iiinova — where your digital self keeps evolving

Matt Crosslin @grandeped.bsky.social · 14/10/2026

And that is also something that editors help with as well: style choices. Running a detector on a published article is just too late in the process, even if you believe it will work. It has been through a process to perfect the writing.

100

Matt Crosslin @grandeped.bsky.social · 14/10/2026

So using Ai detection on a published article, after it has gone through multiple editor to remove all human errors, is pretty much malpractice. You can't detect if something like that is written by Ai, just how close it comes to the writers that Ai was trained on.

200

Matt Crosslin @grandeped.bsky.social · 14/10/2026

But we have to realize that generative Ai itself was trained on millions of writings by people. It can't come up with anything original. So it is not writing its own thing, just stealing what others have written. Ai detection just comes down to catching human errors to say "not Ai!"

100

Matt Crosslin @grandeped.bsky.social · 14/10/2026

There are tools out there that will use Ai to "humanize" Ai-generated articles so that they can't be detected. I have tested their output on many Ai detection services, and none can detect it. Just feeding various examples of human and Ai created writing into a detection tool doesn't really test it.

200

Matt Crosslin @grandeped.bsky.social · 14/10/2026

The true red flags are all the people just acknowledging that Ai detection is garbage but still moving forward with the investigation. You only do that if you are wanting the oppression part.

0142

Matt Crosslin @grandeped.bsky.social · 14/10/2026

Two things are intertwined here: there is no such thing as “reliable” Ai detection no matter how much some wants there to be, and Ai is a tool for oppression of specific groups. Ignoring both of those things and acting like Ai can be used to detect itself will lead to more oppression.

56823

Matt Crosslin @grandeped.bsky.social · 13/10/2026

Paxton would not still be in office if things like this were enough to keep millions of Texans voting for him. Talarico’s strategy seems to be to try to win over people that don’t care about any of this last time Paxton was up for re-election. Which is what Allred, O’Rourke, etc, etc also tried.

010

Matt Crosslin @grandeped.bsky.social · 13/10/2026

Right on schedule, Talarico is trying hard to lose the election. Just like the TX Democrats before him. His new ad talks about how he called Biden out for “failing” at the border. No Republican is going to vote for him based on that. And now Democrats will think he has gone MAGA by repeating them

141

Matt Crosslin @grandeped.bsky.social · 10/10/2026

I guess I have finally leveled up in the teen boy parenting department. He used to ignore me at band events when his friends were around. Now I get half a glance and half a smirk when I say hi to him and tell him “great job!” After halftime. 😂

020

Matt Crosslin @grandeped.bsky.social · 10/10/2026

People seem to be mad at me for… checks notes… pointing out the ChatGPT summary is full of inaccuracies. Okay, sure - go ahead and put that reissue out there that talks about songs that aren’t even ON THE ALBUM…?!?!

010

Matt Crosslin @grandeped.bsky.social · 30/09/2026

Kids totally change after getting hours a day to goof off on what parents call “life skills”? Unheard of! And there is no way that the CEO of Applications at OpenAI is in any way biased here… 😂

George Siemens claiming that CEO of Applications at OpenAI praising Alpha School is a solid endorsement, but not a carefully planned biased propaganda post.

020

Matt Crosslin @grandeped.bsky.social · 30/09/2026

An anti-science antivaxxer is just admitting that they are going to force Ai to spread misinformation, while education thought leaders are still over on Xitter praising Ai failures like Alpha School.

16627

Matt Crosslin @grandeped.bsky.social · 26/09/2026

“Suddenly you are not recording people.” Actually, you are still recording people… and making people hate cute little animated avatars…

243811

Matt Crosslin @grandeped.bsky.social · 24/09/2026

Definitely more accurate and clear. “Hybrid” doesn’t say who or what is hybridized. “Recognition” is too vague - I can recognize an image or speech from across the room, but still not tell you what either one actually is. “Show” “mistakes” etc are all too vague.

030

Matt Crosslin @grandeped.bsky.social · 16/09/2026

Anything but train employees to follow food safety laws and create processes that make it easier to do so. Which would also be the cheaper route than paying big $$$ for fascism. But why empower employees when you can have an expensive illusion of control?

