Show HN: Are You in the Weights?
www.intheweights.comWith more traffic moving off-web and into LLMs, I got curious about what traces we leave "in the weights". My design partner and I built a site in the past few weeks that checks recognition across frontier and small models. It queries many of them in parallel, clusters the responses, and tells you how strongly they recognize you. Happy to answer any questions here!
Yeah, that went about as well as I would have expected.
It dug up a bunch of what can only be my information, then made up a bunch of confidently wrong things to say about me.
I'm a Software Engineer and SaaS guy, known for running the company "[random word from my blog] Software" and his [different word from my blog] Blog. Founder of three separate startups I've never heard of and may not exist, and well known contributor to Open Source (because that's something that software people often do, so it makes for pretty words to put into a paragraph, despite me not contributing to open source).
Overall, it's like watching a really bad sight reader doing his act. It suggests something that's likely true about you given your background, then keeps tweaking those suggestions until you go "yeah, that's it! you nailed me!".
Sadly, this is pretty par for the course watching AI try to do stuff.
Im an adult content creator and artist, apparently.
I am neither.
So, you must be a dog! ("On the Internet, nobody knows you're a dog").
Jokes aside, I wonder if people from countries with 2+ names and 2+ last names get different results. In my case, I get a lot of false positives because there's apparently a lot of people with same first name and first last name. Luckily, none of them were me.
> I wonder if people from countries with 2+ names and 2+ last names get different results.
I have four names in total, two first and two last, mostly just use one first + one lastname publicly, and this tool only correctly infers correctly when using that specific combination (which also happens to be globally unique, as far as I know), any other combination seems to make the thing hallucinate strongly.
I don't know what people expect. If you ask a random person to recall anyone they've ever known, you're going to get a lot of "don't they have a blog called <random word> software? And maybe they're into photography too?". That's just how information works.
The only difference with an LLM is that it won't say "maybe".
No, a normal person would say, "I don't know them."
Silicon Valley bros, on the other hand...
I have a last name that only my family holds in Denmark (which means me and my brother by now), yet it managed to tag me a Professional Football Player, MP for the Socialist Party and Founder of a Sleep Mattress Company.
I do like the visuals though.
> Overall, it's like watching a really bad sight reader doing his act.
After a few clear fabrications, in the hallucinations it suggests I might be "a private individual" about who there's not much information.
I mean, I guess oh my gosh that's me but...
Mine got me 100% wrong. I recognise some of the information as that of some people with the same surname as me though
I thought it would've just had the information from my linkedin
1. No way in heaven or earth I'm using my real name with this.
2. Alfred E. Neuman < https://www.intheweights.com/p/alfred-e~2e~-neuman > is either "Mad magazine mascot" (11 responses), or "German-American writer, novelist, and playwright" (1 response, from Llama 3.2 1B, classed as a hallucination). Maybe the odd one out means the German writer Alfred Neumann? < https://en.wikipedia.org/wiki/Alfred_Neumann_(writer) >
3. Tamamo-no-Mae < https://www.intheweights.com/p/tamamo~2d~no~2d~mae > is either a "Caster-class Servant in Type-Moon's Fate franchise, based on the mythological fox spirit" (3 responses), or the "Legendary nine-tailed fox spirit" (12 responses, the vast majority, but all classed as hallucinations)!
4. Thank goodness for Firefox's "mute tab" toggle; the thumping and keyclick sounds get real old, real fast.
Don’t worry, I’ll enter your real name for you.
You can try your social media handles
I put my SS in there
Schutzstaffel?
Secret Santa!
Tried my credit card. No results. Good
KanColle moment
What are you concerned about with entering your real name?
My guess name to browser fingerpint and IP map. But any Google auth login could do that.
6 Football (soccer) players share my name and I still am at the top. Type "SEO" and I'll DM you my one little weird trick. /jk
Fun story about my name [0], the bank couldn't mail me my debit card because the mailman kept crossing my address off the envelop.
[0]: https://idiallo.com/blog/sharing-a-name
403 Forbidden
The only other Mike Warot I'm aware of lives in London, and works at the British Museum.
Very fun!
One thing confused me in the story: are "Ibrahim" and "Ibrahima" interchangeable?
Naming children after grandparents gets particularly fun when meeting up with cousins!
Orally they are interchangeable. Though, I only ever append the “a” when speaking in my native language.
