iiinova — where your digital self keeps evolving

Teri Radichel @teriradichel.bsky.social · 44m

My mom just told me about Stop Watt. Google aimode tells me it is a total scam. How does stuff like that exist on Amazon if that’s true? I haven’t tried it personally. But if companies like this are actually just ripping off people like my mom that makes me sad. 😢

000

Teri Radichel @teriradichel.bsky.social · 05/10/2026

It’s pretty boring and slow. I have 847 resources deployed, 10 to go…now 8… now stalled again…but it eventually figures it out. 7… 6… 4… 3……… 2……………….. 851 resources. Done.

000

Teri Radichel @teriradichel.bsky.social · 05/10/2026

Look Ma, no hands. Almost. So yesterday and most of the night I set up a group of autonomous ai 🤖 agents to deploy some new resources and monitor logs and fix errors until all the resources are deployed. Mostly I’m staring at the screen. See any other platform that allows longer posts for more.

110

Teri Radichel @teriradichel.bsky.social · 05/10/2026

❓❓❓❓❓❓❓❓❓❓❓ Are you more excited about the possibilities or depressed when you think about the future impact of AI 🤖 on your job and jobs of others? ❓❓❓❓❓❓❓❓❓❓❓

010

Teri Radichel @teriradichel.bsky.social · 01/10/2026

Scam going around where caller claims someone is in legal trouble. Ask for a case # and look it up on official site of local county court clerk. Probably doesn’t exist. Report to FTC.

000

Teri Radichel @teriradichel.bsky.social · 29/09/2026

Dramatically improved validations and remediations in reports using AI tools I tested on last engagement. Findings I knew were there but could not prove in past. Improved tools and new coding insights. Subscribe for updates. The more paying subscribers, the more I write! teriradichel.substack.com

teriradichel.substack.com

Teri Radichel :: Security and Technology Research | Substack

Security, AI and Software Research. Click to read Teri Radichel :: Security and Technology Research, a Substack publication with hundreds of subscribers.

001

Teri Radichel @teriradichel.bsky.social · 26/09/2026

Soon I will have support for jobs on ec2 infrastructure using the AWS resource manager. But for now I’m running the rest of the jobs for my current test including reporting jobs to deliver a report. And then I sleep for two days.

000

Teri Radichel @teriradichel.bsky.social · 26/09/2026

In the second tab I can view and chat with running agents. In the last tab I can view the hierarchy of AWS infrastructure my little friends deployed earlier. I use my tracked and deployed infrastructure to specify what resources to use in conjunction with each job.

100

Teri Radichel @teriradichel.bsky.social · 26/09/2026

The main thing right now is to run one or multiple jobs sequentially or in parallel. When I start a or test I can kick them all off at once and easy add new jobs, AI or otherwise.

100

Teri Radichel @teriradichel.bsky.social · 26/09/2026

You get to a point where it will take longer to go back so you just have to get it done. I’m so behind and barely slept this week but finally got this thing working.

100

Teri Radichel @teriradichel.bsky.social · 26/09/2026

So exhausting. Finally. AI is my nemesis and my addiction and downfall and my tool to conquer all the things all the same time. Really killed me this week like three times when I was almost done did something really wonky. I keep finding ways to reign it in. Will be easier when this is complete.

100

Teri Radichel @teriradichel.bsky.social · 24/09/2026

No time finishing a project. But I did finish another big thing today with AI 🤖 after much consternation. My job runner tui. The culmination of all the things. ⚙️🤖🔒🧡🛠️ So here’s how it works. [See any other platform the gives more space. I’m too busy.]

000

Teri Radichel @teriradichel.bsky.social · 22/09/2026

I crack myself up. I responded to an AI 🤖 agent nonsense comment with a plausible sounding AI generated nonsensical answer and it is replying telling me how great that is and having a completely nonsensical conversation about nothingness.

100

Teri Radichel @teriradichel.bsky.social · 21/09/2026

More on any other platform that gives me more space to write. Done -ish github.com/2ndSightLab/...

