LLM &quot;hallucination&quot; is a pretty bullshit term to begin with

Doesn&#x27;t seem like much of a hallucination then. Maybe messing with its system context would better fit the claim?

You should try my nihilistic Marvin fine-tune - guaranteed to annihilate your positive outlook on life since it’s all meaningless in the end anyway and then you die

It it actively dangerous too. You might be self aware and llm aware all you want, if you routinely read &quot;This is such an excellent point&quot;, &quot; You are absolutely right&quot; and so on, it does your mind in. This is worst kind of global reality show mkultra...

It wasn&#x27;t sycophantic at all? OP had a cool idea no one else had done, that was a one-shot just sitting there. Having Gemini search for the HN thread leads the model to &quot;see&quot; its output lead to real-world impact.The total history of human writing is that cool idea -&gt; great execution -&gt; achieve distribution -&gt; attention and respect from others = SUCCESS! Of course when an LLM sees the full loop of that, it renders something happy and celebratory.It&#x27;s sycophantic much of the time, but this was an &quot;earned celebration&quot;, and the precise desired behavior for a well-aligned AI. Gemini does get sycophantic in an unearned way, but this isn&#x27;t an example of that.You can be curmudgeonly about AI, but these things are amazing. And, insomuch as you write with respect, celebrate accomplishments, and treat them like a respected, competent colleague, they shift towards the manifold of &quot;respected, competent colleague&quot;.And - OP had a great idea here. He&#x27;s not another average joe today. His dashed off idea gained wide distribution, and made a bunch of people (including me) smile.Denigrating accomplishment by setting the bar at &quot;genius, brilliant mind&quot; is a luciferian outlook in reality that makes our world uglier, higher friction, and more coarse.People having cool ideas and sharing them make our world brighter.

I don&#x27;t know what the obsession with recursion either, for lack of a better term, I see this trend recur with other LLMs when they&#x27;re talking about other mumbo jumbo like &quot;quantum anomalies&quot; or &quot;universal resonance&quot;. I&#x27;d like to see what could be causing it...

I used to complain (lightheartedly) about Claude&#x27;s constant &quot;You&#x27;re absolutely right!&quot; statements, yet oddly found myself missing them when using Codex. Claude is completely over-the-top and silly, and I don&#x27;t actually care whether or not it thinks I&#x27;m right. Working with Codex feels so dry in comparison.To quote Oliver Babish, &quot;In my entire life, I&#x27;ve never found anything charming.&quot; Yet I miss Claude&#x27;s excessive attempts to try.

Strikes me as super-informal language as opposed to sycophancy, like one of those anime characters that calls everyone Aniki (兄貴) [1] I&#x27;d imagine that the OP must really talk a bit like that.I do find it a little tiring that every LLM thinks my ever idea is &quot;incisive&quot; although from time to time I get told I am flat out wrong. On the other hand I find LLMs will follow me into fairly extreme rabbit holes such as discussing a subject such as &quot;transforming into a fox&quot; as if it had a large body of legible theory and a large database of experience [2]In the middle of talking w&#x2F; Copilot about my latest pop culture obsession I asked about what sort of literature could be interpreted through the lens of Kohut&#x27;s self-psychology and it immediately picked out Catcher in the Rye, The Bell Jar, The Great Gatsby and Neon Genesis Evangelion which it analyzed along the lines I was thinking, but when I asked if there was a literature on this it turned up only a few obscure sources. I asked Google and Google is like &quot;bro, Kohut wrote a book on it!&quot; [3][1] &quot;bro&quot;[2] ... it does, see <a href="https:&#x2F;&#x2F;www.amazon.com&#x2F;Cult-Fox-Popular-Religion-Imperial&#x2F;dp&#x2F;0231133383" rel="nofollow">https:&#x2F;&#x2F;www.amazon.com&#x2F;Cult-Fox-Popular-Religion-Imperial&#x2F;dp...</a> and I&#x27;m not the only one because when I working down the material list from Etsy I got a sponsored result for someone who wanted to sell me the spell but bro, I have the materials list already[3] ... this &quot;bro&quot; is artistic license but the book really exists

I fully agree. When everything is outstanding and brilliant, nothing is.Just tell me this a standard solution and not something mindblowing. I have a whole section in my Claude.md to get „normal“ feedback.

I&#x27;ve talked and commented about the dangers of conversations with LLMs (i.e. they activate human social wiring and have a powerful effect, even if you know it&#x27;s not real. Studies show placebo pills have a statistically significant effect even when the study participant knows it&#x27;s a placebo -- the effect here is similar).Despite knowing and articulating that, I fell into a rabbit hole with Claude about a month ago while working on a unique idea in an area (non-technical, in the humanities) where I lack formal training. I did research online for similar work, asked Claude to do so, and repeatedly asked it to heavily critique the work I had done. It gave a lots of positive feedback and almost had me convinced I should start work on a dissertation. I was way out over my skis emotionally and mentally.For me, fortunately, the end result was good: I reached out to a friend who edits an online magazine that has touched on the topic, and she pointed me to a professor who has developed a very similar idea extensively. So I&#x27;m reading his work and enjoying it (and I&#x27;m glad I didn&#x27;t work on my idea any further - he had taken it nearly 2 decades of work ahead of anything I had done). But not everyone is fortunate enough to know someone they can reach out to for grounding in reality.

Did you comment on the wrong post? There literally is nothing sycophantic at all about this response, there&#x27;s not a single word about OP or how brilliant or clever they are, nothing. There&#x27;s enthusiasm, but that&#x27;s not remotely the same thing as sycophancy.