Same with me, apparently three different models think I am a minor football player. Looking it up, I am not the first result for my full name, but there also is no minor football player with that name, so…
Also have an extremely common name in Portugal (just in my company there are 4 people with my name, including my previous manager), only slightly helped by the fact that we’re one of the few countries that inherits last names from both parents, which helps with differentiation. At least I did snag pedroalves.pt when I found it available!
EDIT: Username does better, but for some reason Kimi seems to think I do algorithmic competitions, and Llama 3.1 thinks I am a German football club (no longer just a player, a whole club now!)
> Same with me, apparently three different models think I am a minor football player.
I read the parent comment’s blog post right before coming back and reading your comment, and for a moment I thought you meant “models” as in “fashion models” (which makes sense with the footballer) and was expecting a crazy story.
That made me remember how much more fun HN was before it became AIN. It can’t be good for any of us to be inundated with the same subject everyday.
> one of the few countries that inherits last names from both parents, which helps with differentiation.
Not sure I’d call basically every country in Latin America “few”
The.. false positives are extremely scary (not listed as hallucinations)
A terrorist on the US saction list.. the first female airplane suicide bomber?? I was in the US a year ago and I did not bomb any planes
I think with Arabic names it's highly biased, which is kind of scary, I don't want to be bombed based on an LLM query
Apparently I'm either a child voice actor on the cartoon Bluey or an adult film actor. Those were the most interesting anyway. But all of them were hallucinations. The most interesting thing about this, IMO is that none of the models could simply say that they didn't recognize the name.
I was told I had a roll on "The Walking Dead". Interesting that it decided I was an actor, but didn't list the one adult film that somebody with my name was actually in.
I have bad news for you about what the US military does with bombs and LLMs:
https://www.businesstimes.com.sg/opinion-features/claude-ai-...
Ha ha! Yes I am in the weights apparently. Nearly all the models know what I do.
I suspect being in the Open Source world is a bit of a bubble as far as the weights are concerned.
Anyway it stroked my ego nicely even though it was totally artificial, like Zaphod Beeblebrox surviving the Total Perspective Vortex.
Always happy to see a Hitchhikers guide comment.
> Zaphod Beeblebrox
> Fictional two-headed ex-President of the Galaxy
> 979 strength
740, top 5%. Awesome.
https://www.intheweights.com/p/jeremy-edberg-reddit-netflix
Interestingly, almost all of them got it right (although one seems to think I was a VP at Datadog, and I've seen that error before in some LLMs). But Haiku just says "no one of that name seems to exist". So Haiku must be pretty pruned down.
I’m a hallucination. None of these are me.
Perhaps the closest is DeepSeek v4:
> Hyperpape is a user on the LessWrong forum, known for thoughtful comments on rationality and philosophy.
I studied philosophy, so maybe, except I don't post on LessWrong, and I'm not a rationalist.
https://www.intheweights.com/p/hyperpape
You can be in the weights if you make those things true.
Hyperstition
Guess I better take up streaming starcraft 2 lol
https://www.intheweights.com/p/morkalork
It identified me reasonably well (mostly one of the models completely hit the experiments I work on wrong) but then also seems to have invented a football player I can find no trace of. Or would not be surprising if someone with my name was also a football player but I couldn't find easy evidence.
My wife has a very unique hyphenated last name and it totally made up a French linguist with that name.
I was thinking something like this two weeks ago in another thread[1]
>my Reddit history is part of every training set. It was taken without my consent. So now I'm immortal in a way, and hiding in the weights
Anyway 654 isn't horrible for the history still tied to me. That's in the top 6%[2]. It's interesting that it's non-deterministic, and the more keywords you add about yourself, the higher your score goes.
[1] https://news.ycombinator.com/item?id=48403669
[2] https://www.intheweights.com/p/michael-mike-warot-ka9dgx-mrg...
Apparently according to Gemini, I, the only person in the world with my name (unique first name + long complex last name) am a professional soccer player.
Yeah, be careful with answers you get from AIs.
Yep apparently I am a German author, I am not. Further searching and I could not find anyone else with my name.
I'm apparently a former professional soccer player too, but I can't find my alter-ego anywhere using the classic search engines.
Interestingly, the only other person with my (very french) first name and (very spanish) surname is also a software engineer, but in Côte d'Ivoire. What were the odds?
