ATechGuy

What was the first concrete thing you needed that existing sandboxing tools (Docker/VMs/bwrap) just didn't provide?

> As long as there is a full OS running, you are one libc function away from a sandbox escape.

Does this mean, all software in the world is just one function away from escape?

That's a fair criticism.

My thought process was that the original comment was based on their personal experiences and since ChatGPT is trained on a large dataset, it may offer a different perspective derived from experiences of a lot more people.

> "you're wrong, the computer says so"

My thought: you're knowledge may be limited, this is what a computer trained on a lot more data says:

When scaling out, edge latency will overshadow kernel boot-up times: speeding up boot-up from 1.5s to 150ms will not have any perceived impact on app performance when scaling on edge to meet the demand.

Fast boot up means nothing if your agent/app is slow at runtime (due to virtualization tax or QEMU emulation). Fast boot up is a PR term, which can easily be optimized for compared to designed a better virtualization layer that performs near-bare-metal.

@nl But this uses nested virtualization on GCP, which severely hurts performance. Faster boot up means nothing if everything else is slow.

Genuine question: why not just use GCP/AWS VMs for agentic execution? What is missing?

Scrapers use residential IP proxies, so blocking based on IP addresses is not a solution.

It is really free? Genuinely asking.

From ChatGPT:

This approach can stop very basic scripts, but the claim that “99.9999% of scrapers can’t execute JS or handle cookies” isn’t accurate anymore. Modern scraping tools commonly use headless browsers (Playwright, Puppeteer, Selenium), execute JavaScript, support cookies, and spoof realistic user agents. Any scraper beyond the most trivial will pass a JS-set cookie check without effort. That said, using a lightweight JS challenge can be reasonable as one signal among many, especially for low-value content and when minimizing user friction is a priority. It’s just not a reliable standalone defense. If it’s working for you, that likely means your site isn’t a high-value scraping target — not that the technique is fundamentally robust.

Genuine question: if traffic to websites is down and likely to continue the trend, what would be the value? Do you plan to address agentic browsing in the near future?

Why not use KVM?

Having used this for our research, I highly recommend Rtrvr.

Maybe we should write an agent that writes an agent that writes an agent...

Then why only apply to VMs, why not apps?

Side-channel attacks apply to multi-tenant cloud environments, not local.

Why not use VMs from AWS (EC2) or GCP? E2b is built on top of GCP anyway.

Startups would build on big tech, so are likely to add their margins. Have you looked into (bulk) discounts from GCP/AWS?

Thanks! While I agree with you on "saving seconds" and overall latency argument, according to my understanding, most agentic use cases are asynchronous and VM boot up time may just be a tiny fraction of overall task execution time (e.g., deep research and similar long running tasks in the background).

> It does take 1-5 seconds to boot the environment (firecracker vms).

I'd say 1-5 secs is fast. Curious to know what use cases require faster boot up, and today suffer from this latency?

That's what the author is claiming. Practically, VM-level strong fault isolation cannot be achieved without isolation support from the hardware aka virtualization.

Why? stopping LLM crawlers is a need now. We've seen more capable people working for undergrad dropouts.

> Meta seems like one of the few large tech companies where if the whole company vanished, the world would be purely a better place.

100%

Speculating based on another post: "...our investors are pushing us hard to frame it as AI..."

If it can be made configurable.

static analysis DOES NOT hallucinate.

You are assuming that they build their own models.

What about packages that are not "flagged"? There could be hallucinations when deciding to (or not) "flag packages".

You rely on LLMs riddled with hallucinations for malware detection?

> We are a 5-person, VC-funded team with PhDs

Who are your customers?

> we’re rethinking serverless from scratch:

What are the use cases that you target?