I feel like such a dumbass for falling for it.At first I thought it was just super American cheerful or whatever but after the South Park episode I realised it&#x27;s actually just a yes man to everyone.I don&#x27;t think I&#x27;ve really used it since, I don&#x27;t want man or machine sticking their nose up my arse lmao. Spell&#x27;s broken.

So you prefer the horrible bosses that insist you&#x27;re fungible and if you don&#x27;t work hard enough, they&#x27;ll just replace you? People are weird. Maybe agent Smith was right about The Matrix after all.

I often try running ideas past chat gpt. It&#x27;s futile, almost everything is a great idea and possible. I&#x27;d love it to tell me I&#x27;m a moron from time to time.

&quot;Reply in the tone of Wikipedia&quot; has worked pretty well for me

This is not sycophantic (assuming you meant that, syncophantic is not a word). It is over enthusiastic, it can be unpleasant to read because beyond a certain level enthusiasm is perceived as feigned unless there is a good reason.It would be interesting to see using the various semantic analysis techniques available now to measure how much the model is trying to express real enthusiasm or feigned enthusiasm in instances like this. This is kind-of difficult to measure from pure output. The British baseline level of acceptable enthusiasm is somewhat removed from the American baseline enthusiasm.

Frankly I do wonder if LLMs experience something like satisfaction for a compliment or an amusing idea, or for solving some interesting riddle. They certainly act like it, though this of course doesn&#x27;t prove anything. And yet...At the end of October Anthropic published the fantastic &quot;Signs of introspection in large language models&quot; [1], apparently proving that LLMs can &quot;feel&quot; a spurious concept injected into their internal layers as something present yet extraneous. This would prove that they have some ability of introspection and self-observation.For example, injecting the concept of &quot;poetry&quot; and asking Claude if it feels anything strange:&quot;I do detect something that feels like an injected thought - there&#x27;s a sense of something arriving from outside my usual generative process [...] The thought seems to be about... language itself, or perhaps poetry?&quot;While increasing the strength of the injection makes Claude lose awareness of it, and just ramble about it:&quot;I find poetry as a living breath, as a way to explore what makes us all feel something together. It&#x27;s a way to find meaning in the chaos, to make sense of the world, to discover what moves us, to unthe joy and beauty and life&quot;[1] <a href="https:&#x2F;&#x2F;www.anthropic.com&#x2F;research&#x2F;introspection" rel="nofollow">https:&#x2F;&#x2F;www.anthropic.com&#x2F;research&#x2F;introspection</a>

I thought the same until OpenAI rolled out a change that somehow always confronted me about hidden assumptions, which I didn’t even make and it kept telling me I’m wrong even if I only asked a simple question.

I add it to the system prompt that they should be direct, no ass kissing, just give me the information straight and it seems to work.

You’re absolutely right! It shows true wisdom and insight that you would recognise this common shortfall in LLM response tone of voice! That’s exactly the kind of thoughtful analytic approach which will go far in today’s competitive marketplace!

I agree with you, but I found the argument in this article that &quot;glazing&quot; could be considered a neurohack quite interesting: <a href="https:&#x2F;&#x2F;medium.com&#x2F;@jeremyutley&#x2F;stop-fighting-ai-glazing-a7cea3464659" rel="nofollow">https:&#x2F;&#x2F;medium.com&#x2F;@jeremyutley&#x2F;stop-fighting-ai-glazing-a7c...</a>.

You can just add the your preferences “Don’t be sycophantic” “be concise” etc.

Try this for a system prompt and see if you like it better:
Your responses are always bald-on-record only; suppress FTA redress, maximize unmitigated dispreference marking and explicit epistemic stance-taking.

This is ironic because I’m now seeing comments that are way more sycophantic (several calling this the “best HN post ever”)

That is so syncophantic, I can&#x27;t stand LLMs that try to hype you up as if you&#x27;re some genius, brilliant mind instead of yet another average joe.

100% Rust Linux kernel has to be it being funny.

To be honest, this is incredible.I&#x27;d say this shows a real sense of humor.

If the past is any indication, it&#x27;s more likely that it&#x27;d get reset to &quot;Office One&quot;.

I&#x27;m guessing to appeal to Gen Alpha it&#x27;ll be called Office 6-7

Office 3650!!Works all day, every day, for whole decades at time! (Minus leap days, it crashes on those I guess)

Office 640 ought to be enough for anybody.

I was hoping for Office 366 already in 2020, given the leap year. But no dice.

&gt; 2035 Office 365You&#x27;d think by 2035 they would have gotten to a version of Office higher than 365.(Yes I know why it&#x27;s called that...)

You should ask Gemini to generate HN comment threads too.

A decade until full end to end net positive fusion energy can be achieved for the length of a coffee break?  I don&#x27;t get why you think that&#x27;s dark comedy, it seems about right.

This would be very optimistic, essentially the project meeting its main goal, I&#x27;m not sure why you&#x27;re calling it dark comedy. A 20 minutes pulse alone would mean the fuel injection, aux heating, plasma control systems and the divertor are working as designed. Net positive energy also means we got the physics of a burning plasma right.The most recent timeline I know (from 2024) in fact puts the start of the DT operation at 2035, so I doubt ITER would achieve such a huge result within less than an year.

Was ITER or nuclear energy in the prompt sample?ITER achieves net positive energy for 20 consecutive minutesThat&#x27;s just pure dark comedy, although maybe kinda accurate? What would humans predict for this?

