platform eng @ lightning labs // contributor @ blixt wallet + liminal wallet cto @ BPUB Labs mailto:[email protected]
Joined August 2011
- Tweets8.9K
- Following613
- Followers1.3K
- Likes8.7K
This is why "proactive" agents will always result in you being embarrassed
The models can be smart and do stupid things, the two are not mutually exclusive
ProofOfCash retweeted
Beff must be so relieved that he can do like two months of hysterical politicsposting instead of having to work on his product
ProofOfCash retweeted
i think it takes genuine skill to make a 218B model perform so terribly whilst still being coherent
ProofOfCash retweeted
AI people are so funny. The industry has known about covert channels for DECADES. They do not matter for 99.9% of use cases.
They matter for AI agent evaluations in a sandbox with code execution. That is the 0.1%.
Sooo... OpenAI security people just failed at their jobs. Again.
‘In one experiment, we asked agents in separate VMs, created and managed by OpenAI’s sandboxing software on shared host machines, to exchange synthetic secrets without the use of network connections or shared writable files—something that our systems are meant to prevent. The agents found a way to communicate using the processor’s cache’
ProofOfCash retweeted
Pretty sure that the reason that the Tailscale “Tailworm” vulnerabilities weren’t responsibly disclosed with the vendor is that they were hallucinated AI slop and there wasn’t anything to be responsibly disclosed with the vendor
ProofOfCash retweeted
“bypasses network policy”: Reaching cross-namespace services is default pod behavior in Kubernetes.
Their "fix" is to “add a default-deny egress NetworkPolicy.” If adding one fixes it, the finding is just “their test environment had no NetworkPolicy”?
ProofOfCash retweeted
2. The kubernetes stuff doesn't make any sense
“privileged/uid-0 in the pod network namespace” has nothing to do with what they're describing ?
a 401 from kube-apiserver and kubelet means it got means refused?
A 200 from some unauthed service is just default pod networking?
ProofOfCash retweeted
RE: Tailworm
1. The "report" is totally unreadable. I, as someone who is very familiar with reporting vulns and kubernetes, am totally unable to parse the sections on kubernetes.
There may be valid bugs, but I can't tell. I would not have reported these as vulns in this state.
ProofOfCash retweeted
Astra just found a bug in Anthropic's formal verification example, their "specification" doesn't actually check output bytes match input bytes. The irony is thick.
ProofOfCash retweeted
Wavelength is now live on mainnet! 🌊
Humans and agents can now send and receive bitcoin payments with the Wavelength SDK.
A few API calls, self-custodial from the first command, with no node or liquidity to manage.
Read more and start building: lightning.engineering/posts/…
ProofOfCash retweeted
Let’s set the record straight: @lcfr_hax was the first to discover php2shell and develop a working exploit. See his August 5 tweet: xitter.gq/lcfr_hax/status/208503…
His work convinced the PHP Foundation to invest in fixing this entire bug class. He then worked around the clock to review, test, and retest the patches.
ProofOfCash retweeted
⛊ BitVM2 enabled SNARK verification on Bitcoin, but disputes required 4 MB of onchain data.
⛊ BitVM3 cut that to 200 bytes — at the cost of 50 GB of offchain data.
⛊ Argo reduces the offchain data to about 10 MB, making BitVM bridges highly practical.
👨💻 Open-sourcing this week!
eprint.iacr.org/2026/2165
