djha-skin

I recently Googled "Are you not enteretained?" I wanted to see the movie clip. Instead I got "I am thoroughly entertained!" from the AI. Some weird blurb like that. It did have the clip and even mentioned it in the AI summary, but it was still jarring. They have since fixed it, I think, to say "You must be talking about such and such movie quote", so it feels like they are trying to improve it, but it was jarring at the time.

I mean, in terms of property damage, the Boston Tea Party was a pretty big deal. Probably in arson territory in today's dollars.

I used to think this but then improved and found that it was a matter of tooling. Using beads[1] and getting the AI to write its own tickets and wiki articles using GitHub's wiki feature on repos, it can keep track of long horizon projects pretty easily. Then you curate some of the tickets or read through them and have them update the beads or adjust them as they work them. When they want to retain some important piece of knowledge you have them add it to the GitHub repo wiki.

Modern AI programming looks an awful lot like traditional product owner or product management roles.

1: https://beads.gascity.com/

I started a Janet project [1] with the express purpose of doing everything manually for fun, but I ended up using the agent for a lot of stuff pretty quickly anyway. Functions I wanted to be part of the library as a matter of completion, but that were not fun to write. Docs as well. Stuff I probably wouldn't have written in the first place if AI didn't exist.

It reinforced in my head this idea that hobby programming is entering its "sports" or "running" era.

Consider the modern jogging enthusiast. They buy Nike shoes that make it really easy to run long distances, specialized clothing, fancy water bottles. They find spots to drive to and park at before getting out and running downhill for a few miles, getting picked up at the bottom by one of their friends. Stuff that makes for a great social media post.

They do the hard part of running, but curated to just the parts that they enjoy. They drive the rest of the time and when they do run, it looks nothing like the first marathon run of ancient Greece.

It sounds nice and all to talk about "I'm going to code anyway," but when I did this I found that there are always parts of the coding that I really really never did want to do and that I'm happy to just outsource to my agent like the jogger outsources getting to the top with their car. Testing, writing docs, making the readme logo. Stuff that probably wouldn't have happened without the AI but that I'm glad exists.

1: https://GitHub.com/djha-skin/janver

Fresh reader here. I am very interested to learn more about your sentence

> We spent a lot of money finding out that a lock is more performant.

I want to hear about that journey.

> Besides, it's a btree.

I guess it's not a hash table, but I think the point of Vincent's post is still worth exploring. You and he both say that essentially a key/value store is mutexed, and the user's code runs inside that mutex. That is fascinating. Why would that be better? Clearly it is, or you wouldn't have spent the money. What controls did you put in place to ensure user code didn't blow up the performance, or did you even feel the need for such controls? Is WASM VM execution fast enough for this? Is there even a market for that? Who pays to put their code inside that mutex and why? I want to know more.

Ok, but does it need to scale?

This feels ai generated, and doesn't get to the promised point of "what does it mean for us" forecasting in Europe or much more than a few sentences for the US.

HAMTs: https://en.wikipedia.org/wiki/Hash_array_mapped_trie

He addresses this use case in the article.

Reminds me of Rich Hickey's clojure data structures. Yes, they're technically log_32(n) complexity, but it turns out log 32 is basically flat on any normal machine, thus "practically constant".

It cannot be a better C. You cannot implement exceptions in it using set jump for example, but the biggest problem is memory management. You can't implement your own arena allocator in golang.

Chief design goals were radically easy concurrency and speed of compilation.

I suspect this "not a systems language" alludes only to OCaml's rather steeper learning curve and until-recently difficulty with multiple threads. I am sure it could roll just fine as a single-threaded compiler language written by a small team, which indeed, it was.

Can't I just use a `docker run` command to run claude and bind in the current directory into the docker container?

Punishing leadership for perceived strategy mistakes is a great way to scare good leadership away from working for you.

In the age of AI, writing things that used to take years can now be done in months or weeks if you have deep enough pockets for it.

Reimplementation is a particularly juicy target because it's easy to test. Imagine someone writing a better browser than Chrome from scratch in just a year.

Because of this moats around business due to difficulty of implementation are effectively gone.

What is gtm

TrueNAS is on FreeBSD, as well as lots of network equipment. This does affect us more than we think as operators.

New datacenter going up in Utah. 40k acres. For context the local land footprint of a Facebook datacenter is 390 acres. Projected to consume tens of gigawatts. For context the entire state last year only consumed 4 but was able to produce 10. This thing is like Manhattan Project/Apollo Project sized.

> Google Chrome silently installs a 4 GB AI model on your device without consent. At a billion-device scale the climate costs are insane.

OH MY GOODNESS, this is the WORST headline.

If Google Chrome comes with an AI model, and you install Chrome of your own free will, you just gave consent.

The "climate costs" are happening whether or not the AI is there. Sure, maybe it makes the hardware work a bit harder, but like, come on. I'm still using my computer anyway. YOU are the one costing the climate, not Google. You're the one turning the "On" button on.

I don't even know why headlines like these are taken seriously.

Fascinating. Would love to know more if you have a link.

Are you a network engineer?

Are you a network engineer?

Obfuscation. By inspecting packets coming from my network now you can tell what MAC addresses are in my network and also internal network topology. It's part of the reason your cell phone feels the need to randomize its MAC.

Everyone I've talked to with this opinion are typically mobile devs thinking about cell phones. Ipv6 works great there, but NATs are often used in corporate networks for isolation and in particular obfuscation. You can't tell what's behind a NAT by inspecting traffic coming from inside it like you can with no NAT networks. Some of the networks I administrate are contractually obligated to be so isolated.

I am reading about ipv6 nat. I guess it's possible but discouraged?

This contention point confuses me. I consistently get downvoted for this opinion, and I've seen contrarian voices online, but I have yet to meet an actual datacenter network admin who disagrees with me.

Article does not address the elephant: there is no ability to NAT with IPv6. Sure, absolutely, you shouldn't have to NAT, but in my datacenter, NAT is a feature, not a bug. The article specifically asks "did the ipv6 designers go mad" and then they list features I've never heard of or use to prove they didn't. Those features are not why I think they went mad. The inability to create a NAT is.

For this reason, at every shop I've ever worked at, the intranet is ipv4, often with ipv6 disabled, with dual stack on the load balancer for ingress traffic. Note, I do not set it up that way: it comes like that when I've arrived.

I have recently blogged that AI and Common Lisp don't mix, but I've come to the opposite conclusion lately. AI evens the playing field between large teams and single developers. Now all the lone wolves in cl will be able to do large things, like a .net implementation or a yaml parser. I heard one guy say he was using AI to write a c complete in common lisp. I wonder if AI was used here or not.

Such humans are likely not programmers, which generally fits my use case. If I wish for my blind sister to contact me, I just give her my number.

It's simple: draw your email in a paint program and export it as a png. Totally readable by humans.