Good question. I didn't cover this in the post — the binary doesn't run on the host kernel directly. It runs inside a lightweight KVM-based VM with no operating system. The shim is the only thing handling syscalls inside the guest. So strace on the host wouldn't see anything — no syscalls reach the host kernel from the guest. From the host side, the only visible activity is the hypervisor process making syscalls on behalf of the guest.
Inside the guest, there's no kernel to attach strace to — the shim IS the syscall handler. But we do have full observability: every syscall that hits the shim is logged to a trace ring buffer with the syscall number, arguments, and TSC timestamp. It's more complete than strace in some ways — you see denied calls too, with the policy verdict, and there's no observer overhead because the logging is part of the dispatch path.
So existing tools don't work, but you get something arguably better: a complete, tamper-proof record of every syscall the process attempted, including the ones that were denied before they could execute.
I'll publish a follow-on tomorrow that details how we load and execute this rewritten binary and what the VMM architecture looks like.
I guess it’s different since I mostly deal with interns. But I haven’t noticed a lot of H1s being hire at my FAANG, at least ones that are obviously F1? I knew an H1 in another group from an obviously second rate school in China and a masters from a no name in America. He was pretty successful in his career so I guess school brand doesn’t mean that much.
We aren’t an AI tech group or anything like that. I was on a Z (working) visa in China for 9 years though.
>Writing the code hasn’t been the bottle neck to developing software for a long time.
Then we're doing different things.
I didn't like GitHub so I wrote my own. 60k lines of code later... yes writing code was the bottleneck which has been eliminated. The bottleneck is now design, review, and quality assessments that can't be done trivially.
This isn't even the project I wanted to be doing, the tools that were available were holding me back so I wrote my own. It also consumes a few hours a week.
If you think writing code isn't the bottleneck then you aren't thinking big enough. If you don't WANT to think big enough, that's fine, I also do things for the joy of doing them.
Right. But it's not my favorite nerd snipe interpretation that allows me to post low effort comments on hackernews about the headline instead of engaging in a meanful discussion about the article.
Are they? This seems about deceptive or malicious content (i.e., redirecting to ads) rather than “something in my history triggers a JS redirect”. I’ve definitely experienced the latter with MS, but never the former.
In my experience building an open-source project using agent.md and skill.md, I’d say that many “agent skills” are essentially structured system prompts, but the surrounding configuration and orchestration matter more in practice.
One issue I’ve run into is that different models (e.g., Codex vs. Claude/Opus) often require slightly different configurations. This leads to duplicated files like .codex/skill.md and .claude/skill.md, which adds maintenance overhead.
For example, when working in non-English contexts, I need to ensure the AI doesn’t generate non-English inline comments. Constraints like this often have to be repeated across model-specific configs.
In my case, since enforcing English comments is a priority, I tend to put this in skill.md. Otherwise, I would place it in agent.md.
Because of this, I prefer using a single AGENTS.md as a global configuration layer. It’s simpler to manage and reduces duplication.
Maybe not "collapse" in a the sense of going to zero but if there was no profit to trading, then the quant trading industry would not exist, trading profits would collapse.
Meanwhile Two Sigma is hiring alpha quants to be AI research scientists at $250k starting salary + bonuses.
Even if we're just talking about the HFT/sell-side, there clearly exist various anomalous inefficiencies that can be exploited.
It really comes down to JavaScript. The web was fine when sites were static HTML, images, and forms with server-side rendering (allowing for forums and blogs).
In the Mahābhārata, what's going on with the dynasty tree of the Kurus?
That's a view you get in every single book, and it looks really weird here. I feel like it's important to get this really basic stuff right before doing the cool-looking graph visuals.
First of, Bevy isn't just an ECS implementation. It's a game engine designed around ECS. While the ECS part is its core, Bevy also provides rendering, resource management, physics, etc. One of its other tasks, a rather important one too, is system scheduling: once you register systems, Bevy is responsible for dispatching them each frame, at the appropriate time, while also avoid data races over global resources and components.
You asked about Google and what impressed me so much, going through this exercise, while not exactly helpful for me and my work directly (I'm doing similar things but completely in the Azure ecosystem), it is definitely a great display of how agents are more than just an 'LLM' that everyone here seems to think is equivalent to AI.
