swisniewski

5 ms·

1.
by swisniewski 4mo ago
The real problem is that Go produces interface implementations dynamically. The determination wether type T implements interface I is made at runtime. So is generation of the necessary vtables to produce the interface implementation. So you
2.
by swisniewski 4mo ago
But humans do make mistakes like that (driving into oncoming traffic or driving down a light rail track). For example, here’s a case where a human did it to avoid an ambulance: https://www.click2houston.com/news/2012&#x
3.
by swisniewski 5mo ago
Let’s assume the AI does out perform the DR. I still want humans in the loop, interpreting the LLMs findings and providing a sanity check. You can’t hold an LLM accountable. That’s the min responsible bar for LLM authored code, which normal
4.
by swisniewski 5mo ago
How big is the Wire Guard user base on Windows? How often do they ship new versions? My understanding is that: 1. Windows drivers are Attested by Microsoft 2. Windows collects driver telemetry Which means a really good question to ask is: W
5.
by swisniewski 6mo ago
You can use BGP hijacks to spoof another website. You just need to get a publicly trusted CA to mint a certificate for your new site. This can be done, for example, with let’s encrypt, using several of the various domain verification challe
6.
by swisniewski 6mo ago
There is no right way to do Spring Boot.The entire idea is broken. Dependency injection is good. It makes it possible to test stuff. Automagic wiring of dependencies based on annotations is bad and horrible. If you want to do dependency in
7.
by swisniewski 6mo ago
To be honest, I’m not surprised that GitHub has been having issues. If you have ever operated GitHub Enterprise Server, it’s a nightmare. It doesn’t support active-active. It only supports passive standbys. Minor version upgrades can’t be d
8.
by swisniewski 6mo ago
I use a DGX spark, with Cosmic as my DE, and it's super awesome. This is a bit of a franekin-distro, as it's ubuntu + nvdia packgages + system 76 packages, but it works pretty well. I've been using Flatpack chromium, which is
9.
by swisniewski 7mo ago
The premise seems flawed. From the paper: “we find that the LLM adheres to the legally correct outcome significantly more often than human judges” That presupposes that a “legally correct” outcome exists The Common Law, which is the foundat
10.
by swisniewski 7mo ago
Some of this is people trying to predict the future. And it’s not unreasonable to assume it’s going there. That being said, the models are not there yet. If you care about quality, you still need humans in the loop. Even when given high qua
11.
by swisniewski 8mo ago
This is not satire. If you have a large dependency graph, you are going to have a lot of vulnerable stuff. Letting one computer send you patches and the other computer merge it for you when all your tests pass is a good thing.
12.
by swisniewski 8mo ago
Take a look at pr-bot: https://github.com/marqeta/pr-bot The answer to dependabot, or snyk prs is to automatically merge them once all the status checks pass. This free your devs from having to worry about patching. PR
13.
by swisniewski 8mo ago
The article has the headline "AI Misses Nearly One-Third of Breast Cancers, Study Finds". It also has the following quotes: 1. "The results were striking: 127 cancers, 30.7% of all cases, were missed by the AI system"
14.
by swisniewski 8mo ago
Huh? I was commenting that there were no controls and the doctors were given skewed data, so any conclusions of ai ability vs Dr ability seem misplaced. Which seems to be what you just said… so I am confused about what I said that was inac
15.
by swisniewski 8mo ago
The description from the summaries sound very flawed. 1. They only tested 2 Radiologists. And they compared it to one model. Thus the results don’t say anything about how Radiologists in general perform against AI in general. The most gener
16.
by swisniewski 9mo ago
look up the parent tree... There was this statement: > From a Hacker News perspective, I wonder what this means for engineers working on HBO Max. Netflix says they’re keeping the company separate but surely you’d be looking to move them
17.
by swisniewski 9mo ago
I use Cosmic on a DGX Spark, as my daily driver, and it works pretty well. They don’t have a pop os iso for arm64, but they do have arm64 Debian repo. So I just took DGX os (what Nvidia ships on the device), added the pos os “releases” repo
18.
by swisniewski 10mo ago
Generally with large acquisitions, product integration tends to precede infrastructure integration by years to decades. Look at GitHub as an example, they were acquired in 2018, and are just migrating to Azure now after 7 years. Microsoft s
19.
by swisniewski 10mo ago
Not sure why this got downvoted. The technique could be implemented without conditionals, but not in python, and not using iterators. You could do it in C, and use & and ~ to make the cyclic counters work. But, like I mentioned, the cod
20.
by swisniewski 10mo ago
Sigh… Saying the code doesn’t have conditions or booleans is only true if you completely ignore how the functions being called are being implemented. Cycle involves conditionals, zip involves conditionals, range involves conditionals, array
21.
by swisniewski 10mo ago
Another interesting thing… random data has a high likely hood of disassembling into random instructions, but there’s a low probability that such instructions (particularly sequences of such instructions) are valid semantically. For example,
22.
by swisniewski 1y ago
There's a much simpler way to do this: If you want your library to operate on bytes, then rather than taking in an io.Reader and trying to figure out how to get bytes out of it the most efficient way, why not just have the library take
23.
by swisniewski 1y ago
Has anyone else been able to get "secrets" to work? They seem to be injected fine in the "environment setup" but don't seem to be injected when running tasks against the enviornment. This consistently repros even if
24.
by swisniewski 1y ago
I see this a lot ("if you are a startup, just ship a monolith"). I think this is the wrong way to frame it. The advice should be "just do the scrappy thing". This distinction is important. Sometimes, creating a separate
25.
by swisniewski 1y ago
The article is bullshit. AWS has a pretty simple model: when you split things into multiple accounts those accounts are 100% separate from each other (+/- provisioning capabilities from the root account). The only way cross account stu
26.
by swisniewski 1y ago
The simplest approach is to create an account and install our agent. That provides actionable insights out of the box. Our current poc supports datadog and cloudwatch, but in principle we can use just about any telemetry provider. If you dr
27.
Debugging Sucks(debugging.sucks)
4 points by swisniewski 1y ago|2 comments
28.
by swisniewski 1y ago
Digital content is not “published” in the same way as traditional content. Digital content is published by placing data on a computer, connecting that computer to the intent, then running software on that computer that allows software on ot
29.
by swisniewski 2y ago
I usually adopt a policy with my teams: You are not allowed to complain about style in code reviews. If it’s important enough for your to comment about, it’s important enough for you to add a linter to the build and block the build if the s
30.
by swisniewski 2y ago
I think you may be making assumption: that the feedback you are trying to give is clear, articulate, and constructive. When speaking to someone significantly more senior than you it’s entire possible that your feedback may not be. And that

More ›