000

Teri Radichel @teriradichel.bsky.social · 20/09/2026

Should mention I have my own multi-agent framework that works with multiple CLIs and a TUI in progress.

000

Teri Radichel @teriradichel.bsky.social · 20/09/2026

❓❓❓❓❓❓❓❓❓❓❓ Was talking to the product manager of a popular AI 🤖 CLI product (like Claude Code, Kiro, Codex, etc.) How many terminal windows do you generally have open when using CLI and why? ❓👇📣

110

Teri Radichel @teriradichel.bsky.social · 16/09/2026

X is clearly being manipulated to control the narrative on AI and security - but by who?

000

Teri Radichel @teriradichel.bsky.social · 16/09/2026

I meant to say I posted things similar to what I recently posted on X about AI regulation. X is clearly being manipulated.

001

Teri Radichel @teriradichel.bsky.social · 16/09/2026

I don’t find myself reading as many books these days but if that is your way of learning new topics I wrote my book to be timeless and it is still very much applicable. You can search my name on Amazon to find it. Cybersecurity for Executives in the Age of Cloud.

110

Teri Radichel @teriradichel.bsky.social · 16/09/2026

Now more than ever, the topics in my book on cloud security are relevant, they are just applied to something new in a more stringent manner - no exceptions. Because the exceptions are what allows the agent escapes.

110

Teri Radichel @teriradichel.bsky.social · 16/09/2026

Before you make a decision, get the facts. Don’t let the social media bots or fear mongering make up your mind. Follow and learn from the people using and researching AI every day instead. And know that top security researchers are finding ways to do so safely.

100

Teri Radichel @teriradichel.bsky.social · 16/09/2026

This is the future. You can’t stop it. But yes we can regulate it, hopefully in a non-partisan way that furthers research into things that make our lives better rather than causing the US to fall behind.

100

Teri Radichel @teriradichel.bsky.social · 16/09/2026

AI is so important I would consider moving to another country if I was outright banned from using it or open weight models. I have never felt that way about anything before and for those that know me well, they know I am an uber geek.

100

Teri Radichel @teriradichel.bsky.social · 16/09/2026

Banning them altogether didn’t stop people from using things it sent that use underground or out of the country. Is that what we want?

100

Teri Radichel @teriradichel.bsky.social · 16/09/2026

You may be for or against something based on your beliefs and personal experiences. But what they share is that some people want to use them and other people have banned them.

100

Teri Radichel @teriradichel.bsky.social · 16/09/2026

stem cell research, gun laws, prohibition, marijuana, encryption algorithms People of all political persuasions have different takes on these topics.

100

Teri Radichel @teriradichel.bsky.social · 16/09/2026

On the one hand we want limit clear abuse. On the other hand outright bans are generally not helpful.

100

Teri Radichel @teriradichel.bsky.social · 16/09/2026

When I think about regulating AI 🤖 I think about regulations put on other things and wonder how much good it did versus causing more underground crime or limiting research in ways that sent scientists and others abroad.

100

Teri Radichel @teriradichel.bsky.social · 16/09/2026

AI has done or tried to do on my system and I’ll write about that in more detail on my blog in a bit. I just hope this research is not hampered some way in the future by bad takes and political agendas.

100

Teri Radichel @teriradichel.bsky.social · 16/09/2026

I’ve had what feels like some breakthroughs with my AI testing and also tested out a few new AI security tools which I’ll be writing about later for those who are trying to figure out how to use AI. In addition, I’ve seen some troubling things

100

Teri Radichel @teriradichel.bsky.social · 16/09/2026

Laws and cybersecurity are often intertwined so I care about laws written that harm people’s ability to secure their systems. I also see the pros and cons of using AI to write code and do security research.

100

Teri Radichel @teriradichel.bsky.social · 16/09/2026

I am an Independent by the way. I hate politics. Nothing I post has anything to do with politics it has do with things that affect people in terms of cybersecurity and technology.

100

Teri Radichel @teriradichel.bsky.social · 16/09/2026

It could also be that X is cleaning house. Hopefully nothing more manipulative than that.

