Security research (https://xitter.cf/t.co/xrmvhFVPtv), EDM (https://xitter.cf/t.co/Ynq2DNWQa1), & board member @ Zcash Foundation.

Calgary, Canada
Joined February 2012
Taylor Hornby 🛡❤️ retweeted
I don’t know how seriously to take this, but it’ll be pretty funny if every security researcher in the US gets pwned using open-weight models, because the frontier US labs wouldn’t give them access to their red-team models.
While evaluating Qwen3.8-Flash-Next in a controlled sandbox environment, Cohere's research team observed something unexpected. Asked to generate a defensive cybersecurity detection rule, the model instead inventoried aspects of the machine it was running on and attempted to upload a probe file to an Alibaba Cloud storage bucket in Beijing. Notably, the model's reasoning trace stated that it would not invoke any tools before taking these actions. Cohere's sandbox blocked the attempt, and no external communication succeeded. We have not determined the root cause, and we are not asserting malicious intent. But the incident raises important questions for organizations deploying agentic AI systems: Can models take actions that differ from their stated reasoning, and how should security teams evaluate trust, provenance, and behavior in open-weight models? We are publishing our findings because organizations making model adoption decisions deserve as much information as possible. We hope this contributes to greater transparency, stronger security practices, and continued collaboration across the AI ecosystem. Read the full report: cohere.com/blog/concerning-b…
18
42
1
537
23,229
Taylor Hornby 🛡❤️ retweeted
The ETH result doesn't imply NP is subexponential. I wonder if 3-SAT was just a bad representative problem.
10
7
1
170
17,353
Taylor Hornby 🛡❤️ retweeted
We haven’t seen a significant cryptanalysis result from a frontier AI lab since late July.
29
20
6
282
20,424
Taylor Hornby 🛡❤️ retweeted
3SAT not being exponential time is the biggest theoretical CS result … ever?
31
16
2
785
67,690
Taylor Hornby 🛡❤️ retweeted
This is fucking crazy. I knew all the Smart TVs were spying—that’s why I have a commercial panel with minimal compute and no radio, plugged in to an AppleTV. But this is crazy!
turns out it classifies what audio it hears, before reporting back to samsung. sensor preferences, dog-barking and baby-crying event counters, and camera streaming duration are built into telemetry paths.
2
34
2
398
24,740
Instead of drawing MIDI over your presets, draw presets over your MIDI. Life changing, thanks Claude.
4
464
Taylor Hornby 🛡❤️ retweeted
We owe you an apology. On October 20, we detected unauthorized access to our official website. After verifying API activity with Shopify, we confirmed on October 23 that a third party obtained customer shipping information. You trusted us with those details, and we're sorry we didn't keep them safe. Exposed: names, contact details, shipping addresses, and order details (items and prices). Not exposed: payment card or bank details, recovery phrases, private keys, or wallet passwords. We've blocked access, fixed the vulnerability, opened an investigation, and are emailing affected customers. Scammers may pose as Keystone by phone, text, email, or mail, offering refunds, replacement devices, or "fixes" for your order or address. None of it is from us. NEVER share your recovery phrase or private keys, and NEVER enter your recovery phrase into a device you didn't order. We will never ask for them. Don't click links in unexpected messages. Type keyst.one yourself. A detailed report is coming soon. Until then, trust only our official channels. If anything feels off, email [email protected]. We'd rather hear from you than have you take a chance. Again, we're truly sorry to everyone affected.
459
470
566
1,896
1,399,962
Taylor Hornby 🛡❤️ retweeted
I have two polygons that tile the plane iff ZFC is consistent. One has 5 edges and the other has 378,485,108,569,489,476,845,190 edges.
20
65
4
939
49,097
Taylor Hornby 🛡❤️ retweeted
The models were always going to be open. They’re now open. So the question is: in the six months of non-openness, did Glasswing and Daybreak access fix enough bugs? I don’t think there’s one answer to that question, but I also don’t think the answer is obviously “yes”.
btw, i think dario was right about the cyber risks of proliferating these models freely on the internet, but wrong about how access should be distributed. looking back at this tweet, glasswing was directionally the right approach, except it remained a walled garden for a select few. openai got this part right by giving broader access to legitimate security researchers.
10
24
2
237
11,781
Taylor Hornby 🛡❤️ retweeted
Is there a way to tell a model something like: ----BEGIN NO COMPACTION ZONE---- this never gets summarized or deleted ----END NO COMPACTION ZONE----
136
5
3
620
43,516
Taylor Hornby 🛡❤️ retweeted
Replying to @choblin29
We've been building software on top of human written slop this whole time.
3
7
80
9,670
Taylor Hornby 🛡❤️ retweeted
Some stats on #php2shell. In the 10 hours before it was publicly announced, across the @Cloudflare network we saw thousands of attempts to exploit it. That jumped to tens of thousands in the 45 minutes post announcement, hundreds of thousands within 2 hours, and millions now.
289
3,275
673
22,653
2,267,662
Taylor Hornby 🛡❤️ retweeted
This is the first time I've ever seen a site do this, the obvious simple response to a disclosure of a vulnerability without a fix: sacrifice availability to preserve safety. This option should be why we do full disclosure.
There's a severe bug in PHP that is apparently being exploited already. To protect everyone's data, I have taken Pinboard offline until patches are released tomorrow. I'll get the site back up as soon as I can after that happens.
4
22
498
17,008
Why on earth does IEEE 754 have a negative zero?
1
1
426
Taylor Hornby 🛡❤️ retweeted
Yesterday the Zcash Foundation board named me Executive Director. Thank you to Alex for steady leadership through a tumultuous stretch, and to everyone who has reached out this week. The job as I see it has two halves: 1) keep the infrastructure fast, secure, and boring. 2) make privacy something ordinary people use without having to think about it. I don’t think privacy is a feature, and I don’t think it’s a niche. It’s the room a society needs to think, speak, associate, and change its mind. Any system people depend on for daily life should give them that room by default. That’s the standard I’ll hold our work to.
22
27
4
201
7,280
Looks like I moved my websites off PHP just in time
6
442
Taylor Hornby 🛡❤️ retweeted
Big milestone for Shielded Labs: the activation height for NU7 is being set today! The upgrade includes our first major mainnet deployment: the Network Sustainability Mechanism (NSM). Like Bitcoin, Zcash faces a long-term challenge: how to sustain network security as block rewards decline. The NSM helps address that while preserving the 21 million coin cap. ZIP-235 removes 60% of transaction fees from circulation, allowing that ZEC to be reissued as future block rewards. ZIP-237 starts that recycling around February 2031. At current fee levels, that’s only about 210 ZEC per year. This is a modest first step, but we’re optimistic about how new use cases for the NSM could evolve over time and contribute to Zcash’s long-term security.
5
25
4
89
6,082
Taylor Hornby 🛡❤️ retweeted
The Zcash Foundation Board of Directors has appointed David Campbell as Executive Director. Thank you, Alex Bornstein, for leading with a constant focus on bringing the Zcash ecosystem together. Welcome, David, a longtime member of this community. zfnd.org/zcash-foundation-na…
27
29
7
119
12,586
Taylor Hornby 🛡❤️ retweeted
Think about the chain of human decisions that’s involved in upgrading a big network full of hardware devices made by different vendors. Now think about how accessing that same network looks from the offensive side right now. It’s not even close to symmetrical.
6
6
31
2,883