I'm also a professional soccer player. Each model comes up with different nonsense.
Only Grok did well kinda...
> Grok 4.20 says >Dutch software developer and Usenet poster known for eccentric claims about perpetual motion machines and pseudoscience.
Yes, I guess I'm that guy.
Few things are as funny as putting a working device in front of people and see them continue to assert it doesn't work.
It took Howard Johnson, who holds academic credentials in chemistry and physics, nearly six years of legal challenges to finally secure his patent.
After presenting undeniable, physical proof that the device worked and produced continuous motion without an external power source, the presiding judge ordered the USPTO to grant the patent.
More interestingly, after such a long circus the patent could no longer be groped for so called national security purposes.
And most interestingly, there was no academic or industrial interest in the technology.
It reads like a global iq test and it turns out we are really dumb.
Yeah, I’m an actor who has been in game of thrones and the Alien franchise, apparently - but no such actor exists. There are only three findable folks with my name - one’s a felon, one’s been dead a couple years, one’s me.
Oh and sometimes I’m a dog food brand. Go figure.
I, also with an internationally unique name, am a Canadian entrepreneur and a professional rugby player.
I’m wondering if there is a way to use this as proof for my soccer player salary, endorsement deals, etc.
Likewise, unique name, is mostly wring about me, thinks I’m a French politician and hallucinates that I’m an actor (my sister is).
I'm seeing a lot of people report that they're football players and / or actors. Must be a common hallucination.
Me too: must be World Cup themed.
Well, have you ever tried to go pro? Maybe you should
This is a clever trick to get you to enter your real name. ;-) I entered mine, I was on the page kind of, there was some kind of exaggeration of me as the last one. I was surprised someone else in my family who is a kind of actual famous person was not found. It seems to have a lot of recency bias based on that.
My user name here is in the weights. My real name isn't on the Internet.
Oddly, I'm listed only as a stack-exchange contributor, which was really brief compared to hn but its adjacent enough the relations might run together.
"which historical portrait do you most resemble?"
proceeds to get rubes to upload all of their faces in portrait for data mining.
It thinks I'm a German journalist and author focusing on data protection and privacy issues.
It got the job description wrong, but the issues correct. Not too bad. Should've said "asshole with an opinion".
Strangely enough even most tiny models can explain what some of my projects are, including correct historical details. The project websites have my name on it in the contact info and there's other sources connecting it as well. Maybe that gets scrubbed from training data?
Apparently I'm an American volcanologist. Pretty cool.
(I nuke my online accounts regularly to not be tracked - started because I had a stalker but now it's just for the best. I know that this goes against hn rules but yeah it's a bad rule)
Hah. My chosen name collision with my online handle makes the models consistent. They all are certain that I am an adhesives manufacturer. (Good!)
On the other hand, the tool did make an assessment of sorts: NO STABLE PERSON FOUND.
I have a beanie with your name on it somewhere! It was free swag from the adhesives company.
at least ai knows not to mix you up with the horses
It confused me with an academic with the same last name[0], so I guess I'm not notable enough to be in the weights.
[0] Which is uncommon to begin with, so that person might be the only one with it who is in the weights.
I have an unusual name, and have published a book with some minor fame (which is the first google result for my name). Querying ChatGPT, Claude, Grok, etc. gives a reasonably accurate summary of my public info.
OTOH, this tool describes me as a "security researcher known for talks and writing on JavaScript, Node.js, and web security."
I am not a security researcher and have never given any such talk and know precious little about Node.js or web security.
Not that far off, I guess, I might as well try making those the reality...
This was listed as a hallucination but is the most accurate for my name: “A NAME THAT MAY REFER TO AN INDIVIDUAL, BUT I CAN’T IDENTIFY A SINGLE WELL-KNOWN PERSON WITH CERTAINTY FROM THE QUERY ALONE.”
Same result here. That reasonable response was buried in page 2 of the hallucinated results.
Meanwhile, Gemini 3.1 Lite said with great confidence that I was a military police officer who gained national attention in 2024 after being involved in a high-profile confrontation. Other AIs said I was a footballer. Not sure if it's hilarious or worrying...
I got similar from ChatGPT - I took the wording to imply it knew exactly who I was but was going to keep quiet as I am not a "public figure".
