🧾 COLDCARD: WHAT HAPPENED, WHO IS INVOLVED & WHAT PLEBS SHOULD DO
Color code
🟦 CODE / PRODUCT
🟨 WARNING / INVESTIGATION
🟧 PUBLIC COMMUNICATION
🟩 CONFIRMED / FIX / RESPONSE
🟥 THEFT / ATTACKER ACTIVITY
💔 VICTIM ACTION
⬜ UNKNOWN / NEED MORE EVIDENCE
Important: A person’s name appears because they have a documented role in the product, investigation, research, communication, or response. That does not mean they caused the vulnerability or participated in the theft. The attackers remain unidentified.
⸻
🟦 2021: THE VULNERABILITY ENTERS THE PRODUCT
January–March 2021
Coldcard integrates libNgU into its firmware.
People/projects relevant to this stage:
switck #libNgU
Peter D. Gray / doc-hex #ColdcardFirmware
Coldcard/Coinkite developers #Development
The critical integration ultimately allowed wallet seed generation to reach a deterministic software PRNG rather than the intended hardware random-number generator.
For affected older Mk2/Mk3 devices, later analysis puts effective entropy around 40 bits instead of the intended 128 bits.
That turns an astronomically large seed-search problem into something attackers can potentially enumerate.
⸻
🟦 MARCH 1, 2021: THE CRITICAL CODE PATH CHANGES
Coldcard seed generation moves into the new RNG architecture.
The later reconstruction is essentially:
Generate seed
↓
random.bytes()
↓
ngu.random.bytes()
↓
wrong rng_get() implementation
↓
software PRNG
↓
predictable-enough seed space
⬜ Questions still worth answering:
Who reviewed this specific change?
Was RNG behavior tested on actual production hardware?
Who verified which rng_get() implementation ended up in the finished binary?
⸻
🟦 MARCH 17, 2021: FIRMWARE 4.0.0 SHIPS
Coldcard itself described the release as containing major internal cryptographic/BIP39 changes.
People relevant organizationally:
Rodolfo Novak / NVK #CoinkiteLeadership
Peter D. Gray / doc-hex #ColdcardFirmware
switck #libNgU
Again, leadership involvement doesn’t establish authorship of the vulnerable code.
⸻
🟨 MARCH 29, 2021: ANOTHER SECURITY ISSUE IS FIXED
Coldcard 4.0.1 fixes a security problem in 4.0.0.
We have not established that this was the RNG problem.
⬜ But it creates an important historical question:
After finding a security problem immediately following a major cryptographic rewrite, was the rest of that rewrite comprehensively reviewed?
⸻
⏳ 2021 → 2026: THE FIVE-YEAR GAP
This is one of the central mysteries.
The vulnerable code exists.
The source is public.
Coldcard continues being used.
People generate wallets.
Yet the entropy defect apparently remains undetected.
This is why:
Open source means people CAN inspect code. It doesn’t mean somebody DID successfully verify every security-critical path.
⸻
⬜ 2021/22: CLAIMED EARLIER WARNING
James O’Beirne #RNGResearch reportedly later said another person had raised an earlier concern.
We still need the original communication before treating that as established notice.
Status: LEAD, NOT CONCLUSION.
⸻
⬜ 2022: REPORTED SECURITY REVIEW
There are claims concerning an outside security review containing RNG-related recommendations.
The original report remains important evidence to obtain.
Questions:
What exactly did it say?
Who received it?
What recommendations were implemented?
⸻
🟨 MAY 2025: O’BEIRNE SAYS HE RAISES RNG CONCERNS
James O’Beirne #RNGResearch
O’Beirne says he examined the Coldcard RNG architecture, became concerned and raised the matter with Coinkite.
⬜ The crucial missing evidence remains the original May 2025 correspondence.
There is an enormous difference between:
“I don’t like this library.”
and:
“Your seed generator may not actually be using the hardware RNG.”
Until we see the communication, don’t collapse those possibilities together.
⸻
⬜ JUNE 2026: REPORTED AI-ASSISTED SECURITY REVIEW
A review shortly before the attack has been reported.
What we need:
scope
findings
prompts
RNG coverage
seed-generation coverage
Until those records are available, this remains an investigative question rather than proof someone missed a known vulnerability.
⸻
🟥 JULY 30, 2026: THE BITCOIN STARTS MOVING
UNKNOWN ATTACKER(S) #UnknownAttackers
Attackers apparently enumerate vulnerable Coldcard-generated seeds offline, derive addresses and sweep wallets.
They don’t need to steal the physical Coldcard.
They don’t need the victim to plug it in.
They don’t necessarily need to phish the victim.
They can attack the weak seed itself.
Galaxy now confirms attackers began exploiting affected wallets at least by the early morning of July 30.
