News for hackers. Stuff that matters.

jgrahamc

Re:We halved DNS build delay and cut memory use by 85% on Tuesday November 03, 2026 @10:43AM (#50353404)
There's a lot of native code at Cloudflare. Just go look up the optimisations we did around crypto or compression in assembly. Or every time we use a compiled language like Go or Rust or C++.

I'd also say that when you look at JITed language like Lua and JavaScript the performance is pretty damn impressive. As with so many things it depends what you're doing.

Re:We halved DNS build delay and cut memory use by 85% on Tuesday November 03, 2026 @10:42AM (#50353389)
I decided to step down as CTO because I'd been at it for a long time and the board role allowed me to stay involved without being there every single day. I wrote about my "retirement" here: https://blog.cloudflare.com/three-chapters-at-cloudflare-pro...
Re:We halved DNS build delay and cut memory use by 85% on Monday November 02, 2026 @12:57PM (#50340772)
Cloudflare always had a pretty strong culture around optimisation. If you go back and read the blog there are a lot of blogs about reducing CPU, latency, RAM, etc. through software. It's one of the key features of how Cloudflare has been successful and an important part of succeeding in today's world.

Source: I was Cloudflare's CTO.

Re:MacBook Pro with PROLED XDR Display and Touch on Wednesday October 28, 2026 @02:32PM (#50278725)
I use the touch screen on my Panasonic Toughbook: https://blog.jgc.org/2025/07/ode-to-anti-mac-panasonic-tough...
Re:MacBook Pro with PROLED XDR Display and Touch on Wednesday October 28, 2026 @01:21PM (#50277181)
I use Mac mostly, but I also use a Panasonic Toughbook (https://blog.jgc.org/2025/07/ode-to-anti-mac-panasonic-tough...) and it has a touch screen display and I find it's a useful addition to how I interact with it.
Re:Sigsaly: Churchill's top secret hotline to the President on Sunday October 18, 2026 @01:33PM (#50143287)
Also, worth reading about is Turing's Delilah speech encryption system: https://spectrum.ieee.org/alan-turings-delilah
Seems fair; I won't speak to people wearing glasses with cameras in them.
Re:Casio F-91W using Statecharts on Friday October 16, 2026 @04:11AM (#50115570)
Re:Submarine Cable Resilience on Wednesday October 14, 2026 @03:01PM (#50093132)
Everyone in the industry I can find including people who do recycling of submarine cables. It seems that cross sections are sometimes given away as gifts as part of a project but are really hard to find.
Re:Submarine Cable Resilience on Wednesday October 14, 2026 @01:49PM (#50091854)
I have been trying, and failing, to buy a cross section of a submarine cable. I'd love one as an art piece for my house.
Re:Passkeys Are Not a Trap on Wednesday October 14, 2026 @01:15PM (#50091188)
I find the following has the right combination of mental model and security for me:

1. A password manager that generates my passwords stored on an external USB key (which is itself encrypted). I use KeepPassXC.

2. More than one Yubikey doing U2F.

I know how to backup #1 and I have more than one of #2 to handle losing a Yubikey. #2 gives me phishing protection.

Re:bel-doom on Tuesday October 06, 2026 @12:52PM (#49981074)
A little known "fun fact" about Arc and Bel. Back when I was implementing Cloudflare's WAF in 2012 I wrote an implementation that targeted Arc rather than Lua (which Cloudflare used extensively). We eventually went with the Lua version but the Arc version did work.
Re:Surely you have ultra-wideband radios on your bins too? on Sunday October 04, 2026 @09:29AM (#49953768)
This is great fun. For where I live I decided to "under engineer" a solution that required 0 Turing Machines: https://blog.jgc.org/2024/09/i-made-rubbish-clock.html
What I'm struggling with is that the world seems to be carrying on as normal, either talking around the threat or ignoring it altogether.

This was pretty standard behaviour during the Cold War. Growing up in England I was well aware that a nuclear war could happen at any time and that England had absolutely no defence. And so we carried on living.

Re:The AI Race Just Got Awkward on Wednesday September 30, 2026 @01:46PM (#49912102)
Yes, but to expand on my flippant response, the opening of this post is a sign of bad writing. And I don't want to waste my time on bad writing.

It shows that the author hasn't thought about their audience and has assumed that everyone knew and used the same terminology as them. And, worse, assumed that they'd understand immediately why they were using that terminology.

The opening is If you read the news headlines these days, you would be forgiven for thinking that the Western labs are getting spawn-camped by Chinese labs en masse. If you don't know what spawn-camping is you're lost; if you do it's not obvious what that means in this context. Good writing brings the reader along with the writer.

It would have been clearer if they'd written: If you read the news headlines these days, you would be forgiven for thinking that Western AI labs are being outplayed by Chinese AI labs using something similar to the gamer technique of "spawn-camping". The Chinese appear to be waiting for each Western release and then instantly distilling it. A little like gamers waiting for their opponents to reappear (from the dead) at their camp and then kill them off immediately.

This is better because a reader unfamiliar with the idea of spawn-camping learns something and it explains the metaphor. But I could be wrong in my interpretation of why they are using spawn-camping since they fail to explain it.

Re:The AI Race Just Got Awkward on Wednesday September 30, 2026 @12:27PM (#49911013)
I dunno, man, I first got on the Internet in 1986 and was Cloudflare's CTO for years. I've been online quite a bit.
Re:The AI Race Just Got Awkward on Wednesday September 30, 2026 @12:16PM (#49910854)
The first sentence is: "If you read the news headlines these days, you would be forgiven for thinking that the Western labs are getting spawn-camped by Chinese labs en masse."