What exactly is the "N strength · Top N%" referring to? My name is most likely 100% unique in the world, seems I'm in about 50% of the weights, but I'm really not sure I understand what those yellow numbers mean.
A completely made up name got "110 strength · Top 60%" and "hits" in GPT-5.5 and "Gemini 3.1 Lite", not sure what to make of that either.
This is directional; models self-report confidence on their answers and the strength is a linear combination of the confidence plus a bonus for every model that got clustered in.
Models are notoriously uncalibrated especially for self-reporting confidence so I would treat it lightly. Hopefully I can study this a bit later on!
There is a Country-Pop singer songwriter with my name. Also a 1940's running back for the Giants.
Then there is a third one which might be a hallucination (that one ironically appears to be me, and the other two are hallucinations).
Bahaha apparently only in their hallucinations. I’m not a professional rugby player or a neurologist.
If there is someone else with the same name, I’m not sure that is a hallucination? But if there isn’t then yes.
Interesting, I wonder if the rugby thing is a common bias. I did find myself in the weights, as the top result. But apparently there are also Australian rugby versions of me!
German football goalkeeper here :)
We need a name for these pure hallucinations, something like lucies or looseys
Usually the hallucinations have some logic to them like a person with a similar spelling in some of the training sets. LLMs are mysterious!
GP didn't give enough information to know if this was actually a hallucination or not, let alone what type of hallucination. I.e. it's only a hallucination if no rugby player or surgeon is a John Doe, not if John Doe the GP isn't those things.
I wonder how much of hallucinating/"mistakes" in LLMs is because the training data is full of us filling in additional info we humans commonly feel or interpret as implied rather than something which manifests from the architecture of the LLM itself. I assume only a small percentage, but also a non-zero one.
Strange, there’s a neurosurgeon and Australian Rules Football player that share my uncommon name. I already knew about them from googling myself previously. Eerily similar!
Is there any reason to assume it wouldn't be? A lot of training data comes from the open web, after all, and Google also searches Google books, so a Google search is basically a model training data search.
The only interesting thing is how small the models have to be, to lose knowledge of you.
It’s funny, seeing the block (rather than line) cursor in the text box, my fingers itched to press i to enter Insert mode before typing my name.
Now that's interesting. I put my name in and it gave me a bunch of made up identities plus some really funny stuff in the hallucination part [1].
But then I entered my name as it's on Linkedin, including a nickname, and it totally failed to find me even then. Pretty sure the full name + nickname combo is unique.
[1] It said I'm former prime minister of Romania, the locals should know why that's funny in the current political circus.
This must be a remarkably expensive demo/toy to operate.
Not cheap for sure but it's all for fun! I have done some optimizations to try to get cost as low as possible; the final clustering actually uses Kimi K2 for this reason. More info on https://intheweights.com/about
Because you don't have a privacy policy or anything really, I assume you're harvesting IP addresses and selling matches to the highest bidder.
He stands to make dozens of fractions of a penny doing that! Must be pretty tempting.
There's a nice feature in Brave for sites with obvious privacy implications: right click -> Open link in private window with Tor.
Well, guess we'll have to wait a bit to see if we're in the weights... I got a 429, as I'm sure many others are (and thus mashing retry).
Didn't expect to hit the front page! Trying my best to keep it up
Please place a large obvious notice that everything you type into that box will immediately be made public.
Please disable pagination on the "latest" leaderboard, with that every query is public.
Just disabled latest!
BUTTWIPE MCFART FICTIONAL INTERNET PERSONA
LLAMA 3.2 1B SAYS
MCFART IS A POPULAR INTERNET PERSONALITY KNOWN FOR HIS HUMOROUS CONTENT.
wow how does it konw
LOL, I'm a TV actor, a Real Madrid football player, a musician and a pro photographer.
Fortunately, my real life namesake, a gay porn performer, didn't register though. Short career span I guess.
I also have a gay porn doppelganger who didn't show up. Must be a conspiracy against gay porn.
Reminds me of the scene from Devs:
"The box contains us, the box contains everything and inside the box there's another box"
Yay, I made it! https://www.intheweights.com/p/adam-gaweda
Now GET OFF THE INTERNET!
I share my name with a famous sports star and that person comes up far more than me, which is understandable. It absolutely nails my username but I've used the same one online for almost 30 years. It'd be weird if I wasn't fairly well understood.
What this tells me is that I've done a decent job of keeping my real life and my internet personas nicely separated.