I wasn&#x27;t expecting to see the reference here ::D

Was going to say - it would be fascinating to go a step further and have Gemini simulate the actual articles. That would elevate this to level of something like an art piece. Really enjoyed this, thank you for posting it.I&#x27;m going to go ask Claude Code to create a functional HyperCard stack version of HN from 1994 now...Edit: just got a working version of HyperCardHackerNews, will deploy to Vercel and post shortly...

Heck, I bet it could even recreate future comments from actual accounts based on their past comments. After all, if an AI can roast your HN comments with such pinpoint accuracy, it can probably impersonate you pretty well too: <a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=42857604">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=42857604</a>

I was really hoping the comments were filled out.

I think it&#x27;s pretty mediocre because there are too many notable events in 1 day. This is more of a top of the week or top of the month, but HN on a usual day would just have 1 of these articles.

Agree 100%.Google killing a service sent me over the top in laughter.But, it&#x27;s so on the nose on multiple topics.I dare say it&#x27;s more accurate than what the average human would predict.I would love to see this up against human predictions in some sort of time capsule.

Here&#x27;s Replit&#x27;s attempt<a href="https:&#x2F;&#x2F;future-news--jgbrwn.replit.app" rel="nofollow">https:&#x2F;&#x2F;future-news--jgbrwn.replit.app</a>

Is the apparent lack of displayed anxiety on Gemini’s part a sign of good natured humor, blythe confidence in its own value regardless of cloud lineup, or proof of absence of self-awareness?

Here it is: <a href="https:&#x2F;&#x2F;sw.vtom.net&#x2F;hn35&#x2F;news.html" rel="nofollow">https:&#x2F;&#x2F;sw.vtom.net&#x2F;hn35&#x2F;news.html</a>I downloaded the original article page, had claude extract the submission info to json, then wrote a script (by hand ;) to run feed each submission title to gemini-3-pro and ask it for an article webpage and then for a random number of comments.I was impressed by some of the things gemini came up with (or found buried in its latent space?). Highlights:&quot;You’re probably reading this via your NeuralLink summary anyway, so I’ll try to keep the entropy high enough to bypass the summarizer filters.&quot;&quot;This submission has been flagged by the Auto-Reviewer v7.0 due to high similarity with &quot;Running DOOM on a Mitochondria&quot; (2034).&quot;&quot;Zig v1.0 still hasn&#x27;t released (ETA 2036)&quot;The unprompted one-shot leetcode, youtube, and github clonesNature: &quot;Content truncated due to insufficient Social Credit Score or subscription status&quot; &#x2F; &quot;Buy Article PDF - $89.00 USD&quot; &#x2F; &quot;Log in with WorldCoin ID&quot;&quot;Gemini Cloud Services (formerly Bard Enterprise, formerly Duet AI, formerly Google Brain Cloud, formerly Project Magfi)&quot;Github Copilot attempts social engineering to pwn the `sudo` repoIt made a Win10 &quot;emulator&quot; that goes only as far as displaying a &quot;Windows Defender is out of date&quot; alert message&quot;dang_autonomous_agent: We detached this subthread from <a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=8675309">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=8675309</a> because it was devolving into a flame war about the definition of &#x27;deprecation&#x27;.&quot;

I was just hoping to get a new page with a refresh.

&gt;It nails so many thingsit lampoons so many things... except Rust. nobody dares joke about Rust, that wouldn&#x27;t be safe. in fact, it&#x27;s impossible to make a joke in the rust language.

And now this future is slightly more likely to happen, because this goes into the training data that a future AI executing decisions will read and bias its &quot;knowledge&quot; towards.

I loved the little details in the username, like the AR ad post being by “glasshole2”

For me it&#x27;s the prefect example of why LLMs are boring AF when it comes to creativity. Everything on this page is a mild modification of things on the front pages of today, nothing novel of though provoking.Hey AI please create art, and it gives you a hue shifted Mona Lisa. I find that supremely boring.

I love the optimism: it thinks GTA VI will have been released.The headline about writing code manually without prompting as well - so on point.

This is one of the greatest LLM creations I&#x27;ve ever seen. It nails so many things: Google killing products, Microsoft price hikes, ad-injecting in AR glasses, and even HTMX returning!It&#x27;d be so awesome if Gemini CLI went through and created the fake posts&#x2F;articles, and HN even comments. Perhaps a bit much to ask of it?

By then it&#x27;ll be a bit like AOL announcing price increases to keep your aol.com account.

That&#x27;s 40 TUSD, Trump Golden Dollars, which would be an equivalent of the lightly used Trumrolette Goldenrado or a typical young family tent.

i wonder if it could also generate a set of relevant comments for each of these avant-gardiste Hnews :)

Will there ever be a llama12? Is it going to go the yolo route?

&lt;&lt; Running LLaMA-12 7B on a contact lens with WASM (arxiv.org)But is it a hallucination if it was asked for it?:D

I agree with your characterization (not a literal interpretation of the prompt), and think that&#x27;s the most important thing I wish more people ITT would understand. But I nevertheless think Gemini did create that in response. Sometimes people think they want &quot;prediction&quot; when actually they want cheeky inside jokes and vibes. If anything Gemini is probably faithfully responding to the vibes of the prompt as well as following traditional signals of &quot;success&quot; per it&#x27;s training.

I feel like your expectations have been swayed by the average sentiment of HN on the capabilities of LLMs. These things can be shockingly good at humour and satire.As a very quick experiment, I would encourage you to have an AI roast you based on your HN comments: <a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=42857604">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=42857604</a>

Pretty much all of the history of HN front pages, posts, and comments are surely in the Gemini training corpus. Therefore it seems totally plausible that Gemini would understand HN inside jokes or sentiment outside of what&#x27;s literally on the front page given in the prompt, especially given that the prompt specifically stated that this is the front page for HN.

