Mark Percival retweeted
I deleted all my SSH keys
blog.infected.systems/posts/…
Reminiscing on the end of Salon and found this gem from 2006 - Why Johnny Can't Code - salon.com/2006/09/14/basic_2…
It's been 20 years and now with kids of my own I'm in a similar position, but is it even coding?
@DavidBrin We need an update on this!
2
258
This is an insanely cool project and team. You can write UI-heavy apps and games on everything from a 2004 PSP to an ESP32, and on your Mac. Hardware is fun again.
Mark Percival retweeted
Your laptop isn't built for running agents.
Introducing @instabox_sh: cloud computers for AI agents.
We provide multi-OS sandboxes (Linux, Android, Mac, Windows) with GPU and CPU.
We launched in beta <1 month ago, and today, users have created 1M+ boxes.
Give agents their own cloud computers so they don't fight over the same memory, CPU, and ports on your laptop.
Close your laptop, and they keep working.
Get started today → instabox.sh
I need a setting on OpenRouter that automatically disables the key if you use it with anything but Chinese models - You see my key trying to use OpenAI or Anthropic, I've been hacked!
3
5
391
I’ve been a remote coder (vim + vps) for 15 years, so it’s not surprising, but I’ve been 100% sandbox-pilled.
I’d recommend our portco @instabox_sh - but regardless of who you pick, get on a sandbox. It’s like Telnet vs ssh.
5
213
Regular SSH keys suck, they live forever, and there's no way to know if they've been compromised. At this point hardware SSH keys should probably be the default for EVERYONE. You should confirm every ssh auth with a physical button press/bio auth.
blog.infected.systems/posts/…
Mark Percival retweeted
Putting biased competitor comparisons on your landing page is the dumbest things you can do
Especially in devtools/infra. Developers aren’t stupid. They can tell when you’ve cherry-picked a comparison to make yourself look good
Seriously, who are you trying to fool?
And yes, I’ve written these comparisons myself before. Never doing that again
Here’s what I do instead:
1. Write everything I care about into an eval.md, with clear criteria and real-world workflows.
2. Give Claude Code and Codex the eval, plus the CLIs and API keys for InstaBox and all our major competitors.
3. Let them run the same tests and compare the actual results.
4. Turn everything we lose points on into Linear tickets. Our software factory picks them up, implements fixes, and runs QA.
5. Rerun the eval every week. Repeat.
I’m paying our competitors $3,000/month for their team plans just to find out how much better they are than us 😅
The first time I ran it, InstaBox was literally at the bottom of the list. And it wasn’t even close in some areas.
Our competitors are really fucking good. They’ve built some incredible infra, and we’re still behind in a lot of areas
But week by week, our scores have been climbing
We still lose in plenty of areas. At least we know what to fix next
I’d rather know where we suck than pretend we’re the best
Sandboxes force you to confront how you handle dev auth and tokens. Once you've made the move, going back to a world where a program has access to your home dir feels insane.
3
120
Mark Percival retweeted
Hacked my own VPS last night with abliterated models. Took less than 10 minutes and $5 in inference, and got root level access to the box and all credentials.
Lots to process and share soon, but feeling deeply in touch with my eastern european roots this morning...
This is especially galling for those of us with root passwords longer than 255 characters. All that wasted time typing!
1
5
454
Mark Percival retweeted
Firezone Cloud EU is live 🇪🇺
Your account data stays in Sweden and France. Your encrypted traffic takes the shortest path worldwide.
EU data residency. Full-speed global access.
firezone.dev/blog/firezone-c…
Now is the time to go through your connected accounts on Google and ruthlessly cut access to anything not absolutely vital that has elevated access. This is how your email get hacked, even with 2FA - myaccount.google.com/linkeda…
3
4
299
Mark Percival retweeted
Replying to @MalwareTechBlog
MPTCP is just one exploit vector - the use after free is actually in IPv4 options handling.
Simply dropping IPv4 packets with IP options may not be enough to fully mitigate.
See my post here: teapawt.net/jamilbk/status/2116950…
Here are my notes so far on CVE-2026-98432 after digging into this with Claude + defensive access:
1. Not just an MPTCP thing - any received IPv4 packet with options creates the problematic conditions.
2. Also appears that using NetLabel/CIPSO (SELinux or Smack with a CIPSOV4 mapping) creates the same conditions that a packet with IP options would. If you're using these, it might mean that dropping IPv4 packets with IP options might not fully mitigate.
3. At least for our test VMs in Azure (accelerated networking) I wasn't able to get any IPv4 packets with options through, from three different US-based ISPs. Unfortunately Microsoft publishes no docs on whether the fabric drops this or not, so YMMV.
We're using a simple XDP program to mitigate until a patched kernel is available:
github.com/jamilbk/drop-ipv4…
Honestly, your main machine should probably become a very locked-down UI for interacting with sandboxes. The threat to developers is higher than it's ever been.
To be fair, this is where we all should have been headed for years, but we ignored the early risks.
Replying to @guillewrotethis @github
Ok how about rephrasing:
“Remove any agent from your main computer. Now.”
Agents are absolutely dangerous and should only be used in disposable environments















