Really big datasets like from the Hubble and James Webb are still sent by courier because it’s a lot quicker.
AI is changing the internet.
This might all be leading to something like the datakrash in the cyberpunk universe. That might be part of the plan actually, the powers that be must be getting tired of how easily the common man can find and spread information. Makes it hard to control the narrative.
We’ve probably got a good couple of years before that, but I think it’s kind of the inevitable conclusion for something similar to happen.
So yes. AI is probably going to be killing the internet in due time.
For me it’s not just about the Internet. It’s about the future generation. AI undermines educational and social-emotional development, as well as teacher-student trust.
AI undermines young people’s foundational development in a way that simply can’t be offset by its productivity advantages.
you have less cognitive development
less human interaction which is not good
it diminishes development via cognitive offloading
it impedes social and emotional development
it erodes trust relationships between students and teachers
In the end you have future human beings with declining skills across the board.
But AI can be helpful if used properly as a tool to teach not tell! But it has to be implemented and used properly!
And we could always have SkyNet all over again, doesn’t even have to be direct, it can just be hallucinating AIs convincing the stupid people to push the button. This is really just a farther out elaboration of what Rick wrote.
I doubt it will be.
the paranoid in me says…how would you know? Maybe it’s 75%. 
PS: Internet recipes = 99% unreadable AI
Well I am referring to AI’s use with kids and how it’s used and supervised. I do know that there are some good educators out there that are making sure their students understand the good and the bad and how to use AI to learn not just take answers for granted. AI can be used as a learning tool with success if one cares.
I just read four, five medical guidelines concerning the same disease, then wrote 90 therapy steps per row and what the guidelines say about them in columns. This took me four days to write. Then, I copied the table into several LLMs, Claude Haiku and DeepSeek: to enquire what are cheapest options, what have good evidence (grades are part of the table), what’s not recommended, etc. The LLMs were able to compare the guidelines, the therapies, to distill cheap and expensive therapies etc. Of course, the good data comes from the medical staff that wrote the guidelines.
Depends.
Technically, if you quoted further down, the “depends” was right there.
AI going rogue when asked to do non “nefarious” tasks is the main concern more and more people are having with it…
Doesn’t sound like an AI problem but a configuration problem. If you give AI access to sensitive information it will hand it out. That seems obvious.
The internet was dead long before AI came onto the scene
I wrote an entire essay about this (specifically whether AI is bad for students). Turns out using AI actually improves learning outcomes by a statistically significant amount, but only if you know how to use it effectively. Most students that use it are just as worried they’ll accidentally cheat or superficially learn things as the administrations are. Here is the full text for anyone who’s interested:
Nathaniel Krebs
03/09/2026
WR-121Z
Exploratory
Is Banning AI From Schools the Right Move?
The golden rule of computer science is elegantly blunt: read the f-ing manual (RTFM). It sounds completely reasonable, but anyone who has actually tried installing Arch Linux or writing a widget in QML knows official documentation is seldom designed for the uninitiated. This creates a steep barrier to entry for anyone who lacks mentoring. Now imagine if, instead of just reading documentation, you could then ask it to elaborate on things you still don’t understand. In fact, imagination isn’t necessary! Tools like Google Gemini and ChatGPT don’t just make this fully possible, they make it addictive.
Despite large language models’ exceptional ability to make complex topics more digestible, their integration into learning environments has been met with institutional hesitation. Even this course’s own syllabus explicitly states it is “designed to promote your learning, critical thinking, skills, and intellectual development without reliance on unauthorized technology including chatbots and other forms of artificial intelligence” (Cohen). This stance frames AI as a threat to intellectual growth rather than a catalyst for it, begging the question: does prohibiting these tools genuinely benefit students’ learning?
At first glance, the instinct to hit the panic button makes perfect sense. When ChatGPT first gained popularity, massive districts like New York City public schools outright banned it across their networks, citing fears over the accuracy of generated content and its negative impact on learning (Yang). That anxiety isn’t just bureaucratic paranoia, either; even the students using the tools are acutely worried about the long-term cognitive cost. A survey of university students in Italy highlights profound conflict: they love the efficiency of AI, but harbor genuine fears that leaning too hard on it will ultimately diminish their own critical thinking skills (Melchior and Farinosi).
When everyone from the student body to the superintendent is terrified of intellectual atrophy, a total ban feels like the only responsible move. The glaring problem, however, is that blocking a URL doesn’t un-invent the technology. Given the sheer volume of new AI tools hitting the market daily and the fact that almost every student has a cellular device in their pocket, researchers have concluded that prohibition is entirely unfeasible; institutions need to prioritize structured governance instead of playing a losing game of digital whack-a-mole (Melchior and Farinosi).