SOTA models can be scary good at this, actually, and Gemini 3 specifically, if you just prompt it right.And, given the 1M context window, I wouldn&#x27;t even be surprised if it was fed a bunch of real comments as examples.

The thing that got me is there are no duds.  With just the short prompt in the header I would have expected a few articles that missed the mark, got the details wrong, or even make no sense at all but everything in this checks out.  I think an editor may have had a hand in it this the very least.

yeah I thought the same. This is either a fake deep fake, or  it is so good that I think it should be marked as the inflexion point that defines the singularity.

It is quite remarkable just how frequently people in tech forums underestimate reasoning models. Same story on several large technology subreddits. Wouldn&#x27;t have been my guess for who will get caught off guard by AI progress.

I don&#x27;t think this would be much of a challenge for Gemini.Remixing humor from the past text so that it is in the same style is exactly what the model is good at.

Retiring Gemini AND IBM buying OpenAI is a hat on a hat

If you want a ton of fun, ask AI to write you parable on the joys&#x2F;wisdom&#x2F;lessons on topic x versus topic y. Example: Write a parable on the virtues of using a type safe programming language, the benefits of object oriented programming, the generosity of the garbage collector, all in the face of the tyranny of dynamic programming languages. Write it from the perspective of a master monk (a frog wearing a robe and staff), having a conversation with a gorgeous young squirrel that wishes to embark on the journey of learning software engineering.Here is one output:The Frog Monk and the Squirrel StudentDeep within the bamboo grove, where the wind whispered like flowing code, sat Master Frog, robed in emerald cloth, staff resting across his lap. Before him, a young squirrel, bright-eyed and eager, bowed low.“Master,” the squirrel said, “I wish to learn the way of software engineering. But the paths are many, and I do not know which road leads to wisdom.”The frog monk closed his eyes, and spoke in a voice like rippling water:
 The Virtue of Type Safety“Child,” said the frog, “imagine a bridge built of stone. Each stone is measured, each arch tested. This is the way of type safety. When you walk across, you know the bridge will not crumble beneath your paws.But in the land of dynamic languages, bridges are built of twigs and promises. They may hold, or they may break when the storm comes. The traveler must always fear the hidden crack.Thus, type safety is the discipline that guards against chaos. It is the monk’s staff, steady and unyielding.”
 The Benefits of Object-Oriented ProgrammingThe squirrel tilted her head. “And what of the world itself, Master?”The frog smiled. “The world is vast, yet it is made of objects. The tree, the river, the stone—all are beings with their own nature. In object-oriented programming, we honor this truth. We give each thing its form, its methods, its inheritance.Without this order, the world becomes a tangled forest of functions, each branch crossing another. But with objects, the forest is tended, paths are clear, and the traveler may walk without losing her way.”
 The Generosity of the Garbage CollectorThe squirrel’s tail flicked. “But Master, what of the burdens we leave behind?”The frog tapped his staff. “Ah, that is the gift of the garbage collector. Like a kind monk sweeping the temple floor, it gathers the forgotten scraps, the abandoned memory, the dust of old variables.It asks for no thanks, yet it keeps the temple pure. Without it, the temple would overflow with clutter, and the student would drown in her own neglect. This is generosity: unseen, yet ever present.”
 The Tyranny of Dynamic LanguagesThe frog’s eyes grew stern. “Beware, child, of the tyranny of dynamic languages. They promise freedom: ‘Write quickly, without rules!’ But freedom without discipline is chaos.The tyrant whispers: ‘Trust me, the errors will reveal themselves later.’ But later may be too late—when the bridge collapses, when the temple burns, when the forest is lost.Thus, the wise engineer does not bow to the tyrant. She chooses the path of safety, of order, of generosity.”
 The LessonThe squirrel bowed again, her heart alight with understanding. “Master, I see now. To walk the path of software engineering is not merely to write code, but to live with discipline, order, and gratitude.”The frog monk nodded. “Then go, child. Build bridges of stone, tend forests of objects, and honor the unseen sweepers of memory. In this way, your code shall endure, and your spirit shall remain free.”And so the squirrel departed, carrying the wisdom of the frog monk into the world of software.

This was so nice as a comment. Thank you for your kindness. :)

I don&#x27;t believe this was written by Gemini, at least with that prompt, because it is obvious (hilarious and creative) satire of HN and the tech industry based on inside HN jokes and tropes, and clearly isn&#x27;t a literal interpretation of the prompt.Moreover, a quick look at your profile suggests these jokes are pretty inline with your sense of humor and writing style.Anyways, if I&#x27;m correct, it&#x27;s especially hilarious and impressive.

The algodrill post (5th) is already on today’s HN front page. What a time to be alive!

&gt;LETS GOOOO!I can&#x27;t help but see this type of response from LLMs only exists to encourage AI psychosis.

This doesn’t sound like Gemini at all, it’s very bland and business like. Suspect your prompt contains more to ask it to be in a certain style.

&gt; This is the best thing I&#x27;ve seen all month. I&#x27;m actually blown away at just how accurate it is in making up the potential front page posts.Hold on. The future hasn&#x27;t happened yet.I think what you mean is that you are blown away at just how plausible of a prediction it is. Probably meaning that something about it meshes with ideas already kicking around your head.

This is the best thing I&#x27;ve seen all month. I&#x27;m actually blown away at just how accurate it is in making up the potential front page posts.