2442

Matt Crosslin @grandeped.bsky.social · 11/09/2026

Or just say you want no sauce and get 20 things of sauce.

011

Matt Crosslin @grandeped.bsky.social · 11/09/2026

Talarico is pretty much tanking his chances of winning. No Democrat wins in Texas once they throw the trans community under the bus while bragging about GOP endorsements. I’m not saying whether it is right or wrong for this to happen, just that it always does. TX Dems just cannot learn from the past

000

Matt Crosslin @grandeped.bsky.social · 25/08/2026

“ChatGPT, tell me how trade my parents for some that actually care about me?”

020

Matt Crosslin @grandeped.bsky.social · 19/08/2026

That jump he makes from “machine says it will suffer without making art” to assumed personhood is… something. Definitely driven by his bias.

010

Matt Crosslin @grandeped.bsky.social · 18/08/2026

Just when I thought that the term “thought leader” couldn’t get any more useless and annoying at the same time…

041

Matt Crosslin @grandeped.bsky.social · 17/08/2026

You say for the people but then you are against Medicare for all? You be illin’…

010

Matt Crosslin @grandeped.bsky.social · 12/08/2026

(BTW: if Ai can do a good job on your assignments, then stop assigning things that are well-covered everywhere online. "Assessment by regurgitation of the basics" has never really been a good assessment method, and yes Ai can do fairly well on anything that is well organized and reviewed online)

000

Matt Crosslin @grandeped.bsky.social · 12/08/2026

Just because Ai can write a decent paper on some well-covered topic, that doesn't mean it can make sense of entire courses... where there is no rhyme or reason to all the different formats, terminology, assignment instructions, etc.

100

Matt Crosslin @grandeped.bsky.social · 12/08/2026

There is no system out there for organizing and rating these dissertations, much less classify the different ways different disciplines approach dissertations in general. When students feed their dissertation idea into Ai, it spits out well-written smart-sounding nonsense because of this complexity…

110

Matt Crosslin @grandeped.bsky.social · 12/08/2026

When you run agentic Ai and tell it to take a class or build a class (as I have), you just get well-written grammatically correct nonsense. It's the same reason I tell students to not use Ai on dissertations. There are thousands of dissertations online - good, bad, and "never should have passed"…

100

Matt Crosslin @grandeped.bsky.social · 12/08/2026

No one uses the exact same format or even language for structuring a course. Are they objectives or outcomes or competencies or or goals or...? What is the right way to format a rubric? This complexity is confusion for a system just trying to give THE most likely response to a prompt…

110

Matt Crosslin @grandeped.bsky.social · 12/08/2026

Tech bros are desperate for everyone to believe what Ai can do so that they can get more funding to keep the charade going. If you have worked with enough teachers across enough contexts enough, you know that the one thing that will keep any of this from happening is just the complexity of humanity…

100

Matt Crosslin @grandeped.bsky.social · 12/08/2026

Look, when you read an article about what agentic Ai can do (design a course, take a course, etc), look for a link to a source that shows this has actually happened. You usually won't see that, because it hasn't…

121

Matt Crosslin @grandeped.bsky.social · 12/08/2026

Meh. More BS from the NYT. Students turning in AI created assignments find they fail the assignment more often than not, because AI is not interested in passing, just completing a query. I have tested these agents that supposedly can take an entire class. They can’t.

010

Matt Crosslin @grandeped.bsky.social · 06/08/2026

Lawyer types: is it possible for citizens of a state… say, oh I don’t know, Texas… to do a class action lawsuit against their attorney general for being an evil money-wasting complete and utter joke of a human being?

000

Matt Crosslin @grandeped.bsky.social · 30/07/2026

Seeing this article helped me realize this. I think I mentioned this before, but on our first or second date, my girlfriend said "AI? bleck! (pretended to throw up) They keep trying to force us to use that at work, and I just nod at them and then don't."

030

Matt Crosslin @grandeped.bsky.social · 30/07/2026

Might sound weird to some, but back when I was on the dating scene I swiped left on a good friend of mine that would have been a great match for me… except she had developed a weird obsession with Ai. As messed up as the dating scene is… just couldn’t do it.

