devmor

And most people running containers are running a default docker/containerd setup, which does not actually isolate you from anything at all in the first place anyways.

For real.

My agentic workloads happen on two physical machines dedicated only to agentic workloads that share a VLAN with only eachother, with single token access to the specific repositories they are meant to push to and limited scope, hard limited credentials pulled from the hardware store on hardware token/fingerprint request only.

I view the safety around unmonitored agentic work in a similar way to how I view working with viruses or bacteria in a disease lab.

This stuff isn’t magic, it isn’t intelligent, but it is dangerous and needs to be handled with careful intent.

To be four, you could open it to the world.

This is too reductive of a statement to have any meaning. You have re-defined a class of program to encompass all programs. The term “harness” exists to differentiate it from programs that are not harnesses.

I used to be excited about these, but after a year or so of having to drive near them in Midtown Atlanta I would get incredible dopamine out of watching them crushed for scrap.

They cause traffic obstructions during rush hour every single day. Blocking entire lanes of traffic to pick up Uber customers.

Do you think it takes an hour to clean and resupply a restroom? It takes about 15 minutes for the maintenance staff in my office building.

> that being said of course I think it’s going to be way less than $5 an hour if you have a robot do it.

This is about an LLM-powered robot, not just a robot. At standard API prices, it'd run over $5 before it finished opening the door.

> cleaning out a restroom and resupplying it that’s easy as hell.

But is it cheaper than paying someone around $5 in labor to do it? Because until it is, it's not economical to have a robot do it.

Even the most powerful frontier models on SOTA hardware are not fast enough for starcraft, which generally involves multiple informed decisions per second at peak moments.

You can train a specialized model for this, but a frontier model is ill-fitting.

htop and similar TUIs have a refresh rate of 1 frame per second

I don’t know what your “again” is about - you haven’t been speaking to me here, but no, it’s quite simple and every major court case in the US has ruled as I described based on existing law.

It is not more complex in “many cases” at all.

I think you’re missing part of the conversation here. We are not discussing Starlink’s capability to deliver internet signals.

Well, stdout is not typically implemented for the purpose of simulating graphics rendering via character manipulation.

The only source I have ever seen is the US SDA confirming they have seen a demonstration of Starlink’s ISL, in 2024.

Do you have any more detailed or up to date information? I’m not doubting that they built ISL capabilities, since ISL is often used for collision avoidance and small scale satcom - I’m doubting that it is being actively used to route internet traffic.

I understand and do agree with you following your intended interpretation. Thank you for the patience and explanation.

Since when? Last I heard this was just vaporware.

The semantic difference between an agent "going rogue", and a script that generates random valid curl requests is essentially pointless in the eyes of the law and responsibility.

From a technical standpoint, the difference is scale and intent, but ultimately people and/or and organization are culpable for the output.

The original comment I replied to, in which you claimed that the difference between a safe Rust and C++ program with unsound behavior is that the C++ program needs to be fixed, but for the Rust program, the compiler needs to be fixed.

As I see it, in either case, the program can be fixed, or the compiler can be fixed. The only difference as I see it is that the C++ spec does not expect the compiler to be fixed.

Did I misunderstand your post?

When one is “full of shit” as they say, the shit needs to spill out somewhere.

Your shoulds and musts are nice and all but you said that in these situations, the rust compiler needs to be fixed when the C++ program can be adjusted.

I simply cannot imagine a situation where you can’t find an alternate solution to a problem with the compiler. Even if the compiler would optimize your intended solution away, you can write an asm! macro.

It sounds to me like you have simply decided that one language cannot do what the other can because you personally would not do it, which is very silly.

If this is not the case then I restate my request for an example, for the third time - an example would prove the point you are not making through explanation.

That’s not an example, that is you restating your post.

> If a compiled safe Rust program has unsound behavior, the compiler needs to be fixed

Can you provide an example of such a program that can’t be fixed at the program level like a C++ example?

I still like writing web applications in PHP for many of the same reasons listed here - modern PHP anyways.

I shy away from frameworks unless I am building something that I expect a large team to work on, or I specifically need multiple complex features that don't make sense to maintain.

I've written a few applications for my own personal use or targeted at friends/family recently that have no third party dependencies other than Bootstrap, which I still like more than tailwind for anything that doesn't need to look like a marketing page.

Given that we have had journalists prosecuted under criminal hacking statutes for discovering personal information in the rendered HTML of a web page, or publicly available attachments on a WordPress installation, I find the suggestion that intent is required to be dubious.

As someone who does think linguistically, I often end up at odds with people who are expecting me to imply things in my text, when my text is written specifically to convey exactly what I mean and nothing further.

Always happy to see Oxide growing. The way their company is run is pretty close to how I think all engineering orgs should run.

Hoping they hire for a position in my area of expertise some day!

Don't give secrets to agents at all that you don't plan on revoking immediately after.

If you have allowed an agent to access any kind of credential, you should assume it is no longer private.

What an incredible read. Thank you for bringing this into my life - I think it’s going to be one that sticks with me.

That actually looks pretty incredible, including some features I never knew I wanted.

Unfortunately I work for a very large corporation and don’t have the pull to get something like Epiq brought on - but I am definitely going to evaluate it for my non-FTE work.

As a side note - it was impossible to find your project through search engines. I only found it by searching on github itself.

Sounds a bit like a doomed version of Quantum Immortality

https://en.wikipedia.org/wiki/Quantum_suicide_and_immortalit...

I hate Jira because it’s one of the slowest applications I’ve used since the i486 days and the UX is designed to force you to click across 3-4 different fields and modals to accomplish an action that should be one form.

It’s business software for business people that need to be guided through SOP because they have so many different workflows, not engineers that have streamlined work processes to reduce mental load on non-engineering work.