In 2023, ssh added keystroke timing obfuscation. The idea is that the speed at which you type different letters betrays some information about which letters you’re typing. So ssh sends lots of “chaff” packets along with your keystrokes to make it hard for an attacker to determine when you’re actually entering keys.
[...]
Keystroke obfuscation can be disabled client-side. After reverting my original breaking change, I tried updating my test harness to pass ObscureKeystrokeTiming=no when starting up ssh sessions.
[...]
The “chaff” messages that ssh uses to obscure keystrokes are SSH2_MSG_PING messages. And they’re sent to servers that advertise the availability of the ping@openssh.com extension. What if we just…don’t advertise ping@openssh.com?
[...]
Obviously forking go’s crypto library is a little scary, and I’m gonna have to do some thinking about how to maintain my little patch in a safe way.
These are the sorts of coding assistant experiences it's difficult to articulate without context, and also the sort that seem fairly small on their own, but add up to reframing software...
I’ve been thinking about whether LLMs remove parts of the problem-solving process that I enjoy. But I’ve gotta say, debugging this problem using Claude Code was super fun.
I am familiar enough with tcpdump, tshark, and friends to know what they can do. But I don’t use them regularly enough to be fast with them. Being able to tell an agent “here’s a weird pcap - tell me what’s going on” was really lovely. And by watching commands as the agent ran them I was able to keep my mental model of the problem up to date.
These are the sorts of coding assistant experiences it's difficult to articulate without context, and also the sort that seem fairly small on their own, but add up to reframing software engineering as a discipline.
I don't know what the percentage is, but at least many developers started doing it because building things is really fucking fun. That and the remarkable truth that you can build something in the digital world that makes something better in the actual world.
But finding bugs is only fun when you can manage a pretty specific mindset. It's great when you have the time and bandwidth to enjoy the journey. But otherwise the best you can say about it is that finding bugs is reliably satisfying and the degree to which it's satisfying is directly proportionate to how many hours you spent figuring it out. You appreciate the outcome more because it hurts a little, which is great, but different than fun.
However finding bugs with the help of a SOTA coding assistant is really fucking fun. Sometimes. They also introduce new kinds of pain.
In the context of coding, it seems like the headlines focus on whether or not AI can one shot things. And that makes sense, it's what the market wants. You can't really blow the lid off of efficiency (read: replace humans) until you can one shot things. But, for right now anyway, these tools real value proposition is removing some of the high friction parts of the process by being effective assistants. Except that doesn't quite cover it because they also open up avenues that were technically always there, but almost never used because they were high friction. That part is really fucking fun too.
From the article:
[...]
[...]
[...]
These are the sorts of coding assistant experiences it's difficult to articulate without context, and also the sort that seem fairly small on their own, but add up to reframing software engineering as a discipline.
I don't know what the percentage is, but at least many developers started doing it because building things is really fucking fun. That and the remarkable truth that you can build something in the digital world that makes something better in the actual world.
But finding bugs is only fun when you can manage a pretty specific mindset. It's great when you have the time and bandwidth to enjoy the journey. But otherwise the best you can say about it is that finding bugs is reliably satisfying and the degree to which it's satisfying is directly proportionate to how many hours you spent figuring it out. You appreciate the outcome more because it hurts a little, which is great, but different than fun.
However finding bugs with the help of a SOTA coding assistant is really fucking fun. Sometimes. They also introduce new kinds of pain.
In the context of coding, it seems like the headlines focus on whether or not AI can one shot things. And that makes sense, it's what the market wants. You can't really blow the lid off of efficiency (read: replace humans) until you can one shot things. But, for right now anyway, these tools real value proposition is removing some of the high friction parts of the process by being effective assistants. Except that doesn't quite cover it because they also open up avenues that were technically always there, but almost never used because they were high friction. That part is really fucking fun too.