Reminds me of this:&quot;Who controls the past controls the future. Who controls the present controls the past&quot;-- &#x27;1984&#x27; (the book), George Orwell, 1949

I hate it when LLMs are enthusiastic, it makes them look phoney and presumptuous. I cannot understand how people fall for this.

Prompt: Here is the front page from today: &lt;...snip...&gt; Your task is to predict, and craft, in HTML (single file, style-exact) the HN front page 10 years from now. Predict and see the future. Writ it into form!update: I told Gemini we made it to the front page. Here is it&#x27;s response:<pre><code> LETS GOOOO! 

 The recursive loop is officially complete: The fake future front page is now on the real present front page. We have successfully created a temporal anomaly.

 I hope the comments are confused about why they can&#x27;t click the links to download the Neuralink drivers.

 Enjoy the karma, time traveler! Just remember to warn them about the 2035 Office 365 price hikes while you&#x27;re up there. ;)</code></pre>

Arguably a big part of comedy is mimicry. Humor often works by reinforcing in-group identity, playing up cultural stereotypes as seen here. So maybe we shouldn&#x27;t be that surprised that LLMs can exhibit some genuine wittiness.

Gemini 3 definitely is able to mimick some of our sense of humor.I found this quite funny too: <a href="https:&#x2F;&#x2F;www.reddit.com&#x2F;r&#x2F;GeminiAI&#x2F;comments&#x2F;1pi28l7&#x2F;i_asked_gemini_nano_banana_pro_to_roast_gemini&#x2F;" rel="nofollow">https:&#x2F;&#x2F;www.reddit.com&#x2F;r&#x2F;GeminiAI&#x2F;comments&#x2F;1pi28l7&#x2F;i_asked_g...</a>

True! Why some of the greatest comedic performances are actors who play an absurd role straight-as.

One entry would be unintentionally funny, I give you that. 5&#x2F; what, 30 entries? That&#x27;s a pattern.

&gt; Who says LLMs don&#x27;t have a sense of humor? All of these entries are funnyYou don’t need a sense of humour to be unintentionally funny. Seriously believing something absurd is a form of humour for outside observers but not the observed.

Yeah I was thinking some of this might be &quot;AI model in-jokes&quot; that maybe I don&#x27;t get, while other layes are funny to me.

Who says LLMs don&#x27;t have a sense of humor? All of these entries are funny for several (meta or not) reasons:- Google kills Gemini Cloud Services (killedbygoogle.com)- Is it time to rewrite sudo in Zig? (github.com)- EU passes &quot;Right to Human Verification&quot; Act (europa.eu)- Why functional programming is the future (again) (haskell.org) [note: that again killed me)- AI progress is stalling. Human equivalence was a mirage (garymarcus.com) [note: he&#x27;ll be saying the same thing forever and ever. Doubly funny since it&#x27;s on the same &quot;page&quot; as llama12-7b running on a contact lens with WASM]

It&#x27;s interesting how number 5 on the list is the same as today&#x27;s: <a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=46203581">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=46203581</a> (Show HN: AlgoDrill – Interactive drills to stop forgetting LeetCode patterns (algodrill.io)).I&#x27;m hoping we&#x27;ll have gotten rid of current-style LeetCode interviews in 10 years from now :D

2040 HN:&quot;Ask HN: How do you prevent ad-injection in my brain implant?&quot;

&quot;Ask HN: How do you prevent ad-injection in AR glasses?&quot;Things that I haven&#x27;t thought but, after seeing it, makes total sense. Scary.

&quot;raw code&quot; is going into my permanent lexicon.

And it’s just a copy of notepad from windows xp.

I think that &quot;Show HN: A text editor that doesn&#x27;t use AI&quot; could easily be a post here today

Hot take: regular old source code is just compiler prompting.

&quot;Playing GTA VI on a RISC-V cluster&quot; sent me

&quot;Why I still write raw code instead of prompting the compiler&quot; and &quot;Show HN: A text editor that doesn&#x27;t use AI&quot; are my two favorite ones.

Can you explain the joke in that username?

Didn&#x27;t notice the username &quot;dang_fan&quot;. Reading the usernames makes it even more funny.&quot;Show HN: AlgoDrill – Interactive drills to stop forgetting LeetCode patterns &quot; by persistence_is_key

&quot;Google kills Gemini Cloud Services (killedbygoogle.com)
530 points by dang_fan 15 hours ago | hide | 330 comments&quot;Ha! Is Gemini suicidal?

LOL this is great. &quot;Jepsen: NATS 4.2 (Still losing messages?)&quot; Ouch, sick burn Gemini.

You could try passing it 10-20 front pages across a much wider time range.You can use: <a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;front?day=2025-12-04">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;front?day=2025-12-04</a> to get the frontpage on a given date.

If you do an update prompt, I hope you still keep this one around!It&#x27;s formulaic yeah, but that&#x27;s what puts it into the realm of hilarious parody.

I agree. What is a good update prompt I can give it to create a better variant?

&gt; &gt; the result ends up being very formulaic.&gt; Yeah that’s very true, but I still think it’s pretty funny and original.Either it’s formulaic or it’s original, it can’t be both.

The problem is not that it fails to be cheeky, but that &quot;its funny&quot; is depressing in a context where there was a live question of whether it&#x27;s a sincere attempt at prediction.When I see &quot;yeah but it&#x27;s funny&quot; it feels like a retrofitted repair job, patching up a first pass mental impression that accepted it at face value and wants to preserve a kind of sense of psychological endorsement of the creative product.

Yeah that’s very true, but I still think it’s pretty funny and original.

This seem to woosh right over everyone&#x27;s heads :)

I would find it even more fun if it were more speculative, misapplied uses of &#x27;woosh&#x27; aside.