⸻
🟨 JULY 30: PLEBS START SOUNDING THE ALARM
Victims report unexpected wallet drains.
Then investigators begin connecting apparently unrelated cases.
Kevin Loaec #EarlyWarning
warns Coldcard users to check balances.
↓
Rodolfo Novak / NVK #CoinkiteResponse
initially expresses skepticism that the reports demonstrate a systemic Coldcard failure.
↓
Rob Hamilton #OnChainInvestigation
examines blockchain activity and identifies a larger pattern.
↓
Kevin Loaec #RNGResearch
moves toward the weak-entropy hypothesis.
↓
James O’Beirne #UserWarning
publicly warns affected users to move bitcoin.
↓
Rodolfo Novak / NVK #CoinkiteResponse
changes his public position as evidence accumulates and Coinkite investigates.
↓
Greg Sanders / instagibbs #TechnicalReproduction
independently reproduces/demonstrates the technical failure.
That’s the critical transition:
🟨 We think something is wrong
becomes
🟩 We can reproduce what’s wrong.
⸻
🟩 JULY 30–31: COINKITE CONFIRMS THE ENTROPY PROBLEM
Emergency remediation follows.
And here’s the thing every pleb needs burned into memory:
Updating firmware does NOT fix a vulnerable seed.
The weakness is embedded in the seed that already exists.
A vulnerable wallet needs to be migrated to a newly generated safe seed, following current official guidance.
The newest reporting continues to describe the failure as a software PRNG being reached instead of the intended hardware entropy source.
⸻
🟧 THE WARNING NETWORK FORMS
People helping communicate/respond include:
Matt Odell / ODELL #UserWarning
Calle #SecurityResponse
Ben Perrin / BTC Sessions #PublicEducation
James O’Beirne #RNGResearch
Kevin Loaec #EarlyWarning
Rob Hamilton #OnChainInvestigation
Greg Sanders / instagibbs #TechnicalReproduction
They appear here because of their respective public roles.
They are not being identified as attackers.
⸻
🟥 JULY 30 → AUGUST 6: MULTIPLE ATTACKERS
Galaxy has now confirmed something particularly important:
There wasn’t merely one identifiable attacker footprint.
Researchers have identified at least 33 additional attacker footprints beyond the major waves and say with high confidence that multiple attackers were exploiting the vulnerability.
They cannot determine whether every footprint represents a different person.
That means don’t turn:
“multiple attacker footprints”
into:
“33 hackers.”
We don’t know that.
⸻
🟥 CURRENT CONFIRMED DAMAGE
As of Galaxy’s August 14 accounting:
190 victims directly contacted
8,600+ addresses
1,778.84 BTC confirmed stolen
approximately $112.7 million at the valuation Galaxy used.
And importantly:
1,531 BTC was still sitting unmoved in attacker-controlled addresses.
Approximately 246 BTC had moved onward.
Of those moved funds, about 65% entered CoinJoin transactions, while the remainder continued through other on-chain paths, sometimes peel chains. Small amounts reached exchanges or interchain bridges.
Galaxy has supplied attacker-address information to exchanges, compliance companies, investigators and law enforcement so funds might be frozen if they reach centralized intermediaries.
So:
Stolen does NOT necessarily mean investigators should give up.
⸻
💔 NOW: “MY BITCOIN IS STILL THERE”
This person needs action, not investigation Twitter.
First determine:
Coldcard model
firmware that generated the seed
approximately when the seed was generated
whether dice entropy was added
whether a strong unique BIP39 passphrase exists
whether BIP85 child wallets were derived
whether multisig is involved
Then follow the current official Coldcard migration instructions, not random screenshots or DMs.
Do not simply:
update firmware → keep same vulnerable seed → assume safe.
That misses the core problem.
⸻
💔 “MY BITCOIN IS ALREADY GONE”
Do not wipe everything in panic.
Preserve evidence.
Create a folder containing:
Coldcard information
Model
Firmware
Purchase date
Approximate seed-generation date
Wallet information
Public Bitcoin addresses
Transaction history
Whether singlesig/multisig
Whether BIP85 was used
Whether dice were used
Whether a passphrase existed
Theft information
TXID
Date/time
Amount
Destination address(es)
Screenshots
Supporting records
Coldcard purchase receipt
Exchange withdrawals showing how BTC reached the wallet
Relevant support emails
Warnings received
Communications concerning the incident
But:
🚨 NEVER PUT THE SEED PHRASE IN THE EVIDENCE PACKET.
Also never send:
BIP39 passphrase
XPRV
PIN
private keys
dice-roll sequence used to generate the wallet
Investigators generally need the public transaction evidence, not the secret required to spend what’s left.
