Grumpy Dev | D T Dubs 🇺🇸 (@DTWaling) header
Grumpy Dev | D T Dubs 🇺🇸 (@DTWaling) profile picture

Polyunsaturated Fatty Pragmatics Impossible, no. Ask better questions. crew @AmericaMission_ Above all, love Christ Off the wall: https://xitter.gq/t.co/ynlUBp3ofx

Up the road a piece, Mitten
Joined December 2009
Some storms are not meant to break you Some storms are meant to make you The hard storm we have long weathered has made us stronger because we stood together Now We The People ARE the storm! Now with truth, love, and pure strength of will Our world will become stronger still!
19
19
4
120
7,502
"I guess, trust the science... that's what they say... I'll get in trouble if I tell you otherwise..." (made me wonder what Holistic Honey would be like as a super chill redneck, lol - h/t @doordashbliccy)
60
LMAO The petty tyrants ...er, I mean kids ...are grumpy & restless, and looking like their about to make a mess of the kiddie table.
Of course they had to meet because they weren’t invited to the adults table…
1
33
LFG! ❤️‍🔥
We've partnered with @Microsoft to bring Hermes Agent to the enterprise. Hermes Enterprise now works inside Microsoft Teams, the Microsoft Store and Azure. In Teams, employees hand Hermes work like financial analysis, document drafts or inbox triage, and review the result in the same chat. Agent workloads run in isolated Azure Container Apps sandboxes. When someone finds a good way to do a recurring task, Hermes saves it as a skill. With Collective Wisdom, they can share that skill with their team after review, so the next person starts out with the best approach anyone has found to accomplish that task. The skills then grow along with the company. Companies with strict data residency or procurement rules can run Hermes Enterprise in their own Azure tenant. Our CTO Jeffrey Quesnelle presents how Hermes uses these sandboxes at Microsoft Ignite today, in session BRK448. Thanks to @M12vc and the @Azure Container Apps team. nousresearch.com/hermes-ente…
17
Guys ...I think we can make this work.
1
36
I got wind of it within minutes and fixed & re-deployed stacks a while ago. I hope y'all have figured out you should be *actively* monitoring CVEs as a matter of practice. If not & paying clients rely on your web product, you're putting them at risk too, don't sleep on this.
‼️New phpAfterShell Attack Bypasses PHP2Shell Fix to Enable Remote Code Execution Details: cybersecuritynews.com/phpaft… A new PHP attack dubbed phpAfterShell can lead to unauthenticated remote code execution on vulnerable servers, even if they have the October patches for PHP2Shell. Researchers disclosed the flaw on November 19 and demonstrated how it could be combined with a separate bug in PHP to run code remotely. The vulnerability, tracked as CVE-2026-126714, has a CVSS v4.0 score of 9.2. The flaw sits in PHP’s default file-based session handler. Sessions let web applications remember a user’s login status, shopping cart, or other details between requests. #CybersecurityNews #phpAfterShell
51
How loud is your projection?
🧠 The Equation That Explains Human Behavior
19
Great post in case you wondered why it has been so difficult to combat security issues like prompt injection. The very thing that is an LLM's greatest strength (that 2nd "L" ...Language) is also it's greatest weakness. It isn't much different than miscommunication you hear between two people talking past each other. This particular issue is probably the most fascinating SI/AI topic to me. Language is powerful, but also highly flexible & dynamic (not to mention constantly shifting) which leaves a lot of room for error ...we see it every day between humans. What if all the labs agreed to limit LLMs to a more rigid, standardized, vocabulary? The trade-off would be a massive step backward in its usability, not the least of which is the semantic flexibility that allows for use of natural language (including all kinds of slang and colloquialisms). That is the very thing that makes it so user-friendly, such that *anybody* can use it for technical work well beyond their own capability (often beyond their understanding). Moreover, we'd land somewhere with accessibility/usability dialed back to users only marginally beyond the domain expertise requirements and still technically proficient, effectively deflating the very promise of possibility that AI is riding on today (...untenable for a large number of people for many reasons). It is an interesting dilemma and a much more difficult problem to solve than most people even realize. It is also why many of us see much of the big labs' "alignment" and guardrail efforts as laughable and counterproductive - but that's a whole other conversation on its own, lol. Nevertheless, security remains concern, and growing lazy about security is still the biggest weakness in any system. I could go on, but that's enough for now. And take heart, we can work on understanding the facts without getting all doomy... Just stay diligent as well as vigilant my friends, and don't sweat it too much - there has never *not* been plenty of work to do, there's still plenty ahead, and the future can still be bright. That is all. That's my 3 1/2 cents. Happy Hump-Day and have a good evening.🤓
It can’t. Not in large LLMs. In a trillion parameter model, prompt token space is too big. The gap between a sentence that is biomedical science and engineering a biological weapon is a few words. That translates to a tiny delta in logits being selected. I know this because I’ve hacked Muse, Claude, ChatGPT and nearly every other model via prompt engineering. @elder_plinius is the GOAT so h/t to them for some techniques I’ve used but it really isn’t that hard. Even with sentinel classifiers.
53
🎯 Everybody's got something to say, and how many are hoping you just believe 'em without looking at the numbers? The plot thinnins...
32
Grumpy Dev | D T Dubs 🇺🇸 retweeted
DEBRIEF: I have attached a copy of my statement before the Wayne County Board of Canvassers prior to yesterday's 2026 Election certification vote (H/T Braden Giacobazzi). Prior to my statement, a representative of the MI Bureau of Elections issued a statement to the effect of... "The Department will review the information provided by Senator Colbeck once certification is over" Translation? "Pound sand." Please note that I have submitted a similar request (with significant evidence supporting my concerns) to the MI State Board of Canvassers for a Physical Configuration Audit (PCA) of voting systems prior to the statewide certification vote. The MI State Board of Canvassers is scheduled to certify the #2026Election on Monday, November 23 at 10 am ET...prior to any confirmation that the election was conducted on certified voting systems. This vote will occur in Room 1100 of the Binsfield Office Building, 201 Townsend Street, Lansing, MI. @RealLindellTV @realMikeLindell @TalkMullins @EmeraldRobinson @MrJustinBarclay @Charlieleduff @BobDutko
65
447
27
725
8,993
Grumpy Dev | D T Dubs 🇺🇸 retweeted
DHS Secretary Markwayne Mullin had enough of the games today. Democrats went after him for flying to a football game, so he laid out the facts. He is required to take the government plane, he reimbursed the government, and he paid for the tickets himself. Then he called out the members who would have billed the same trip as a campaign event. Do your job and stop playing games. Video: House Homeland Security Committee Support and join the Grassroots Army as we take back our country, rooted in faith, built on family, and fighting for freedom. 🫡🇺🇸
1
5
18
479
I still hear from people who are avid SI/AI users who have not tried (...or heard of) Hermes Agent! If you want options in a *provider-agnostic* SI/AI agent harness, nobody does it better than @NousResearch! $15 off 1st month is plenty for a test drive portal.nousresearch.com/r/du…
54
Grumpy is as Grumpy does. ....where's my coffee? ... ...and who keeps moving my chair!? xitter.gq/i/spaces/1mdqzEJaYovKJ…