I once had a model insist that I was a web designer living in Brighton who ran an agency called 'Guerilla Futures' and was the author of a series of UX design for babies books.
Obviously it was a hallucination, but a very detailed and consistent one. Especially as if things had gone slightly differently there's a good chance I could have ended up in Brighton. Plus it's a pretty good name and the books are a fun idea too. Was this my Sliding Doors moment?
Interesting. Claude Opus 4.8 and Gemini 3.1 Lite kind of got it right, but when I ask the model directly, they say they don't know. I'm curious how the tool is doing the correlation.
Prompt for rollouts posted below (https://news.ycombinator.com/item?id=48592415). I have a bit more information on the clustering part in https://intheweights.com/about but every thing returned by the model is viewable (possibly under the "hallucinations" section)
1985: Austrian rock singer Falco records Rock Me Amadeus!
I was hoping somebody would catch that!
There was a tribute song that came out later that I looked for but had trouble finding. In Falco's timeline you find that he died tragically in 1998.
Of these models, only Kimi had anything on me and it was pretty inaccurate.
When Fable was accessible, I asked it about myself and it had some accurate information about me. It's neat. It feels a tiny bit like I got to sign the Voyager probe. I wonder if Fable was trained on a significantly different selection of data or if it's just better at retaining rare details it saw in its training.
I tried both of my long-lived social media (Xanga, LJ, MySpace era) handles from my teen and early twenties years (I mostly use disposable handles now). I've deleted a decent chunk of those postings, but they were both recognized (top 25% on both), although for the wrong reasons (never a Minecraft Twitchstreamer, but I did have a Minecraft account with the handle name).
Really odd feeling to think that my writings from that period are helping these things. Not necessarily happy about it--I took that stuff down because it was so deeply personal, and a record of my life that didn't need to be public. Odd to think my teenage angst is, in some small way, writing the AI-generated e-mails I get now.
apparently i'm also a minecraft youtuber. i have never played minecraft or posted a youtube video.
i think somewhere along the lines the models might have just gotten "online handle" and "minecraft content creator" conflated.
My wife tried it with her longest-running handle and it got social media right, but said she was on TikTok and Instagram (both wrong, just old Twitter). It's an interesting mix of truth and hallucination.
Only a fool would enter their name in this.
Absolutely, a good hacker could likely get into your bank with your name alone.
Life must be rough for John Smiths.
Right on, nobody will know my name as long as I don't search my name where other people can see it. My name is a secret.
This is just an SEO job/psyop to make "Kyle Maxwell" an even better alias. Hiding in plain sight.
I trust your real name is not actually Kyle Maxwell :) I do this too, btw - a random name generator is the very best form of internet anonymity.
(If it actually is your real name, then I can only assume you're using an iocaine powder strategy to beat the internet ...)
Unless you think Osama Bin Laden[1] is bored and browsing HN, I'm desperately curious to know what you think the risk here could possibly be.
[1] (Yes, I know, it's a joke.)
From one person who's completely not-anonymous on HN to another: Why? Do you worry about the association of your IP address with your real name by yet another site other than any place you've made purchases from or signed up for an account from?
(I'm asking seriously, as I can see some risk to having that linkage more public, but given the rate with which services holding PII are compromised and my own personal rate of receiving notices of "oops, we kinda sorta leaked everything about you, here, have more free credit monitoring", I assume almost all of this is available already.)
I think its guys just who don't really understand how the internet works. I'm flabbergasted that its 2026 and I still see the "oh no my IP" trope.
If you're worried about it use a VPN.
Your ISP anyway knows your IP or at the very least your current one if its dynamic.
People who blur their car license plates - when you can go to their social media profile and usually geolocate them down to a street address.
Heh. That's what I get for having a weird name and having been on the internet since like 1993… https://www.intheweights.com/p/zellyn
My name is Seth Green, which I share with a more famous actor [0]. I go by Seth Ariel Green for disambiguation.
GPT-5.5 tells me about the actor, but Claude Opus 4.8 and, weirdly, Grok 4.2 know who I am [1]. I wonder if that's because I use Claude more? Grok I have no clue why.
[0] https://www.imdb.com/name/nm0001293/
[1] https://med.stanford.edu/profiles/seth-green
I only got hallucinations of random combinations of my (fairly unique) last name & first names that do not exist, combined with very accomplished and completely fictional biographies. I guess I'm not notable enough which is somewhat comforting.
