Wes Drone

"infosec programming experience" = I can read code in various languages, copy/paste my way to my own working code, but there is no way i'm letting anyone else read my junk code without multiple self-deprecating statements and maybe a small panic attack

In an Uber on my way home from MSFT HQ yesterday I had a very similar thought. The only reason to be afraid of AI taking your job is if you’re not adopting AI in your job today. Many parallels to the rise in personal computer use in business 30+ years ago.

Stagger Inn, Edwardsville IL No dial tone… no hardline access.

It’s happening… CLIs are all the rage. Now we just need AI identities with configurable permissions and access filters.

I’ve always been impressed with yolo. This takes it to the next level.

Real-time object detection will never be the same. Traditional YOLO needs NMS to remove duplicate boxes; it's slow and inconsistent. YOLO26 skips it entirely: single-pass predictions, faster inference and up to 300 detections per image. Download model: platform.ultralytics.com/ultralytics/yo…

This is how you get “The Matrix”

I asked openclaw to watch a Facebook Reel of a meal prep recipe, create the recipe from the video, description and comments, add an event to my calendar with the recipe and add the ingredients to my shopping list. 1st Run, needed to fix access. 2nd Run - flawless execution. 🔥

@HackingLZ I’m doing a lot of AI work these days and there are parallels to AI benchmarking, but largely in that world you’re not expected to get to 100% and if you do after a long time then it’s time for a new benchmark. Maybe there is hope for what’s next but marketing is gonna market.

It's somewhat amusing because all the red teams have labs with these products and can tell you where the gaps are. Someone needs to force them into third party evaluations that include both known and unknown variants of TTPs with configurations that mirror the majority of their client base. Nobody should be getting 100% in these tests

Marketing departments are going to be in shambles at Blackhat with all these EDR companies backing out of ATT&CK Evals

Everyone’s arguing about the HHS and CDC and completely ignoring the fact that most US adults get their health information from fitness influencers on TikTok.

bbc.com/news/articles/… Most food science is junk science that almost always relies on people remembering what they ate a week or two after they ate it. Usually you have to track down the actual study for figure that out. Hat tip to BBC for putting it in the article.

In some of my one on one meetings we only talk about body building, lifting and food prep. If you’re remote and having work only conversations I encourage you expand your conservative horizons.

@ipurple @samkscholten We never published it. Probably still lurking in the archives at Proofpoint. We built an att&ck styled framework for bec and related schemes for phone and email.

@wesdrone @samkscholten Was this work disclosed somewhere? I would like to have a read if possible :)

This is pretty cool @samkscholten Looks similar to some work we did in the past.

Great example of model strengths, weaknesses, and work around prompting. “Sometime getting the right answer is as simple as asking the right question. “

OpenAI’s GPT-5 is available today in Azure AI Foundry. This new suite of first-class reasoning models from OpenAI delivers the most intelligent and safe responses of any model we have released to date: perfect for building AI applications and agents. Learn more: msft.it/6011s3vvX

AI is going to radically change many facets of technology and security. Does it live up to the hype, probably not, but it’s also not that far behind. Two years ago I wasn’t a believer, in two more years things are going to be very different.

I’m very interested in the fitness data that will be generated by the presidential fitness test and how it compares to previous years. I hope they publish it. newsweek.com/presidential-f…

Finished the book “I am Pilgrim”. It was good. Going to check out the authors followup book.