If you haven’t noticed yet I have a new blog so my posts here are a bit sporadic. You can find the most up to date information here:

Teri Radichel :: Security and Technology Research | Substack

The posts are free for the first 7 days before they get archived.

My most recent post is part of an on-going research project to track the time it takes to develop a real world application that is more than a script and not a hot mess to maintain. It also provides a new approach to configuring and deploying organization resources on AWS.

Managing an AWS Organization With AI Generated Code: A new take on the AWS Control Tower concept

Managing an AWS Organization With AI Generated Code

That is part of my whole series on creating secure AI agent infrastructure:

AI Agent Infrastructure Security

I’m tracking all my progress, mistakes, and cost in a GitHub repository as outlined in this post:

Tracking AI Success Rate On An AWS Infrastructure Project

You can also get notified when I post possibly eariler on the AWS Builder Center at https://builder.aws.com. It’s becoming a pretty cool place to find technical content and based on user feedback and the AWS wishlist it’s getting better all the time. A lot of AWS heroes are publishing over here now.

Teri Radichel - AWS Hero | Profile | AWS Builder Center

I’m trying to be a bit more methodical about how I’m publishing the posts. Instead of a daily stream of consciousness like I used to do on this blog, I’m first getting a bunch of things done, then writing about it. There are a few reasons for that. First of all I can’t afford to write so much in the day and age of AI. I need to get things done. Additionally most of what I write is slurped up by AI anyway.

I’m hoping that people will become paid subscribers on my other blog, a model I like better and is already proving to produce more income than Medium, to support the research, but so far it is not enough to cover my costs. Not anywhere close. So I have to spread out my posts and write as time allows between other things. But I have published a number of posts over there on my current research into using AI effectively. Check it out if are interested, and even better, if you want to — subscribe!

Thank you 🩵

— Teri Radichel

P.S. I don’t write anything on my blog with AI and when I research things using AI, I tell you. It’s just me. You can tell by the typos 🙃 which I try to fix but occasionally I miss some. Happy vibe coding. Safety first.

Tracking Costs, Time and Mistakes On An AI Project was originally published in Cloud Security on Medium, where people are continuing the conversation by highlighting and responding to this story.

YouTube Video of presentation at AWS Security Community Day at the Computer History Museum 2026

The presentations videos have been released from AWS Security Community Day. Here’s my presentation on How I use AI for Pentesting.

Presentation on You Tube: How I use AI for Pentesting.

How I Use AI for Pentesting on YouTube

If you are interested, here’s last’s year’s presentation on Threat Modeling a Batch Job Framework.

Threat Modeling a Batch Job Framework on YouTube

Subscribe here for more posts like this and follow Security Bugs.

— Teri Radichel

Teri Radichel :: Security and Technology Research | Substack

How I Use AI for Pentesting - Video released was originally published in Cloud Security on Medium, where people are continuing the conversation by highlighting and responding to this story.

I recently spoke at the Computer History Museum in Mountain View at AWS [Security] Community Day. That conference has expanded beyond security but it was still heavily security-focused which I love. Thanks so much to @john.varghese and AWS for helping me get back there again. I’ve added a link to my slides on How I Use AI For Penetration Testing here:

How I Use AI for Penetration Testing

You probably heard all about Mythos from Anthropic if you are even remotely following AI or security topics. I wrote a bit about that here which is that we cannot really know anything about it until we can actually try it. But the larger question I have has to do with the viability of relying on AI models for business given their transient and non-transparent nature. The inability to rely on value per token is a problem.

Anthropic Mythos

One of the issues I was having related to the whole Anthropic Opus 4.6 meltdown that affected an inconsistent number of people recently — and seems to be heavily skewed to security research users — was implementation of the Yubikey authentication architecture I’ve been working on and wrote about here:

Yubikey Push To Run A Lambda Function

I dove into some of the details of a Yubikey authentication system in this post:

Where are the Private Bits of FIDO2 Compliant Passkey Xtored?

I took some time off from working on that hoping Anthropic would figure out whatever nerfed Opus 4.6. Now I’m back at it and have extended my Lambda Troubleshooter in an attempt to quickly resolve problems with this and all future Lambda architectures. The larger point here is how effective use of AI non-deterministic solutions in combination with deterministic solutions can slow your token burn rate and help you achieve ROI with AI solutions.

Reducing Token Burn Rate With A Well-Designed Architecture

Subscribe on Substack for more posts and more timely updates.