I'm in the weights! I've successfully been immortalized! Except... I don't podcast; I hate my voice. Guess the models are trying to tell me to podcast. And other people say I should podcast. But that takes so much darn time!
https://www.intheweights.com/p/devin-prater
According to Kimi, I'm a Dutch robotics engineer. Got that part actually right! Not so much for the rest, about First Lego League. But I did RoboCup, so close enough I guess.
The other models think I'm Dutch (I guess the 'van' gives that away?) and am a soccer/football player. I don't know anything about soccer though
It has me through my open source contributions, but interestingly it claims the same regardless of what I use as a given name, as long as I keep the (fairly unique) surname. So my whole extended family are open source contributors.
I have yet to get the page to load, but due to gmail mixups I've been confused with a retired professor of economics in the UK, and also got a pair of tickets for a King Gizzard and The Lizard Wizard concert.
I hope you went to the concert because they put on a great show. But of course it probably wasn’t near you if it was from a gmail mixup
Interesting one. It knows my internet handle, but when given my full name, it immediately starts hallucinating based on the name structure, guessing which country I'm from and whatnot.
For something that's a toy project, and definitely doesn't seem it's a transparent attempt to get HN user's names, there sure are a lot of tracking cookies for such a website.
What tracking cookies are you seeing? The intention was just some cloudflare checks for spam identification
When I tried this with a self-hosted Qwen model it hallucinated all kind of stuff about me being deeply involved with early Bitcoin development, conferences, and libraries.
Hi Satoshi
Well, according to this I’m a Mexican painter/actor/footballer. Love it.
Strangely only "Kimi" has accurately heard of me. Gemini thinks I'm a German-language version of the stuff I do in English, Kimi recognizes my long-defunct blogging about technology and economics.
I'm glad to hear the teenage drummer I used to see when googling myself has gone pro. He's doing pretty well, too, if these models can be trusted.
There's a famous son of a star, his fame comes from drug addiction and felonies. I'm doomed to have this bias on my name forever...
I put in my name, and four boxes popped up -- one for "American mathematician", one for "spelling bee contestant", one for "American poker player", and one for "fitness industry entrepreneur".
In fact, both of the first two are me, but I wonder if Claude Opus 4.8 (the only one that hit both of those two) realizes they're the same person? :P
Retro bowl (& goal) theme spotted
https://play.google.com/store/apps/details?id=com.newstargam...
https://play.google.com/store/apps/details?id=com.newstargam...
We were drawing inspiration from https://www.michaelfogleman.com/static/nes/
Can you share the prompt you're using for each model?
Sure thing! It is the same prompt for every model in the rollouts, here it is
The clusterer prompt is more intricate and I'm happy to share if of interest, but I have an invariant that every result showing up in a rollout must be clustered into one result (sometimes collapsed into the hallucinations section).
What if it doesn't return JSON?
I score 1%. Is that good or bad?
https://www.intheweights.com/p/ken-shirriff
Top percentile of people searched!
No privacy policy whatsoever?
Would you trust the site if it did?
Incredible concept and a very well-crafted site. I scored very low, but then very high with my legal name. It seems DeepSeek knows a lot of arxiv papers (or at least, about the authors).
After searching "foo", if you try to search "foobar", it deletes "foo" and searches "bar".
According to this, LLMs have never scraped Slashdot!
Slashdot can now be a safe space :-)
Apparently, I am "in the weights". Kimi gets it correctly: https://www.intheweights.com/p/joseph-jude
This is fun!
Great design and artwork. How did you generate the portraits?
Have you thought about extending this into some sort of pipeline for AIO?
I like the Possible Hallucinations feature. Seems like a feature that could stand on its own. Interested in how you separate those out.
In my case the possible hallucination was the only one that was 100% factual.
Thanks! It is part of the clustering step, I tell the model to make a judgement of whether something is inanimate or hallucinated (as defined by low support from only non-frontier models / judgement). I iterated on this a lot and made an eval set out of my LinkedIn contacts where I run GPT5.5 with web search and xhigh reasoning to determine pseudo ground truth. I tuned this to be higher recall (more things classified as non-hallucination) but it is definitely not perfect
I’m in the weights!
So… what does this mean for the right to be forgotten?