⸻
🏛️ REPORT IT
For U.S. victims, the FBI explicitly asks cryptocurrency victims to report:
wallet addresses
amounts
type of cryptocurrency
transaction hashes / TXIDs
dates and times
plus other identifying information surrounding the incident.
FBI Internet Crime Complaint Center
Victims can also contact their local FBI field office.
Don’t assume:
“Bitcoin transactions can’t be reversed, therefore reporting is pointless.”
Blockchain transactions themselves aren’t reversed, but tracing can identify centralized touchpoints, and exchanges can freeze assets through their own processes or legal process.
⸻
🚨 THE SECOND ROBBERY
This needs to become one of your loudest warnings.
After somebody posts:
“I lost 8 BTC.”
They may receive:
“I can recover it.”
“I work with law enforcement.”
“I traced the hacker.”
“Send me $5,000 and I’ll freeze the wallet.”
“Connect your wallet here.”
“Send your seed so I can verify whether you’re affected.”
NO.
The FBI specifically warns that cryptocurrency recovery scammers target people who already lost crypto, including fake recovery companies and people pretending to have law-enforcement connections.
Private recovery companies cannot issue seizure orders.
Law enforcement doesn’t charge victims an investigation fee.
And the current IC3 homepage itself warns that scammers are impersonating IC3.
Nobody legitimately helping investigate this needs your seed phrase.
⸻
🧡 HOW ORDINARY PLEBS CAN HELP
This may be the most important part of the entire project.
You do not need to become a hacker.
Help find Coldcard owners who aren’t terminally online.
The person most endangered now may not be arguing about entropy on Nostr.
It may be someone who:
bought a Coldcard in 2022
generated a seed
put bitcoin on it
put the Coldcard in a drawer
and thinks:
“Hardware wallet. I’m good.”
Those people need the warning.
Plebs can also:
Archive public evidence.
Preserve URLs, dates, screenshots and original context.
Help victims organize evidence.
Don’t ask for their seeds.
Point victims toward official reporting.
Don’t promise recovery.
Translate the technical explanation into ordinary language.
Weak randomness made some seeds guessable.
That’s enough for most people.
Watch for recovery scammers.
They’re going to hunt the victim population.
Don’t dox victims.
Knowing that somebody owns substantial Bitcoin is itself dangerous information.
Don’t accuse named people without evidence.
Developers, executives, researchers, auditors, promoters and funders have different roles.
Relationship ≠ responsibility.
Responsibility ≠ criminality.
Criminality requires evidence.
⸻
⚖️ AND KEEP OUR INVESTIGATION SEPARATE
Our unanswered historical questions remain:
⬜ Who reviewed the March 2021 RNG integration?
⬜ What exactly happened during release security review?
⬜ What was the unrelated 4.0.0 security issue?
⬜ What did the reported 2022 review say about RNG?
⬜ What exactly did James O’Beirne communicate in May 2025?
⬜ Who received it?
⬜ What was the scope of the reported June 2026 AI review?
⬜ Were earlier unexplained wallet drains actually connected?
⬜ Who were the attackers?
⬜ Can centralized touchpoints eventually identify or freeze attacker funds?
Those questions matter.
But they come after protecting people who may still be exposed.
⸻
🧾 THE ENTIRE THING IN SIX WORDS
CHECK → MIGRATE → PRESERVE → REPORT → WARN → VERIFY
And underneath that:
Don’t trust a DM. Don’t share seeds. Don’t accuse without receipts.
‼️🚨🔔 The Sandbox SAND bridge was exploited today. An attacker abused compromised cross-chain bridge permissions to mint unbacked SAND on Base and BNB Smart Chain. The bridge paths were disabled and exchanges including Upbit and Bithumb halted SAND deposits/withdrawals.
And BitBox disclosed/fixed three hardware-wallet vulnerabilities this week in firmware 9.26.5. Importantly, the manufacturer says these are not another Coldcard situation: there are currently no reported stolen funds and existing BitBox seeds are reportedly unaffected.
So something bigger is becoming visible:
Coldcard wasn’t necessarily an isolated “crypto had a bad day” event.
2026 is producing a cluster of:
old dependencies + complex integrations + hardware-wallet/security infrastructure + cheap AI-assisted vulnerability discovery + attackers and defenders searching the same code simultaneously.
And today’s reporting makes that race explicit.
Which means I would add another track to our investigation:
DISCOVERY RACE
🟢 Defensive researchers find vulnerability
↓
Patch privately
↓
Users update
versus
🔴 Attacker/AI finds vulnerability
↓
Exploit developed
↓
Money moves
↓
Defenders discover it afterward
Coldcard demonstrated how catastrophic the second path can become. TRM currently describes it as the largest hardware-wallet exploit of 2026 and is still monitoring the stolen funds.