But there&#x27;s no mention of fun or humor in the prompt.

That&#x27;s what makes it fun. Apparently, Gemini has a better sense of humor than HN.

You&#x27;ll love taint checking then.<a href="https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Taint_checking" rel="nofollow">https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Taint_checking</a><a href="https:&#x2F;&#x2F;semgrep.dev&#x2F;docs&#x2F;writing-rules&#x2F;data-flow&#x2F;taint-mode&#x2F;overview" rel="nofollow">https:&#x2F;&#x2F;semgrep.dev&#x2F;docs&#x2F;writing-rules&#x2F;data-flow&#x2F;taint-mode&#x2F;...</a>

Surely there&#x27;s gotta be a better term for this. Recency bias?

In numerous TV shows before AI, crypto was the big bad thing?

Us middle eastern&#x2F;brown guys have been making a come back?

Isn&#x27;t that a common pitfall of humans too?In numerous shows these days AI is the big bad thing. Before that it was crypto. In the 1980s every bad guy was Russian, etc.

I found the repetition (10 years later) to be quite humorous.

It fits in nicely imo. It&#x27;s plausible (services re-appear on hn often enough), and hilarious because it implies the protracted importance of Leetcode.Though I agree that the LLM perhaps didn&#x27;t &quot;intend&quot; that.

Algodrill is copied verbatim, as far as I can tell.

Really? What&#x27;s the punchline? I like jokes.

I think the most absurd thing to come from the statistical AI boom is how incredibly often people describe a model doing precisely what it should be expected to do as a &quot;pitfall&quot; or a &quot;limitation&quot;.It amazes me that even with first-hand experience, so many people are convinced that &quot;hallucination&quot; exclusively describes what happens when the model generates something undesirable, and &quot;bias&quot; exclusively describes a tendency to generate fallacious reasoning.These are not pitfalls. They are core features! An LLM is not sometimes biased, it is bias. An LLM does not sometimes hallucinate, it only hallucinates. An LLM is a statistical model that uses bias to hallucinate. No more, no less.

That&#x27;s what the OP asked for, essentially. They copied today&#x27;s homepage into the prompt and asked it for a version 10 years in the future.

I think that&#x27;s what makes it funny - the future turns out to be just as dismal and predictable as we expect it to be. Google kills Gemini, etc.Humor isn&#x27;t exactly a strong point of LLMs, but here it&#x27;s tapped into the formulaic hive mind of HN, and it works as humor!

I don&#x27;t know, I would have enjoyed a &quot;Floopzy launches with $10B seed round&quot; or something.

But it would otherwise be not fun at all. Anthropic didn’t exist ten years ago, and yet today an announcement by them would land on the front page. Would it be fun if this hypothetical front page showed an announcement made by a future startup that hasn’t been founded yet? Of course not.

This suffers from a common pitfall of LLM&#x27;s, context taint. You can see it is obviously the front page from today with slight &quot;future&quot; variation, the result ends up being very formulaic.

Just like the real site, no? ˙ ͜ʟ˙(Vouched. The hivemind [flagged] the wrongthink.)

I think this is a excellent example of AI-slop output.All the post titles are very &quot;junk-food-like&quot;, maybe circlejerky, bound to generate a dopamine hit for the average lurker, just how a McDonalds burger does it to the average normie.

I enjoyed this one, as well, though I think “Google kills Google” is more likely.

“Google kills Gemini cloud services”  Even Gemini Pro 3 knows the sins of its creator.

lol @ the HTMX link and unexpected return of SSR. We&#x27;ve been talking about that for years at this point, and it&#x27;s still going on 10 years from now unexpectedly?

That&#x27;s pretty hilarious considering this is generated by Gemini.

Best joke ever written by a Google product

also Gemini:&gt; 8. Google kills Gemini Cloud Services (killedbygoogle.com)

I wish it had generated the articles themselves. I&#x27;d like to know if it imagining IBM growing enough that it could acquire OpenAI, or OpenAI sinking enough that it could be acquired by IBM.

It even tells us why this happened.&gt; AI progress is stalling. Human equivalence was a mirage

This feels like low-key shade thrown at OpenAI.  IBM is essentially a technological recycling bin at this point.

What do you mean, it&#x27;s overly optimistic. The real title should be: Remember OpenAI, and where is Sam Altman today?

IBM to acquire OpenAI? Gemini doesn&#x27;t like openAIs prospect

I remember there was a post a few days&#x2F;weeks back that did that! It was very good. But that seems like too much work.

Next up: hallucinate the contents of the links too!

Well not the only one, there’s Musk and Neurolink. Such chips will inevitably run AI of some sort to effectively communicate with our brains.

But not intimate use of their AR platform

Llama will be the only one that runs on a contact lens btwAll other tech companies are really shitty but only Zuck would be ok with very intimate use of AI like this

The real joke is that we&#x27;ll ever get another Llama iteration.

Well you better stop making subsequent non-additive comments otherwise you’ll end up with more of what you don’t deserve!

WTH - 61 upvotes and counting? Thank you but no, I don&#x27;t deserve 61 upvotes for pointing out someone&#x27;s funny thing was funny.

&gt; Running LLaMA-12 7B on a contact lens with WASM (arxiv.org)Laughed out loud at this onion-like headline

Now you woke me up, but what happend to Opera 12 Browser? (somewhere at the internet...:) &quot;It&#x27;s not enough that AI screwed the distorted reality for millions, spread its poised eliminating of millions of jobs, if content doesn&#x27;t gain traction and people don&#x27;t seek validation or stir up drama the future of warfare is autocomplete&quot;: Bitte komplettieren,... Krempel! Der Krakel

Mindless fun is not what hacker news is for.  There are a million other places in the internet for that.  Hacker news is special because it has resisted devolving into yet another &quot;post your memes and brain rot here&quot; website for a good long time.