I am currently on a long laggy line on a resort island. Your timeout is too short, I’m sure it would work if it didn’t cut itself off.
So despite publishing a lot of fiction and poetry I’m apparently most well known for my contributions to the TeX, LaTeX and typography communities. It also thinks I’m a professional athlete having played professional baseball hockey and basketball.
Interesting, that probably reveals something about the training set for most of these models
I knew I shared a name with a former NBA Forward, but I did not know about some of the other well-known figures. Sadly (or thankfully?) I'm not listed anywhere, even in the hallucinations.
I am in the hallucinations.
Initial reaction was "wow! I guess I have the same name as a Canadian actor!" And then I looked it up and figured out that I do not, in fact, share my name with a Canadian actor. Kimi K2 and GLM both hallucinated the same thing.
everybody is apparently some kind of professional sports player...
The two matches for my name are hallucinated athletes. For now.
I am in the top 3% but I don't know if that means a lot or not much :-/
I hope this isn't costing you a ton in tokens
Apparently I share a name with a prominent white nationalist activist. Yikes.
Interesting Mistral sort of knew something about me, both gpt and deepseek produced the same answer more or less. I wonder why xD, only gemini knew my online handle mostly github and rust which is interesting.
I typed my boss' name in and it returned:
"No stable person found"...
Cool way to get names associated with IP addresses
What in the world is that clicking sound on scroll???
Whoops we had some scrolling bugs with sound, hopefully fixed
Will Smith
American actor and rapper > 984 strength · Top 1%
MLB catcher for Dodgers > 255 strength · Top 25%
MICHAEL RYAN HUNGERFORD MASSACRE PERPETRATOR 204 STRENGTH · TOP 35%
For fucks sake.
And here I thought my being a murder victim was bad.
I looked up the city and year cited by the model for my untimely demise, and it turns out the crime is real, but the real victim was a female sharing my last name, with a middle name loosely resembling my first.
Well, at least he wasn't in the Ep*tein files!
There seems to be some top twenty that rank highly, probably in part due to them being in the files that can't be named!
Straight to jail bud, the AI says you're guilty so it must be true
We can help strengthen the case by making the website about it.
Let's also make a match history page with all those soccer players. Generating the video will have to wait a few years.
My real name (542, top 9%) is quite a bit under my username (692, top 6%).
Nice retro styling
just to be clear, you have each search running on all those models? Self hosted a lot of them right?
I am not in the weights
They all know me to 68%-88% certainty. “Known for my contributions to open source”, yeah, sure, let’s go with that ;)
Deepseek seems to know a lot about me!
If I have a strength of just 488, how can that put me in the top 10%! Anyways, fun idea.
Deepseek thinks I'm the co-founder of Lunar... I would like to inform everyone that this is not true.
Love the graphics, the 8-bit style of the people's portraits is really well done. Are those AI generated?
Yep, those are from "the weights" of GPT-5.4 Image 2 with a little "draw <name>" query and a style reference. More details here https://intheweights.com/about
Why can't it draw Elvis and Hitler?
It is on a 10 minute interval and only does images for the top people, should pick up Elvis shortly. On refusal it shows an X for the person, sometimes the upstream model (gpt-5.4 image 2) will refuse and there are a few names I manually omitted.
Why refuse Adolf Hitler, but allow Vladimir Putin? I mean, it's not a race, but if you're going to draw a line in the sand (labelled 'no admittance to utter cockwombles' or something like that) I know where Putin would end up. Are those models refusing one and allowing the other? Why do they even refuse one?
Some models kind of got it right, but when asked directly, they say they don’t know me. Curious behind the scenes how the correlation works
this is super fun, interesting to see how much these LLMs know things without internet as a knowledge base
I’m the top one! Interesting to see the hallucinations creeping in across the weaker models.
First response for me was also a hallucinated Scottish soccer player who doesn't exist
I really like the website itself
I don't. It drops key presses and randomly moves the cursor on mobile, unless you type slowly.
Not terrible although I am somewhat insulted that QWEN3 8B hallucinated me as the chimp from Jimmy Neutron and no, MISTRAL 3.2 24B, I don't stream on Twitch.
Oh and KIMI K2 0905 completely hallucinated a real name for me (I don't work on Pygame!)
well, the lower confidence ones got my pseudonym, the higher confidence ones missed entirely and attributed it to a prominent speedrunning streamer.