20
Wew! So glad all the influencoors warned us of the incoming chaos. Were it not for them how could we have weather this heroing event! 😏 PSA: Always be prepared - shit happens (nefarious and not). But ignore doomer influencoors larping like they have thee date & inside scoop.
The nationwide Emergency Alert System test is now complete. See you for the next test in 3 years, per federal mandate. We are aware of online claims that the test failed or did not happen because FEMA was hacked. Those claims are false.
31
🎯 Dear Globalists, Sure, please, go ahead and *try* to arrest our president.🤣 --- DFQ 🇺🇸 Know who holds the reigns - Amen.
Trump and Rubio getting arrested? Here's what actually happened. JoinTAB.us 17 November 2026 #freedom #leadwhereyoulive #tab #ICC
1
31
Making the Bezien Math Sherpa the next in-demand job role... xitter.gq/i/spaces/1mdqVLXjVonLM…

2
44
Grumpy Dev | D T Dubs 🇺🇸 retweeted
Let me tell you what’s really happening with this agenda against our military. To EVERY strong man working in America. Watch this before it’s too late.
This will probably land me tons of hate but I don't care. I'm a civilian, so I'm gonna talk about civilian things. The fact of the matter is this: there are tons of professions which are far more dangerous than serving in the US military. Even those actually in combat. Why should a veteran be "entitled" to anything that a logger, fisherman, or roofer isn't entitled to? What does our country need more of at the moment? Construction workers and loggers, or killing brown kids on another continent for the benefit of Raytheon's shareholder dividend? I have more respect for these guys than I do for most "run of the mill" servicemembers. Some servicemembers did incredibly brave and heroic things far above comprehension. Some of them moved crates between bases "outside the wire" and never saw a single round head their way. I worked in commercial insurance prior to this industry and saw plenty of permanently disabled and/or killed manufacturing workers over 10 years. Nobody's talking about those guys or their families being exempted from property tax or some other special entitlement. Can we all just be honest about this?
115
507
40
1,365
30,609