It's seriously the opposite feeling of imposter syndrome at this point, I'm in my 30's, a senior data engineer myself at a F200 company; I can't believe so many of my peers are so behind and ignorant of what is going on, confident enough to makes publicly lasting comments about how 'unreliable', 'bad', 'slop'; 'AI will never this or that'.
> A minuscule minority overrules the vast majority.
This is not a game of numbers, numbers would have some weight if decisions were affecting everyone equally, but the EU isn't homogeneous, foreign policy issues affect different countries differently depending on geography, culture and history e.g. Central Europe is much less exposed to adverse events, dependencies and risks than the periphery.
As I said before, now it's not the time for coercive changes, wait until integration takes its course and makes the political environment approximately the same for all members, the EU is far from that now.
> but it was the only way to convince countries to form the EU.
Then don't alter the deal and ask everyone to pray for not altering it any further.
> There is no alternative to supporting Ukraine as much as possible.
Depends on what each of the EU members considers possible and what risky - forcing some countries to go against their economic and political security will most likely lead to re-partitioning of Europe and another age of European wars.
It's quite disturbing to observe the continuous lack of an honest conversation about the political realities in Europe, which is precisely how the leading countries of the EU blew Ukraine / Russia... removing veto power only reduces the incentives for that to ever happen.
Cool, now maybe let's do something about all the shit I have to clear out out my face before I can read a simple web page. For example, on this very article I had to click "No thanks" for cookies and then "No thanks" for a survey or something. And then there was an ad at the top for some app that I also closed.
It's like walking into some room and having to swat away a bunch of cobwebs before doing whatever it is you want to do (read some text, basically).
There's approximately a zillion devices with 7 segment displays, and half a zillion cameras that you can point at things. It's non-invasive and simple to understand. "Just disassemble it and solder in some extra wires" wouldn't fly for monitoring a thermostat in a basement or something, but a camera + rasperry pi might solve the problem at the right price + reliability intersection.
Back in the olden days of hardware RSA tokens, there were stories of people bridging out the TOTP values via setups like this. I've got an amplifier in the garage hooked up to a smart outlet that'll turn it on, but I have to go over there manually and rotate the knob from 20db to 50db.
I've been tempted to hack up some sort of "RC-Car Wheel + WebCam" to be able to remotely control/nudge the volume, but luckily my lack of free time prevents me from going off the rails like that. My other option is some sort of arduino and hacking at the rotary encoder circuit directly from the inside, but again (luckily) my lack of free time has kept the box closed.
The world's bigger than what you've seen so far, I'm guessing!
On the other side, why should one crazed/corrupt judge in some state which has nothing to do with me be able to infringe on my freedoms and make my life worse? Worse, why is it possible to jurisdiction shop for the single bad actor and impose your will on the entire country?
You're not wrong, but (like most issues in a 350M-person country) it's complicated. The system is tailored to some expected level/type of corruption and bad actors. If you expect that the government is basically fine and that out of 50M people per region surely somebody will file suit if the issue is important then the current system makes a lot of sense. You get judges with more knowledge and awareness of your local issues, anything important still gets addressed, and you're resilient to some degree of random bad judges and bad actors. If those expectations are out of whack then you get worse outcomes.
In reality, the world is complicated enough that even boiling down the lists of judges and whatnot to that simple of a description is misleading at best. Neither solution is anywhere near optimal by itself. So...what next?
> In contrast to studies cited in recent decisions to end CWF in Utah, Florida, and elsewhere, we find no evidence that CWF is negatively associated with adolescent IQ or adult cognitive functioning.
As if we didn't know.
Not being snarky at the scientists, but the people who passed insane, invasive laws against fluoridation.
> when the calculator came about, being forced to compute in your head wasn't an advantage.
I'm not sure, whether that is true, because when educators want you to learn how to compute you are "locked out" of calculators. You don't get to use a calculator until after you learned basic arithmetic and you won't use a CAS when you are supposed to learn calculus.
Inside the guest, there's no kernel to attach strace to — the shim IS the syscall handler. But we do have full observability: every syscall that hits the shim is logged to a trace ring buffer with the syscall number, arguments, and TSC timestamp. It's more complete than strace in some ways — you see denied calls too, with the policy verdict, and there's no observer overhead because the logging is part of the dispatch path.
So existing tools don't work, but you get something arguably better: a complete, tamper-proof record of every syscall the process attempted, including the ones that were denied before they could execute. I'll publish a follow-on tomorrow that details how we load and execute this rewritten binary and what the VMM architecture looks like.