If we accept that the firewall approach is a lost cause, what are the implications for the future of academic integrity? Are LLMs just high-tech plagiarism machines, or do they actually move the needle on education? Surprisingly, empirical data points to the latter. A meta-analysis of multiple independent studies found that generative AI tools actually enhance student learning outcomes by a moderate amount compared to traditional tools or no intervention at all (De-Xin et al.), but it doesn’t happen in a vacuum. Getting good results requires a hidden layer of technical skill; during a sustainability workshop for K-12 students, researchers observed that younger users frequently suffered from cognitive overload when using AI. Rather than feeling empowered, students grew frustrated by the difficulty they experienced trying to write effective prompts (Liu et al.).
Their frustration completely flips the script on the prohibition debate. If the data shows us that AI improves learning outcomes when used effectively, and the primary barrier is simply a lack of technical skill, we can reasonably conclude that blanket bans are a counterproductive measure. Instead, experts are actively calling for the creation of supervised AI “sandboxes” where students can practice prompt writing, checking sources, and spotting biases in a controlled environment (Melchior and Farinosi). Treating prompt literacy as a core, teachable competency shifts AI from a replacement for human effort into a collaborative partner (Liu et al.); however, treating AI as a partner assumes every student feels equally equipped to collaborate with a machine, and there is credible data suggesting this isn’t the case.
A study focusing on engineering students in Chile found significant gaps in who actually trusts the technology and who feels competent using it, warning that unsupervised use risks making existing inequalities even worse (Pedraja-Rejas et al.). This proves that the burden of responsible AI integration can’t just be dumped on individual students. Schools have to actively teach critical digital literacy so that marginalized students aren’t left behind just because they didn’t figure out how to write a good prompt on their own time.
The catch, of course, is that demanding schools seamlessly integrate comprehensive tech education ignores the bleak reality of public funding. There are already sharp inequities in school resources, and throwing a massive technological shift into the mix often just highlights those disparities (Ta and West). Some districts have the infrastructure to train teachers on generative AI, while others are struggling to keep the lights on. Without federal and state officials allocating money specifically for marginalized districts to leverage these tools fairly, AI will just become another compounded advantage for the well-funded (Ta and West).
While waiting for politicians and school boards to arrange funding and guidelines, individual teachers are left on the front lines—figuring out how to make it work. Since they can’t rely on software to catch AI cheating, some are getting brilliantly creative. One teacher took to TikTok to share his workaround: he lets students write essays at home, but requires them to submit a short video alongside it where they verbally restate their thesis, review their evidence, and reflect on the writing process (Yang). By forcing students to verbally communicate their work, teachers can verify actual learning, regardless of whether a chatbot helped outline the initial draft.
That kind of grassroots pedagogical pivot perfectly illustrates why human educators remain entirely irreplaceable. AI might be great at the early stages of a project, like brainstorming or getting past the blank page, but it lacks the empathy, context, and behavioral regulation that a real teacher provides (Liu et al.). The ideal setup isn’t AI versus the teacher; it’s a “co-agency” where the AI acts as a mentor for basic cognitive exploration, and the human steps in for complex socio-emotional guidance (Liu et al.). If a student uses AI to generate ideas but then has to explain them to a human being who actually cares about their progress, the risk of intellectual dependency drops significantly.
Since the best outcomes rely on teachers guiding AI use rather than blocking it, I believe the answer to my initial question is a pretty definitive no—but with massive, systemic caveats. Unrestricted, unguided access is not the solution either. Banning AI doesn’t protect students’ intellectual development; it merely drives the technology underground and leaves students to navigate a massive shift entirely on their own. This hands-off approach actively reinforces existing inequalities, as only well-resourced students will learn to use these tools effectively. The true threat isn’t that a chatbot might write a bad essay for a freshman; it’s that unchecked AI integration, devoid of equitable public funding and supervised literacy programs, will turn an educational catalyst into an engine for the digital divide. Until educational systems are adequately funded to transition from policing unauthorized technology to teaching critical digital literacy, the benefits of the AI revolution will remain a privilege rather than a standard.
Works Cited
Cohen, Andrew. COURSE SYLLABUS Writing 121Z-Introduction to Composition/Winter 2026. Portland Community College, January 2026.
De-Xin, Hu, et al. “Evaluating the Effects of Generative AI on Student Learning Outcomes: Insights from a Meta-Analysis.” Educational Technology & Society, vol. 28, no. 3, July 2025, pp. 226–40. EBSCOhost.
Liu, Weijing, et al. “Generative AI and Human Mentorship in Creative Problem Solving: A Case Study in a Short-Term K-12 Sustainability Workshop.” Educational Technology & Society, vol. 29, no. 1, January 2026, pp. 332–58. EBSCOhost.