Since I have no idea what "spawn-camped" means I gave up reading the rest.

Re:GPT-6 Astra Solves a WWI German Radio Cipher on Saturday September 19, 2026 @06:28AM (#49765207)
Oh yes, certainly, I am not blown away by what was done here. The AI had access to the cipher type and a list of keys.
Re:GPT-6 Astra Solves a WWI German Radio Cipher on Saturday September 19, 2026 @06:22AM (#49765170)
It is not a simple substitution using the polybius square with pairs of letters from ADFGVX mapping to 25 letters of the alphabet.

The first step is to take each letter and turn it into a pair of letters from ADFGVX but the second step is then a keyed columnar transposition.

Thanks for reposting that. Re-reading my own words made me well up a bit at the memories of that time and working with him. And the final technical meeting I had with him where I couldn't understand why he was being oddly stubborn about an API definition. Looking back it was all part of his illness.
Re:Why getting your hands dirty is good for you on Tuesday September 08, 2026 @11:07AM (#49611442)
I parodied this sort of thinking via my alter-ego, Brad Bradstone [1], in a post called "My Treadwater Desk" (https://web.archive.org/web/20190802085907/http://doublestea...).

--- My Treadwater Desk

If you're the CEO of company, like me, you do a lot of sitting and talking. And all the evidence points to sitting being a disaster health-wise. As a father it's my responsibility to take care of my health over the long term so that my kids, Dagwood and Spaniel, have their father around for a long time. Sure, they don't see much of me because I'm building Yellow Yellow, but I want to be there when we can all enjoy the fruits of my labors.

So, I looked into the current fads running through Silicon Valley. NoSitting has displaced NoSQL as the current buzzword and there are two current solutions: standing up and walking.

At Yellow Yellow we like to blaze our own trail and in NoSitting we're no exception. But rather than standing up all day (which kills your legs and leads to varicose veins---a major problem if you plan on spending your retirement years in the Caribbean on the beach) or walking on a treadmill (which causes wear and tear on your ankles, knees and hips) I've come up with a far better solution.

The Treadwater Desk.

In our basement we had a disused well from back before our brownstone had municipal water. It didn't take much to fill the well with water and a bit of carpentry later and I had my laptop and an Apple Magic Mouse (wrapped in Saran wrap).

Now my day starts by changing into a suit (in this case a wet suit) and plunging into the chilly waters of my treadwater desk. That blast of icy water is better than any espresso from the Daily Grind and really gets me in the mood for work. Also, the vasoconstriction of most of my body means more blood for my brain to keep Yellow Yellow running.

Then it's treading water while Skyping with my team. I've found that using Siri I can dictate emails without a problem and the Magic Mouse is working well.

And let's face it, we all came from the sea originally and so treading water is natural for us all. It's also aerobic and far easier on the joints than walking. It's also a full body workout. Rather than just working on my legs I get every muscle moving. And the health benefits of that are well known.

It's also a metaphor for what it's like to be a startup CEO: I'm literally keeping my head above water all day long and if I don't keep making an effort I (and Yellow Yellow) would go under.

Finally, I'm planning to add Epsom salts to the water to increase buoyancy and help keep my skin in tip top condition.

[1] The inventor of double stealth mode (the public doesn't know what we're building and neither do we). https://news.ycombinator.com/item?id=4166183

Yeah. I own an LG TV also of about the same vintage. It doesn't get to talk to the Internet.
Re:Intellectual Fly Is Open on Sunday September 06, 2026 @09:39AM (#49586450)
My worry is that if people can't express clearly, in writing, what it is they want to say then an LLM won't help them much. I've seen LLMs get the wrong end of the stick over and over and produce a huge chunk of text that's certainly readable and says something. It's just not clear that the something is the thing the person originally really wanted to say.

If you can write clearly enough to express your idea to an LLM, then perhaps you should just send that to the person you're writing to.

Re:Intellectual Fly Is Open on Sunday September 06, 2026 @09:22AM (#49586312)
I take a different stance on your final point. I think people should absolutely not use AI tools for functional communication as a crutch because they struggle with writing clearly.

They need to learn to write clearly. I believe there's a direct connection between clear thought and clear words.

Re:Intellectual Fly Is Open on Sunday September 06, 2026 @08:55AM (#49586125)
I think the most important line in this post is: But LLMs are also lousy writers and (most importantly!) they are not you.

When I was editing the Cloudflare blog I imposed very little in terms of style so that the style of each individual writer could come through. It was almost as important as the actual content that the reader could feel that an actual individual wrote the text (with all their personal quirks intact).

One of the "evils" of premature optimization is how much time you spend on the optimization vs. the benefit you get from it. If your goal is correctness and shipping fast and you're not memory constrained then spending time using the least amount of memory is a waste of time specifically because you want to ship fast.

Another interesting thing that happens is you don't necessarily know what form your actual optimizations will need to take. Later when your systems grow you discover the suboptimal parts you hadn't optimized for.

Very early on at Cloudflare I worked on part of the DNS infrastructure that took DNS records from the UI and got them in a state for actual authoritative serving. The system had been constructed anticipating Cloudflare having millions of customers with unique domains, but it had not been constructed for a single customer with a single domain with millions of records. This caused a periodic slow down in DNS record updating while the system churned on that one customer.

In a different job I worked on a piece of optimization software that needed to keep track of "node" A is reachable from node "B". This had been implemented as a matrix (literally a malloced NxN matrix of ints storing 0 or 1) which worked really well for small systems. But you'd be out of memory really fast on a large project. I replaced the matrix with a hash table and all was good because the matrix was actually really sparse.

Older »