Ryan Duff

AI Security @ TrendAI. Former Cyber Operations Tactician - USCYBERCOM. Tradecraft stickler. Tech Policy/Law Hobbyist. Sneakerhead. USAF vet.

[Ryan Duff @flyryan]

@QuinnWorthWork @Ascion_Next @cwdegnan @matanSF Six people who haven't posted to twitter in years all decided to jump on the same day and post similar responses uncoordinated? You REALLY believe that?

[Ryan Duff @flyryan]

@trq212 Any way we can get more than a 30 minute cap on a watch monitor? It's annoying to have to restart a monitor every 30 minutes. The alternative is an indefinite one-shot wait on a bash script, which isn't ideal. 🙏

[Ryan Duff @flyryan]

@Uber_Support @Uber_Support cancelled and ordered a taxi. You guys really blew this entire trip.

[Ryan Duff @flyryan]

@Uber_Support I tried. Moving my business (and business account) to Lyft.

[Ryan Duff @flyryan]

@Uber_Support @Uber ^

[Ryan Duff @flyryan]

@Uber_Support To add insult to injury, the car I booked for my airport trip (my last uber trip) hasn’t even been assigned and it’s 8:30. When I message support, my only option is 6-12 hour messaging. What happened to Uber? We chose you for our business because you had service.

[Ryan Duff @flyryan]

@trq212 Any plans to adjust the messaging in Teams mode? I had to write my own file based messaging system to insure immediate delivery to agents because the messaging tool seems to queue up after the turn resolves, which could be 20 minutes or more.

My workaround is the lead writing to a command control file and the subagents setting up a monitor that fires (after a 10 second debounce) when there are changes to the file. That fires immediately on the next call and the agent gets the message. It would be awesome to be able to have the agent messaging do that.

[Ryan Duff @flyryan]

@vulnresearchlab Any relation to https://t.co/IxxXe7hZS3 ? Your company popped up in my Google Alerts.

[Ryan Duff @flyryan]

@BLKMDL3 Picking mine up tomorrow!

[Ryan Duff @flyryan]

@elonmusk @Tesla this is what I ordered (CLEARLY says T2). It never said T1 anywhere on the estimate. The fact that the tech admitted they would have gave me them “if I asked specifically” but they gave me last gen instead is the most upsetting part. It’s like giving me a 2025 model when I ordered a 2026 and then saying it’s my fault because I didn’t say I wanted a 2026…

[Ryan Duff @flyryan]

@elonmusk as much as I love my @Tesla, the service experiences are so awful that I’m unlikely to buy another.

Ordered new set of T2 tires to have installed. They initially put wrong sizes on the rear and then told me it was right when I called it out. Had to press to get it fixed.

Then, when I picked up my car, they are the last generation T1 tires and apparently it’s my fault for not explicitly telling them I wanted the specific tires I ordered.

Now I have last-gen tires I paid $2100 for that aren’t what I ordered.

[Ryan Duff @flyryan]

@chompie1337 Someone posted this tweet on our internal Teams (I'm at TrendAI now) and I was so stoked to see it was you! I hope things are good with you! Love seeing you still poppin shells. :-)

[Ryan Duff @flyryan]

@Drew_OnAI @runwayml Or literally just try it yourself.... Did you even click the link?

https://t.co/5qEH6irxvv

[Ryan Duff @flyryan]

@AlphaSignalAI I can't find anything that says Anthropic wrote Claudini. Could you provide a source?

[Ryan Duff @flyryan]

@ryunuck @jasager029 @AnthropicAI What an analogy dude.... by your logic, AI should be the absolutely most secret thing where we don't share anything and procecute people when they do...

How does the manhattan project at all make for the right comparison here?

[Ryan Duff @flyryan]

@MTorygreen @AnthropicAI Your logic is horrible.... you think these labs don't have access to the same internet? If that's all it was, why would they need to distill from Anthropic?

[Ryan Duff @flyryan]

@antigravity I wish you guys were more up-front about there being limits in Antigravity even for paid users. I upgraded to Ultra when I ran out of quota and it obviously did nothing. Now I can't refund it even though I'm unable to use it right now.

[Ryan Duff @flyryan]

@mikeyk Sonnet 4.5 is definitely more honest! But it still seems to reward hack... Getting it to do a long task seems difficult sometimes. The new auto-compact is awesome though. I assume it's making use of the new APIs. :-)

[Ryan Duff @flyryan]

@karljobstgaming Theory: Getting so many things wrong is part of Call of Shame's plan. It baits a video from you, drives more engagement for him, and gives him content for a new video. An AI slop channel only cares about ad revenue.