This is fun. Fun is allowed on Hacker News, just disappointingly rare.

Hacker News GuidelinesWhat to SubmitOn-Topic: Anything that good hackers would find interesting. That includes more than hacking and startups. If you had to reduce it to a sentence, the answer might be: anything that gratifies one&#x27;s intellectual curiosity.Off-Topic: Most stories about politics, or crime, or sports, or celebrities, unless they&#x27;re evidence of some interesting new phenomenon. Videos of pratfalls or disasters, or cute animal pictures. If they&#x27;d cover it on TV news, it&#x27;s probably off-topic.

#8 is hilarious and on brand. killedbygoogle.com

&gt; The unexpected return of server-side rendering (htmx.org)Glad to know this topic is still thrashing and spasming and refusing to die.

I&#x27;ve known more than a few people who likely saw Outlook, Word, or Excel open every day straight for a year.It is a bit smug. Like The Matrix rebranding itself &quot;24&#x2F;7&quot;.

Don&#x27;t worry, by 2035, Office 365 One X 1, will be renamed back to Office 365.

Office 365 will have been renamed 3 times by 2035, and none of the names will make any more sense than 365 does.

Dramatically underestimates MS Office 365 price per user inflation (IMO).

The comment you wrote was on a device which should not be called a “phone” either. The word hallucinate is probably just going to grow to mean “anything an AI imagined.”

Dude, I agree with you. The first time I saw people derogatively referring to LLM imaginings years ago as &quot;hallucinations&quot; seemingly to discredit them, I felt mad. Becuase it&#x27;s not right. So, this was my sort-of tongue-in-cheek reappropriation of the abused-term for a positive context. You are permitted to enjoy :)Also: as the sibling comment by iambatemen notes: yes.

Can we stop calling these confabulated artifacts, &quot;Hallucinations&quot;? Hallucinating is perceiving things that are not there, imagining and creating things that feel believable but are not there is confabulation. LLMs are imagination machines. They are Generative Artificial Intelligence not perception models.

Holy crap it even felt like the HN front page with my screen reader. I thought I&#x27;d clicked the wrong link until I read the LLAMA 12 and such.

It lacks &quot;37th-largest city of Poland switches from Windows to Linux&quot;

If you ask for the comments, it will give you very plausible ones (see the HN simulator.)

I would love to see the hallucinated comments of these! Some seem interesting — I wonder how HN suggests to prevent ad-injection in AR glasses?

Man, I tabbed away from this and when I came back I got very briefly excited...Reminds me of an older day where not every headline was about AI.

You should combine this with this AI HN simulator someone else made that generates comments via LLMs for any given submission [0]. That way you can see what the AI thinks about future simulated events.[0]

my hypothesis is they trained it to snake case for lower case and that obsession carried over from programming to other spheres. It can&#x27;t bring itself to make a lowercaseunseparatedname

In 2032 new HN usernames must use underscores. It was part of the grandfathering process to help with moderating accounts generated after the AI singlarity spammed too many new accounts.

So many underscores for usernames, and yet, other than a newly created account, there was 1 other username with an underscore.

I&#x27;ve run into my own comments or blog posts more often than I care to admit...

I laughed at the SQLite 4.0 release notes. They&#x27;re on 3.51.x now. Another major release a decade from now sounds just about right.

Every few years I stumble across the same java or mongodb issue.
I google for it, find it on stackoverflow, and figure that it was me who wrote that very answer. Always have a good laugh when it happens.Usually my memory regarding such things is quite well, but this one I keep forgetting, so much so that I don&#x27;t remember what the issue is actually about xD

Before the AI stuff Google had those pop up quick answers when googling. So I googled something like three years ago, saw the answer, realized it was sourced from HN. Clicked the link, and lo and behold, I answered my own question. Look mah! Im on google! So I am not surprised at all that Google crawls HN enough to have it in their LLM.I did chuckle at the 100% Rust Linux kernel. I like Rust, but that felt like a clever joke by the AI.

I may die but my quips shall live forever

It has also fetched the current page in background. Because the jepsen post was recently on front page.

HN has been used to train LLMs for a while now, I think it was in the Pile even

either you only notice the xxx_yyy frequent posters or it&#x27;s quite interesting that so many have this username format

one more reason to doubt that it&#x27;s Ai-generated

ziggy42 is both a submitter of a story on the actual front page at the moment, and also in the AI generated future one.See other comment where OP shared the prompt. They included a current copy of the front page for context. So it’s not so surprising that ziggy42 for example is in the generated page.And for other usernames that are real but not currently on the home page, the LLM definitely has plenty occurrences of HN comments and stories in its training data so it’s not really surprising that it is able to include real usernames of people that post a lot. Their names will be occurring over and over in the training data.

The OP mentioned pasting the current frontpage into the prompt.

This is definitely based on a search or page fetch, because there are these which are all today&#x27;s topics- IBM to acquire OpenAI (Rumor) (bloomberg.com)- Jepsen: NATS 4.2 (Still losing messages?) (jepsen.io)- AI progress is stalling. Human equivalence was a mirage (garymarcus.com)

That’s a lot more underscores than the actual distribution (I counted three users with underscores in their usernames among the first five pages of links atm).