Teri Radichel :: Security and Technology Research | Substack

— Teri Radichel

Reducing token burn, Anthropic Mythos, How I use AI for pentesting was originally published in Cloud Security on Medium, where people are continuing the conversation by highlighting and responding to this story.

I’m honored to be speaking at the AWS Security Community Day again this year. Come have a chat about the impact of AI on penetration testing, bug bounties, and learn about how I use AI for pentesting.

Here’s a teaser…

Register here:

AWS Community Day, 2026

See you!

— Teri Radichel

New Blog:

Teri Radichel :: Security and Technology Research | Substack

How I Use AI for Penetration Testing was originally published in Cloud Security on Medium, where people are continuing the conversation by highlighting and responding to this story.

🦊 New Blog — Security and AI Research: https://teriradichel.substack.com/

I’ve published this before but I’m now blogging over at Substack. Here are some of my recent stories. Stories are free for first 30 days. Subscribe and follow the sections that interest you.

• What I’ve vibe coded in the past 2.5 weeks that I couldn’t get done previously in like 4 years.
• A multi-agent work flow including security considerations, running things sequentially or in parallel
• KISS — Keep It Simple Silly — to improve AI coding results
• Securing custom agents — lots of considerations

• I’m currently working on an updated version of kicking off batch jobs with MFA — or more specifically, a Yubikey. Currently my two factors are IP address and a Yubikey but I’ll probably add a password into the mix later.
• I showed how to set up a secure pentest environment for Burp Collaborator
• I create a script for finding rogue TLS traffic based on Certificate Fingerprints.
• There’s a post on spotting a Man In The Middle Attack on AWS which is not always going to show up in VPC flow logs…

I have a section for those who are less technical now called Cybersecurity Me, or anyone trying ot improve personal cybersecurity.

• One day I logged into my Fidelity account and CoPilot was connected — what? I have never used CoPilot.
• Beware of what OpenClaw can do if you use it …
• Got tired of phone scammers on my phone — here’s how you can report them
• Does your bank support Yubikeys? They should. Not Passkeys. Yubikeys.

Wrote some posts on security bugs. Expect more in this category soon.

• Took a look at a “vulnerability” in Amazon Bedrock when creating agents in a sandbox
• I showed how to find security bugs with AI while coding
• Took a look at which programming languages are best at preventing serialization bugs that lead to vulnerabilities

• Where are the crytographic bits stored when you register a FIDO compliant passkey — it depends
• I create a script that monitors network traffic including showing you the executables that are generating that traffic
• Took a look at what caused the Lexus Nexus Breach
• Dove into an authentication architecture and things you should be asking your vendors about that

I have links to all my class materials over there. May migrate everything to that platform in the future.

Then there’s my link to get the code on GitHub currently…which may move in the future.

If you are just getting started with AI coding here’s how:

Implement A Simple Project With Kiro CLI and a README.md

Subscribe at this link and follow the sections that interest you most.

teriradichel.substack.com

— Teri Radichel

On March 7th, 2026 I published a post about creating a way to push a Yubikey to run a job. As always one thing led to another and here I am again, essentially trying to set up a common infrastructure to execute batch jobs (or if you must, AI agents, though not all my jobs use AI).

In this post I’ll tell you what I have accomplished since that date as of today, March 24th in the wee hours of the morning. That’s 17 days later. About 2 and a half weeks of vibe coding. I have a reusable bootstrap script that deploys an AWS environment in an OU with multiple accounts and resources. Skip to the bottom of the post for the full list of resources.

What I’ve Vibe Coded In 2.5 Weeks

On March 7th, 2026 I published a post about creating a way to push a Yubikey to run a job. As always one thing led to another and here I am again, essentially trying to set up a common infrastructure to execute batch jobs (or if you must, AI agents) …

This is essentially an implementation of the security automation framework I have been trying to complete for like four years here

Read full story

Subscribe to my Blog and follow Good Vibes for more stories like this.

— Teri Radichel

Stories this week published to the Substack blog — teriradichel.substack.com

🦊 New Blog — Security and AI Research: https://teriradichel.substack.com/

I’m finding that Substack is not as good at SEO as Medium. If you don’t already have a following it may be easier for you to get started on Medium. But overall I’m liking the functionality that Substack has to offer. For those not following me over there yet here are some of the latest stories.

