At Red Hat, Scott McCarty is Senior Principal Product Manager for RHEL Server, arguably the largest open source software business in the world.
Akron
Joined March 2009
- Tweets8K
- Following353
- Followers5.1K
- Likes4.8K
Really good testing. Like Stephen J. Gould, I approve of the statistical methods used in this. Experiment setup is so important. Great findings too! #security blog.checkpoint.com/ai-secur…
Torvalds responded in his trademark fiery way, claiming that people refusing to use AI to code should also forgo coding with an assembler and writing with an editor, finally calling the idea of setting such a rule "utterly stupid." #linux #opensource
xda-developers.com/linus-tor…
I agree that this attack vector is the next one to pay attention to - defense in depth. Also, it's why I use Trentina (github.com/crunchtools/trent…) for everything I do #security #mcp
cohere.com/blog/concerning-b…
Are you curious what all of the fuss is about Clearinghouses? Worried about the CVE apocalypse? I hope this piece helps! #linux #opensource
infoworld.com/article/423609…
I've had a few customers ask, so figured I'd share this! #rhel #satellite #linux
crunchtools.com/synchronizin…
I've been building Trentina with mix of deterministic and non-deterministic guardrails. Always good to read how others are thinking of these problems...
developers.redhat.com/articl…
Good content, helping people get started with bootc on RHEL!
youtube.com/live/0siLuf0bv-E…
I hope this piece will help people understand why Clearinghouses exist, what they're for, and why they're so interesting to people right now. #lightwell #redhat
infoworld.com/article/423609…
With my ever-increasing token usage, the advent of Framework computers with Ryzen chipsets withb192GB of RAM, and open models that fit, it's more tempting every day...
developers.redhat.com/articl…
Even in the post-Mythos world, with AI-assisted discovery and exploits, it's still important to emphasize that collaboration on responsible disclosure is still an important part of the solution. #linux #security #rhel #fedora
redhat.com/en/blog/responsib…
Bcvk makes it really easy to spin up a machine with bootc! #rhel #linux #fedora
developers.redhat.com/articl…
Interesting news. Even Linus is using LLMs now... lwn.net/Articles/1100449/
Your PDF thumbnail step was a 9.8 RCE. Defensive AI found it before attackers did: wordfence.com/blog/2026/10/w… #infosec
"Can I bet my platform on this?" is what customers actually ask when asking for dates. Marty Cagan says the real trade-off is predictability vs outcomes, and a date today buys mediocre software on time: svpg.com/predictability-vs-o… #Product
Errors are the attack vector. Make an agent's secure fetch fail and it runs attacker code. Trentina stops it. crunchtools.com/defending-ag… #AISecurity
Vendors offering to pre-inject customizations into your container images sounds convenient, but you're trading sovereignty, agency, and security for ease. Keep it in your own CI/CD. crunchtools.com/your-contain…
Open weights ≠ open source — but even the former OSI head says he'd take them. Anthropic is the one major AI company that didn't sign the open-weight coalition. Whose business model depends on proprietary winning? zdnet.com/article/open-weigh… #AI
RHEL's MCP server now runs AI-generated commands with three guardrails: gatekeeper AI, human approval, systemd sandboxing. Not zero — three.
redhat.com/en/blog/dynamic-t…
#RHEL #AI
I use LLMs every day but hadn't dug into how Transformers actually work. This LWN comment thread on Debian's AI policy accidentally became the best education on token selection, temperature, and determinism I've found.
lwn.net/Articles/1061544/
#LLM #OpenSource
Got hacked twice in one month. AI helped the attackers chain CVEs into a sophisticated multi-stage breach. AI also helped me kick them out, patch, and harden — in 3-5 prompts, on work calls. The balance is tipping back. crunchtools.com/my-personal-… #infosec
