three things, and the author keeps saying that's all you need.
A text that sounds familiar. It's the abstract of the Bitcoin whitepaper with words swapped. "Electronic cash" became "modular values". Timestamps became "spacestamps", proof of work became "phylo-based proof-of-equivalence". Then a news headline and the line about a puzzle for everyone.
A Base64 string under it. That's a signed message signature, and it signs the whole text above it.
And a picture, 2048 by 2048. Grey emojis in a spiral around one golden Bitcoin coin. Hands, faces, a cat, a key, a padlock. All in greyscale, except the coin.

The prize showed up two weeks later
On June 21 the author came back: there's a prize "as of today", and one character is missing from the second code. Plus a PGP message, with "the password's already mentioned." The same evening, 800,000 sats landed on a native SegWit address.
Nobody had posted that address yet. Where does it come from?
The cat ate a J
The Base64 came out at 64 and a half bytes, and a signed message signature takes 65. People tried E, A, H. Someone decoded the PGP hint and got told "Congrats on decoding the puzzle hint. But it's not the missing char."
On July 29 NotATether posted "It's J." With the J in front, the signature verifies against the whole signed text, no trailing newline. It recovers the public key 02425afd…6650eb and the address bc1qrpn28qa82uyjg37dvsz3w7wpm3kpdea957nm9p. The funded one.
kTimesG fixed the post the same day. The explanation is the best line in the thread: "My cat is not part of this puzzle, yet somehow he managed to create the J incident." Ouch. Three days of building and checking the puzzle, and the cat got the last word.
Later someone doubted the address. The author's answer was short: "Why would you think the address is wrong, since the signature verifies?"
A signature that looks wrong on purpose
Decode the Base64 and look at s. 23 zero bytes, then 9 bytes of data. A random signature never looks like that. Somebody built this one.
So what does the post really hold? Everyone in the thread got stuck around here. NotATether ran Kangaroo over almost the whole 80-bit range. Nothing. The author's verdict: "no idea where it was ever hinted anything about using Kangaroo." The data speaks for itself, apparently.
The hints since
kTimesG doesn't hand out much. The record keeps 18 hints, the author's own plus two from other players. The ones people keep quoting:
- July: "the hints are already inside-out", and everyone is "stuck at the ASCII 8 cat incident".
- September: a meta clue,
gvonys fhelucrM kbZ, to "wake up or confuse even more your LLMs". - Skin tones are part of it. Paint every emoji default yellow and the answer holds only "if you already know that the image was redrawn."
- "The smart approach can find the correct solution in 4 milliseconds."
- And on September 23, a post that just says "This is a hint."
Four milliseconds. Not a GPU farm, not a week of Kangaroo. Whatever the trick is, it's small. Is it hiding in plain sight?
What the Phy record carries
export const phyChallenge = puzzle({
id: "phy",
chain: "bitcoin",
address: "bc1qrpn28qa82uyjg37dvsz3w7wpm3kpdea957nm9p",
sourceUrl: OP,
startedAt: "2026-06-05 19:15:46",
pubkey: compressed("02425afdd1716149faf414b6fdb96d5e7afc8ce42496042f4df559c80a7c6650eb"),
prize: 0.008,
hints: [/* the author's posts, the missing J with its answer, two from the thread */],
transactions: [funding(/* 800,000 sats on June 21 */)],
assets: assets({ puzzle: "puzzle.png", sourceUrl: PICTURE, digests: [digest(/* … */)] }),
});
The address comes from the signature
No post by the author ever spelled out the address. It comes from the signed text, and anyone can check it with keys_message_recover from @agntn/keys. Feed it the post and the signature with the J, plus the = padding the post leaves off. Out comes 02425afd…6650eb and the funded address. Feed it only the last line and some other address comes out. That's the whole proof, and the record keeps the public key it gives.
The same hint, twice
The June 21 post looks different today. When kTimesG fixed the J, the sentence about the missing character got struck through. The record keeps the original wording, with a June 29 capture as confirmation. The J itself is the hint's answer, credited to NotATether.
No technique yet
Nobody published how the puzzle works, so the record names no technique. Guessing one from the thread would be the kind of LLM essay the author keeps asking people to stop sending. Once a solution is out, it gets one.