I tried out AI in a Lambda function (Querying CloudTrail Logs and getting Recommendations to Fix Them), played around with an Amazon Bedrock agent, and showed how to create a simple AWS Kiro CLI project with a README.md file. I also explain how to create cookie cutter projects to save on tokens.

AWS pushes steering docs for Kiro but I’m having decent success with a simple README.md file for my purposes. I might not be doing what everyone else is doing but for me this simple approach is working well so far.

I also wrote about a number of strategies for using AI to find security bugs for those who are interested in such things.

My lab code for classes is now on GitHub and I explain some of the things I attempted to do to minimize the risks associated with giving students access to cloud resources in classes.

I’ve also got a link to my GitHub posts in the old-school looking CODE category.

You can follow for free or subscribe and choose only the categories which interest you rather than everything if some of the topics do not pertain to you.

Enjoy!

— Teri Radichel

teriradichel.substack.com

Teri Radichel :: Security and Technology Research | Substack

AI in a Lambda Function, Amazon Bedrock Agents, Cookie Cutter Vibe Coding, Finding Security Bugs… was originally published in Cloud Security on Medium, where people are continuing the conversation by highlighting and responding to this story.

Stories published this week ending February 13, 2026

🦊 New Blog — Security and AI Research: https://teriradichel.substack.com/

As most of you know I’m primarily publishing on substack now. You can follow me over there for the most recent stores. I’ll still publish a few here and there on Medium. The cool thing about Substack is you can subscribe to the sections I write about that interest you most instead of getting everything I write. I realize some people are more interested in AI than security and vice versa so subscribe to whatever applies to you in particular over there.

Here are the stories I published this week:

Non-Deterministic

The most important word you need to understand about AI

https://teriradichel.substack.com/p/non-deterministic

Non-Deterministic

Report Phone Scams to the FTC

Fake calls about loans in progress exist all over the BBB scam website but that is not where to report them to get them to stop

https://teriradichel.substack.com/p/report-phone-scams-to-the-ftc

Report Phone Scams to the FTC

How To Spot Impostor Traffic When The Network Logs Look Correct

How can you tell if someone spoofed an IP address and has compromised encryption in transit?

https://teriradichel.substack.com/p/how-spot-impostor-traffic-when-the

How To Spot Impostor Traffic When The Network Logs Look Correct

Spotting a Man In the Middle Attack on AWS in Network Logs

When your IP logs will tell you that your traffic is being intercepted and when they won’t

https://teriradichel.substack.com/p/spotting-a-man-in-the-middle-attack

Spotting a Man In the Middle Attack on AWS in Network Logs

Securing Pentest and Bug Bounty Research Networks on AWS

Make it harder for an attacker to compromise your pentest infrastructure and snipe your bounties

https://teriradichel.substack.com/p/securing-pentest-and-bug-bounty-research

Securing Pentest and Bug Bounty Research Networks on AWS

https://teriradichel.substack.com/

The most important word you need to know about AI, traffic not visible in network logs, securing… was originally published in Cloud Security on Medium, where people are continuing the conversation by highlighting and responding to this story.

(Except the get the code section coming later)

Mission accomplished. I have a post in every section on the new blog. Except the last one where I’m going to hopefully publish actual code — that’s coming later for various reasons. One of the cool things on substack is that you can follow only the sections that interest you most.

One of the sections is for less technical users and people who want to secure their own assets. I created that section because invariably I’m giving a presentation on something unrelated and people just want to know how to “secure their own stuff.” I also get questions from my Mom who says she has no idea what I’m talking about. I’m going to try to keep it simpler in that section and explain technical things hopefully in terms anyone can understand, but maybe just on the edge of getting technical for those who need and want that information.

That section is called Cybersecure Me and I just wrote a post about Bank Accounts and Yubikeys which seems to be pretty popular.

The other section I hadn’t yet published in is called Security Insight. This is a catch all category for people who work for organizations trying to reduce cybersecurity risk. I might analyze breaches or provide insights that help companies lower their risk by taking various actions. This category might also be interesting for those who invest in cybersecurity products.

I just posted about Using AI to Analyze Network Packets which I think is going to be a game change for defenders working in SOCs. As a side note I gripe about products and companies spamming the Internet with packets in the name of security.

Overall I think this blog is more representative of me at the current moment than my existing Medium blog. It hits the categories I want to talk about more directly and it targets the different groups of people who have different interests a bit better. We’ll see how it works out.

Subscribe for free. Posts will end up behind a paywall after 30 days the way I currently have it configured.

https://teriradichel.substack.com