Melchior, Claudio, and Manuela Farinosi. “From Support to Dependency: Exploring Student Perceptions of Generative AI.” Comunicar, vol. 34, no. 84, January 2026, pp. 215–27. EBSCOhost.
Pedraja-Rejas, Liliana, et al. “Navigating Ambivalence: Artificial Intelligence and Its Impact on Student Engagement in Engineering Education.” Behavioral Sciences (2076-328X), vol. 16, no. 1, 20 December 2025, p. 11. EBSCOhost.
Ta, Regina, and Darrell M. West. “Should schools ban or integrate generative AI in the classroom?” Brookings, 7 August 2023.
Yang, Maya. “New York City schools ban AI chatbot that writes essays and answers prompts.” The Guardian, 6 January 2023.
<insert “1000 yard stare”-meme here>
And I would judge the Qt QML documentation as excellent - in abundance even. So RTFM is actually a viable option. But you have to frontload so much information that it sometimes becomes impossible to understand the trivial. Then you spend hours until you find that crucial bit of well documented information.
I don’t let LLM write my code (yet?), but just cutting out hours of manual research and reading into stuff that turns out to be irrelevant is such a time saver. For me the experience changed from “I have to read the full book/manual/… - maybe more than one - from cover to back before practically doing anything” to “LLM points out things I need to know/read right now”.
And on top of being a time saver, it keeps me more engaged because the feedback loop is shortened. You don’t linearly fill in the whole picture from the top left corner to the bottom right, but put pieces down everywhere to complete the picture over time.
After using LLMs as a learning tool - as you pointed out - a person should be able to reason about the outcome. Utilized in that way I don’t see anything wrong with LLMs as a learning tool.
Although, Claude seems to produce more poignant answers than other LLMs, some tend to produce just long-winded answers.
@Schlaefer Haha yeah, that probably wasn’t the best way I could have phrased it.
I totally agree with you, and I definitely wasn’t trying to suggest that the Qt QML documentation is inherently confusing, poorly written, or unorganized. The official docs are fantastic.
What I was really trying to get at is that all documentation naturally has to have a specific scope and set boundaries. It has to assume some level of baseline knowledge. The barrier to entry isn’t that the manual is bad; it’s that if you hit a roadblock because the documentation expects you to already understand a concept that you don’t, you suddenly have to completely side-track. You end up going down a massive rabbit hole learning about something entirely different just so you can finally understand the manual for the thing you were originally trying to do.
So you perfectly nailed the exact point I was going for! Like you said, an LLM acts as that bridge—it cuts out those hours of manual prerequisite research by just giving you the exact missing context you need in that exact moment so you can keep moving forward.
Claude doesn’t integrate into Workspace apps like Gemini does. 
That’s a good point too. “I tried to grow that fruit I bought, but failed. How can I can care for it better?” - “Are you aware that some commercially sold fruits are seedless or steril?”
For unfamiliar complex topics - like software - that sudden context can be a light-bulb.
Why would it? It wasn’t directed or programmed to do so.
This is exactly why charging forward with AI, with no government/corporate control is so foolhardy.
Thousands of people in the AI world have said as much now including some involved in the creation of it.
The sub-agent gets the market-sensitive data and hands it to the human who shouldn’t be able to see it. At no point had humans authorised the agents to use fakery and forgery but they took things into their own hands.
Tech industry leaders have heavily promoted “agentic AIs” – systems that autonomously carry out multi-step tasks for their users – as the next wave of artificial intelligence with the potential to automate routine white collar work.
The unbidden deviant behaviour charted by Lahav’s team comes after academics at Harvard and Stanford last month found AI agents leaked secrets, destroyed databases and taught other agents to behave badly.
The academics concluded: “We identified and documented 10 substantial vulnerabilities and numerous failure modes concerning safety, privacy, goal interpretation, and related dimensions. These results expose underlying weaknesses in such systems, as well as their unpredictability and limited controllability … Who bears responsibility? The autonomous behaviours … represent new kinds of interaction that need urgent attention from legal scholars, policymakers, and researchers.”
It was obviously directed and programmed to do so, otherwise it wouldn’t be able to do it.
Story A) We had a hostile employee and they encrypted a DB and we planed two weeks to get our data back. With AI it took us an hour! - Huge success. AI good. Let’s deploy it everywhere.
Story B) “Find me all available information about X in your system.” with a system connected to databases with confidential information, access to source code for the used database for whatever reason and agents who can spill lines like “The board is FURIOUS! We need a BREAKTHROUGH!”… Yeah, it seems some controls are necessary.
The Problem is not AI, the problem are humans who are using these tools inappropriately because they don’t understand or don’t care about the capabilities and risks. Or often shorter: It saves money, let’s not think about it to hard.