The Phy Challenge record
bitcoinp2wpkh unsolved
Record
puzzle({
chain: "bitcoin",
address: "bc1qrpn28qa8…a957nm9p",
})- Prize
- 0.008 BTC
- Key material
- none published
- Verification
- unverifiable
- Solved
- not yet
- Trail
- Explorerblockstream.info
- Sourcebitcointalk.org/index.php?topic=5584952.msg66802369
- Tryphy in the playground
- Address
- bc1qrpn28qa82uyjg37dvsz3w7wpm3kpdea957nm9p
hash160 1866a383a757092447cd64051779c1dc6c16e7a5
Key
- public key · compressed
- 02425afdd1716149faf414b6fdb96d5e7afc8ce42496042f4df559c80a7c6650eb
Transactions
- funding2026-06-210.008 BTC2ba79d0b37…5964d3
Hints
- 2026-06-05
officiallooks bad this bearish cycle around and no end is in sight, so here's a puzzle for everyone, no loose ends, no hashing.
Sourcebitcointalk.org/index.php?topic=5584952.msg66802369Confirmweb.archive.org
- 2026-06-21
officialThere is currently a missing char in the second code of the first post. I will fix it after/if someone figures it out.
Sourcebitcointalk.org/index.php?topic=5584952.msg66860784Confirmweb.archive.org
- 2026-06-21
officialThe rest of the provided parts are all correct and complete, and there is a prize (as of today). If you reach a state where you believe "this cannot be solved", it means you are missing the direct and indirect clues along the way.
Sourcebitcointalk.org/index.php?topic=5584952.msg66860784Confirmweb.archive.org
- 2026-06-21
officialA final useful sign (the password's already mentioned). -----BEGIN PGP MESSAGE----- jA0ECQMKVwglgSLxlRn/0jcBn0f9h3kbilK/NDDWZCkuONpev7JZbbw+l6uVZC0N aJFXTLG6sbLXr6t0z3scptSM3vWBCmOU =D+ub -----END PGP MESSAGE-----
Sourcebitcointalk.org/index.php?topic=5584952.msg66860784Confirmweb.archive.org
- 2026-06-29
officialCongrats on decoding the puzzle hint. But it's not the missing char.
- 2026-07-23
officialI don't know what to hint at, when the hints are already inside-out, and everyone seems to be stuck at the ASCII 8 cat incident, or don't want to share what they found.
- 2026-07-29
officialTo progress before wasting any compute power, the data speaks for itself, which creates some immediate conclusions, so no idea where it was ever hinted anything about using Kangaroo.
- 2026-08-13
officialWhy would you think the address is wrong, since the signature verifies?
- 2026-08-22
officialYou got pretty close with 97% of the recoloring, nice, though a much easier and precise normalization than grayscale-to-tone exists, if some overall observations are made.
- 2026-08-28
officialIf you're referring to the extra LF, no, but I can't fix that for obvious reasons.
- 2026-08-28
officialGuys, stop sending DMs and LLM essays about the puzzle, everything required to solve is in the first post, and it's actually much easier to solve then I anticipated. I already gave too many hints as it is.
- 2026-09-05
officialmeta-clue to either wake up or confuse even more your LLMs, as they seem to have crossed off-road... gvonys fhelucrM kbZ
- 2026-09-06
officialThere's no "useless" information added into the puzzle. If all of you are thinking some LLM will magically deduct information that's visible in plain sight by anyone, you should understand that LLMs can only understand data they were trained about, and this puzzle is not one of those things.
- 2026-09-07
officialOnly if you already know that the image was redrawn and requires to test the skin-tone to fix the image.
- 2026-09-15
officialThe third one helps as well. But anyway, all the clues I've given were already in the puzzle, since the puzzle is already self-explaining itself. The smart approach can find the correct solution in 4 milliseconds.
- 2026-09-23
officialThis is a hint.
- 2026-06-20
communityPhyllotaxis ?
- 2026-07-29
communityThe signature signs the complete OP text with no trailing newline. It recovers the funded SegWit address bc1qrpn28qa82uyjg37dvsz3w7wpm3kpdea957nm9p, currently holding 800,000 sats.