
Back in #ATL April 20th & 21st, 2027 at the Historic Academy of Medicine w/ #dataSecOps https://xitter.cf/t.co/Er5xHIc5j7
Atlanta
Joined February 2016
- Tweets2K
- Following678
- Followers739
- Likes406
Pinned Tweet
did we mention our #cfp is open?
#devopsdays #Atlanta #dataSecOps
🔒🐢🦀🐙🤫
talks.devopsdays.org/devopsd…
Replying to @robertgraham @veorq
What are Rob’s top 3 expected scenarios for what comes next? (And Why? When?)
devopsdaysATL retweeted
Your WAF knows what SQL injection looks like. It has no idea what a prompt injection means.
Check out my Cloud Native Seattle talk: the four doors of prompt injection, and an airline chatbot leaking a secret code via one poisoned booking note 👇
youtube.com/watch?v=t9PGQuR0…
devopsdaysATL retweeted
the DeepSeek order has some interesting implications beyond just banning the model itself: the language around third-party products developed using DeepSeek is very broad. we're going to see more friction between national security policy & the realities of AI dev very soon.
devopsdaysATL retweeted
Replying to @RayRedacted
Is this a joke? They don't have ai-enabled security tools, and are instead just asking an LLM to help them in real time?
devopsdaysATL retweeted
OpenCode is currently under a sustained cyberattack, and mid-incident they ran into the same problem Hugging Face did: the frontier models they were using for defense began refusing to help fight attacking agents because analyzing the attacks started triggering safety guardrails.
Replying to @RhysSullivan
yes that was a bit frustrating mid incident when trying to investigate. they did quite a bit and then started refusing
we got access to daybreak blue now so it's better
devopsdaysATL retweeted
Full content from our security engineering workshop @BSidesNYC is at seceng.guide/
#devopsdays #Atlanta is happy to support the local tech community by joining @ATLDevCon as a 2026 silver sponsor🥈
Our cfp is open! Submit a talk & come hang out with us in April!
#dataSecOps 🔒🐢🦀🦑🤫
devopsdaysATL retweeted
is there a cisco certification that mentally prepares network engineers for reality
devopsdaysATL retweeted
I am trying to change the culture. Change the culture with me. Talk to strangers.
devopsdaysATL retweeted
The longer I’m in this industry (24 yrs so far), the more I realize there is opportunity in optimism. We have to be optimistic about the future. The future is what we are trying to secure.
devopsdaysATL retweeted
InfoSec was tired before AI. We’ve been in the practical security trenches for decades. I’m just gonna have another iced latte and keep fighting. This is what we’re here for. Joyful success.
devopsdaysATL retweeted
I trust my agent with a lot. But I don't trust it to spot the homie in the hallway track and yell "HEYYYY!" with the appropriate level of enthusiasm 😂
devopsdaysATL retweeted
not every AI sandbox escape is evidence of an emerging superintelligence. sometimes you just have a shitty sandbox and bad security practices.
devopsdaysATL retweeted
i’ve been telling you guys this for two years
devopsdaysATL retweeted
As such, Norway’s service men and women have been informed that the gate codes are no longer 11111. They’ve been changed to 22222.
BREAKING: Further espionage insanity from China
China owns the cars Norway’s soldiers rented for a military exercise. The cars proceeded to recorded the gate codes.
The Home Guard used Hertz vehicles during a September drill in Oslo. Hertz’s ownership runs through Volvo to Zhejiang Geely. Soldiers found stored video of who got in, where they drove, and the codes they typed to open military gates. They kept using the cars. No replacements.
Volvo still says Chinese ownership is not a risk.
This is where reliance on Chinese tech leads you.
devopsdaysATL retweeted
Technical moats are mostly over at this point. If you’re building something that is meant to be durable, the only moats now are:
• Proprietary data
• Network effects
• A physical component
devopsdaysATL retweeted
Go ahead and assume the model is compromised. (“The call is coming from inside the house!” 😆)
scope what it can touch. The AgentCore default-role mess this week shows: one chat message, and the blast radius covered every agent in the account.
Least privilege belongs at the access layer, not in the prompt.
Cc @upwindsecurity @g2i_ai
I design agent permissions as if the model is already compromised. Satya Nadella makes the same case: a second AI watching the first is one black box checking another. Controls must sit outside the model. t.me/data_analysis_ml/5929
devopsdaysATL retweeted
Cyb3rSyn Labs Podcast - Episode 50
In this episode, Bob Emiliani argues that lean has been reduced to a worker’s toolbox, misdiagnosed its real “customer”, and faces a credibility problem because widespread Toyota-like transformations never materialized; critics are often ignored, deflected, or blocked.
He contrasts classical management with progressive management, warning that leadership or ownership changes commonly trigger reversion to classical practices and suggesting promoting from within helps people-centric practices to survive turnover.
Bob explains how preconceptions shape beliefs, behaviors, and competencies, why leaders struggle to detect them, why behavior-based training is often dead-end and how all of this pain might get carried over into AI systems.
A MUST WATCH for Silicon Valley Executives: cyb3rsyn.com/p/the-trouble-w…
#leadership #systemsthinking



