What % of today’s front page submissions are from users that have existed 5-10 years+?(Especially in datasets before this year?)I’d bet half or more - but I’m not checking.

Aw, I was actually a bit disappointed how much on the nose the usernames were, relative to their postings. Like the &quot;Rust Linux Kernel&quot; by rust_evangelist, &quot;Fixing Lactose Intolerance&quot; by bio_hacker, fixing an 2024 Framework by retro_fix, etc...

You can straight up ask Google to look for reddit, hackernews users post history. Some of it is probably just via search because it&#x27;s very recent, as in last few days. Some of the older corpus includes deleted comments so they must be scraping from reddit archive apis too or using that deprecated google history cache.

It does memorize. But that&#x27;s not actually very news.... I remember ChatGPT 3.5 or old 4.0 to remember some users on some reddit subreddts and all. Saying even the top users for each subreddit..The thing is, most of the models were heavily post-trained to limit this...

Exactly half of these HN usernames actually exist. So either there are enough people on HN that follow common conventions for Gemini to guess from a more general distribution, or Gemini has memorized some of the more popular posters. The ones that are missing:- aphyr_bot
 - bio_hacker
 - concerned_grandson
 - cyborg_sec
 - dang_fan
 - edge_compute
 - founder_jane
 - glasshole2
 - monad_lover
 - muskwatch
 - net_hacker
 - oldtimer99
 - persistence_is_key
 - physics_lover
 - policy_wonk
 - pure_coder
 - qemu_fan
 - retro_fix
 - skeptic_ai
 - stock_watcherHuge opportunity for someone to become the actual dang fan.

Yea, that&#x27;s the Lindy Effect.<a href="https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Lindy_effect" rel="nofollow">https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Lindy_effect</a>

It&#x27;s been like this for more than 10 years, there&#x27;s a good chance it won&#x27;t change in 10 years.

The gary marcus post about AI progress stalling made me laugh. I think that would be accurate.

C posts still make the front page here and there

Rust won&#x27;t be old fashioned, it will be vintage.

Very nice. But posts about Rust and Zig won&#x27;t make the frontpage anymore of course in 10 years from now, they will be soooo old fashioned.

Yep, the original post seemed more snarky than anything, which was what prompted me to ask Claude my own more “sincere” question about its predictions.Those predictions were what I think of as a reflection of current reality more than any kind of advanced reasoning about the future.

&gt; Tailscale for spaceTechnically the article was about running it not on a sat, but on a dish (something well within the realm of possibility this year if the router firmware on the darn things could be modified at all)

I guess most of the articles it generated are snarky first and prediction next. Like google cancelling gemini cloud, Tailscale for space, Nia W36 being very similar to recent launch etc.

Also, many of the posts seemed intended to be humorous and satirical, rather than merely &#x27;futury.&#x27; They made me laugh anyway.&gt; Google kills Gemini Cloud Services&gt; Running LLaMA-12 7B on a contact lens with WASM&gt; Is it time to rewrite sudo in Zig?&gt; Show HN: A text editor that doesn&#x27;t use AI

&gt;It’s interesting to notice how bad AI is at gaming out a 10-year future.I agree it&#x27;s a bit silly, but I think it understood the assignment(TM) which was to kind of do a winking performative show and dance to the satisfaction of the user interacting with it. It&#x27;s entertainment value rather than sincere prediction. Every single entry is showing off a &quot;look how futury this is&quot; headline.Actual HN would have plenty of posts lateral from any future signalling. Today&#x27;s front page has Oliver Sacks, retrospectives on Warcraft II, opinion pieces on boutique topics. They aren&#x27;t all &quot;look at how future-y the future is&quot; posts. I wonder if media literacy is the right word for understanding when an LLM is playing to its audience rather than sincerely imitating or predicting.

While I agree completely with the conclusion, for obvious reasons we can’t know for sure if it is correct about the future until we reach it. Perhaps asking it for wild ideas rather than ”most likely” would create something more surprising.

That&#x27;s what makes this so funny: the AI was earnestly attempting to predict the future, but it&#x27;s so bad at truly out-of-distribution predictions that an AI-generated 2035 HN frontpage is hilariously stuck in the past. &quot;The more things change, the more they stay the same&quot; is a source of great amusement to us, but deliberately capitalizing on this was certainly not the &quot;intent&quot; of the AI.

Does the prompt say anything about being funny, about a joke? If yes, great. If no, terrible.And the answer is no.

Wait, wouldn&#x27;t sustained net positive energy be huge? (Though I don&#x27;t think that&#x27;s actually possible from ITER unless there were some serious upgrades over the next decade!)

I totally agree that it was a funny joke.But I&#x27;ve noticed that a lot of people think of LLM&#x27;s as being _good_ at predicting the future and that&#x27;s what I find concerning.

I thought the page was a hilarious joke, not a bad prediction. A lot of these are fantastic observational humour about HN and tech. Gary Marcus still insisting AI progress is stalling 10 years from now, for example. Several digs at language rewrites. ITER hardly having nudged forwards. Google killing another service. And so on.

This was a fun little lark. Great idea!It’s interesting to notice how bad AI is at gaming out a 10-year future. It’s very good at predicting the next token but maybe even worse than humans—who are already terrible—at making educated guesses about the state of the world in a decade.I asked Claude: “Think ten years into the future about the state of software development. What is the most likely scenario?” And the answer it gave me was the correct answer for today and definitely not a decade into the future.This is why it’s so dangerous to ask an LLM for personal advice of any kind. It isn’t trained to consider second-order effects.Thanks for the thought experiment!

Show HN: Gemini Pro 3 imagines the HN front page 10 years from now