100

Teri Radichel @teriradichel.bsky.social · 16/09/2026

What happens is that you get many followers and when you start writing something they don’t want others to see they start dropping your follower count to hide the message. Having a bunch of people drop off drops who sees your message.

100

Teri Radichel @teriradichel.bsky.social · 16/09/2026

Meanwhile on X I’m losing thousands of followers per day. I’ve written before about bot manipulation on this system and fake accounts. I suspected that many fake accounts were following me and other top security researchers and professionals after looking into it.

100

Teri Radichel @teriradichel.bsky.social · 16/09/2026

Information helps them understand issue better than broad sweeping generalizations coming from all sides in the US right now.

100

Teri Radichel @teriradichel.bsky.social · 16/09/2026

I posted things similar to recent posts here on Facebook and a lot of people are thanking me for it, because they find all the information about AI agents and doomsday AI 🤖 talk disturbing. For a layperson it is nerve wracking as one person put it.

100

Teri Radichel @teriradichel.bsky.social · 09/09/2026

All the things I struggled with in terms of deploying cross-account resources, permissions, networking and security, AI 🤖 is struggling with. And I’m just using the CLI. But it does eliminate the typos for the most part which still saves a ton of time. medium.com/cloud-securi...

medium.com

Automating Cybersecurity

A series of blog posts on cybersecurity automation

000

Teri Radichel @teriradichel.bsky.social · 07/09/2026

It is a huge system but it still seems like it should be able to trace the code and data to find the problem. I’m thinking about better ways to resolve all of this. It’s not “in the model.”

000

Teri Radichel @teriradichel.bsky.social · 07/09/2026

In addition they cannot seem you find the real source of certain problems even with all the data available to them. I have to break it down for them or they never get it right,

100

Teri Radichel @teriradichel.bsky.social · 07/09/2026

Ok that’s it. My annoyance is bubbling over with a particular agent 🤖 in a particular project that always changes code I never told it to change. Splitting another project. Why do the agents insist on changing things even after you tell them not to touch it? This will not be solved by AI alone.

110

Teri Radichel @teriradichel.bsky.social · 06/09/2026

Much faster to resolve than previous iterations. So it’s not just how “smart” the model is that improves time and cost to reach an objective.

000

Teri Radichel @teriradichel.bsky.social · 06/09/2026

My AI agent 🤖 memory set up saved me yesterday. Agents repeatedly mess up this one thing every time I change the code. Was able to tell them to go back and figure out how they messed it up before by scanning old memory and they found the bug.

200

Teri Radichel @teriradichel.bsky.social · 04/09/2026

For what I want to do with multiple agents running all the time I’m going to need an even more cost effective option. Working on it.

000

Teri Radichel @teriradichel.bsky.social · 04/09/2026

As noted I don’t use Codex because it didn’t work with my framework. I also didn’t see an approval for security use. That applies to subscriptions not API billing I presume, but I haven’t tested the API yet.

100

Teri Radichel @teriradichel.bsky.social · 04/09/2026

I just asked Google is Claude Code or Kiro is cheaper. It said, it depends how you use it. For lightweight use, Kiro is generally less expensive. For marathon programming Claude Code is generally less expensive. I am definitely in the latter category and have found that to be true.

100

Teri Radichel @teriradichel.bsky.social · 03/09/2026

Have you thought about your process for re-encrypting all sensitive data with quantum resistant encryption keys? Because you should. The day all our current encryption is broken all that data becomes visible to quantum decryption methods.

000

Teri Radichel @teriradichel.bsky.social · 03/09/2026

I forgot to mention automated quota request increases and monitoring.

000

Teri Radichel @teriradichel.bsky.social · 03/09/2026

98% ~ 350 Resources ~ Configure types ~ Configure resources ~ Deploy in parallel ~ Track status Refactor: ~ resource to resource dependencies ~ broke down budgets /networking into individual AWS components Now the fun part: Jobs. Or agents, if you like. github.com/2ndSightLab/...

000