280

Matt Crosslin @grandeped.bsky.social · 30/07/2026

How long until the Data Center Bros are all saying “this noise isn’t bad if you just don’t pay attention” like they are saying “that’s not much water usage compared to…” and “that’s is not much energy usage…”?

130

Matt Crosslin @grandeped.bsky.social · 22/07/2026

This article has real MTV LeRun Skate-Bike commercial vibes to it… please buy one, we will take anything… even checks! m.youtube.com/watch?v=M6de...

m.youtube.com

MTV LeRun

YouTube video by Random Stuff I Find on VHS

010

Matt Crosslin @grandeped.bsky.social · 01/07/2026

Bluesky was forcing some age verification thing on me. Seems gone for now, but I’m pretty sure it will be back. Not really going to mess with that “privacy breach waiting to happen” nonsense. I’ll probably just be gone from here suddenly at some point.

000

Matt Crosslin @grandeped.bsky.social · 29/06/2026

Giving me some major trauma-flashbacks here! Surprised anyone remembers the dual-layer MOOC thing. Their website is still up - not sure if it is a zombie site or still active, though: prosolo.ca

prosolo.ca

ProSolo – Competency-based Learning Platform

210

Matt Crosslin @grandeped.bsky.social · 20/06/2026

Keep running into small pockets of Talarico supporters that are trying so hard to manufacture some kind of “disrespect” by Crockett towards Talarico. They are pulling junk out of thin air. But it reeks of trying to set her up to take the blame if he loses. If that is you… grow up.

000

Matt Crosslin @grandeped.bsky.social · 18/06/2026

Oh, yay. Especially since the current president has such a great track record at nominating people that actually help the American people… what could go wrong?

070

Matt Crosslin @grandeped.bsky.social · 07/06/2026

Didn’t they also invent 911 because people couldn’t even memorize three emergency numbers?

000

Matt Crosslin @grandeped.bsky.social · 06/06/2026

You maybe memorized your home phone and your parents work phone. Maybe a good friend or significant other if you called enough. Not to mention there were a lot fewer numbers before cellphones took off. One number covered an entire household.

110

Matt Crosslin @grandeped.bsky.social · 06/06/2026

As much as I dislike Ai… who were all these people memorizing phone numbers? No one I knew. We all had a paper taped to the wall with phone numbers. They used to sell all kinds of book things to write numbers in and store them next to, under, etc phones. They sold those by the millions…

331

Matt Crosslin @grandeped.bsky.social · 03/06/2026

“Frontier” makes me think of Oregon Trail and everyone dying. Not a great image there Mayo.

080

Matt Crosslin @grandeped.bsky.social · 29/05/2026

This is where I wish authors would use better language with Ai. Ai doesn’t “believe” anything, nor does it have concepts of true or false. It just completes the pattern in whatever fashion its algorithm deems the best outcome.

010

Matt Crosslin @grandeped.bsky.social · 27/05/2026

“Observably organized?” Only one way to do that. These people keep saying they want biological checkpoints at bathroom and locker room entrances. Trans people just want to go use the bathroom or locker room. Only one of these groups are perverts, and it’s NOT those that just want to pee in peace.

010

Matt Crosslin @grandeped.bsky.social · 26/05/2026

How fast will that “the gains of Ai will be shared globally” part turn into “we gave you slop, now give us your water”?

030

Matt Crosslin @grandeped.bsky.social · 21/05/2026

Wow. Way to read the room on how much Ai sucks.

110

Matt Crosslin @grandeped.bsky.social · 17/05/2026

So glad Ai is so awesome. I really love spending twice as long grading because I have to describe twice as many things that ChatGPT got wrong. Well, it’s all factually correct, but wrong for the assignment. I can already tell you what is in the Ai Declaration statement before I even read it.

000

Matt Crosslin @grandeped.bsky.social · 11/05/2026

But hey Canvas went down for a day so obviously online is the. worst. ever. and now time to brag about how analog is the coolest way to go… even though all that paper stuff was technically digitally created on the same devices that connect to the LMS.

020