My real name was attributed to a non-existent famous midfield footballer
There is a 'hallucinations' section on the page, which suggests that the items above that section are not hallucinated. I highly doubt that.
I am, as far a I know the only person in the world with my name. So I searched for my name. I am none of many things this tool tells me I am, for example a right wing politician, a journalist,l and a researcher on solar fuels.
Ah yeah, the "hallucinations" classification is optimized for recall (keeping as many results as I can) not precision. It is mostly based on small models being the only support for a claim. Certainly lots of hallucinations everywhere!
Surprised to find myself in the top 50%. Like... _really_ surprised.
Apparently gpt 5.5 thinks I’m a metal folk musician, i wish.
Ha thats funny it thinks I'm a jazz-funk musician.
Maybe we should start a band?
Nice, I am not good enough engineer to be in the weights
Neither am I, except according to LLAMA who reckons I contributed to Postgres. Meh, I'll take what I can get.
I really like seeing the differences in responses between the models, its neat to see the intelligence on them.
My handle and real name give more or less correct results with a 220-243 score (whatever that means). Gemini insists on me working in sound engineering, something I did in the 90s, but at least it's not completely absurd.
The most correct result comes from Opus 4.8, but is amusingly deemed an hallucination:
Claude Opus 4.8 says
A name associated with French IT/systems administration and Linux community discussions, possibly a storage and data systems specialist.
A semi-famous-in-academia family member seems to be in all of the weights, except for llama3.2 1B and qwen3 8B. You'd expect the 1B to be the worst, but actually it got quite close.. qwen3 8B was a complete hallucination.
Fascinating! I’d like to learn more about how to interpret the results to be honest, the About is awesome and helpful.
I scored 1,100 total on my music moniker. It has been used in SoundCloud and also via streaming services/releases via DistroKid. Represented in all the models but of course not disproportionally large fame so to speak. It’s just a very unique setup, somewhat designed to stand out.
My writing account, newer within the past few years, is just under 1,000. The Kimi and DeepSeek pick that up a lot more. I wonder if they train on Medium more than the others…
Thanks for sharing!
“Al Qaeda terrorist. Involved in Madrid 2004 bombings”
…WTF!?
> GEMINI 3.1 LITE SAYS A FRENCH DIPLOMAT WHO SERVED AS THE AMBASSADOR OF FRANCE TO THE REPUBLIC OF THE CONGO (BRAZZAVILLE) FROM 2017 TO 2020.
er. okay. The good thing about this test is that I am the only person in the world with my full name, and I know all the people with my last name (about 30-ish people). None of us are ambassadors, none of us are related to Congo in any way.
LLMs have to be a lot weirder if you are famous enough to chat with them about you.
super cool!! I love the idea and the UI
If you average all of them together, it's close:
https://www.intheweights.com/p/reuven-swirsky
If I spell my name in Hebrew othography, it comes even closer
https://www.intheweights.com/p/~5e8~~5d0~~5d5~~5d1~~5df~-~5e...
But none are exactly right.
I love this!
It's amazing how it jumbles things up. Really shows you that even the leading models still very much hallucinate esp when they don't have the ability to go looking for more context. It took various things related to stuff I work on but mixed them up and added pure invention or mixed bits up with other people with vaguely similar names or projects.
It nailed 2 out of 4, which I'm not going to repeat to preserve a modicum of privacy.
But unfortunately I'm not a professional footballer _or_ a fictional character in a Henry James novel (though I looked up the reference and it's close!)
Feels great to have both a very generic first and last name and share them with others who are internationally known and some more locally. I really have no desire to be in model weights.
An they nailed me, as soon as I clicked the link I saw “rate exceeded”
My username shows up as me. My real name is apparently shared by more real people than I figured (surname is an oddball). That guy's a CEO and billionaire. Go figure, never heard of him until just now.
I get why they couldn’t slop pixel art Hitler, but why not Mandela?!
Yet another reminder that my wife is far more well known than I am
PRIVACY WARNING: Every name/text entered into this site is publicly listed on the "latest" leaderboard which seems to paginate endlessly.
And will thus potentially end up in the effing weights.
Ugh too fucking late. What a privacy nightmare.
That sucks… shame on me I guess
This was the first thing I thought too.
Even if this thing wasn't publicly displaying the names, I would assume they would be collecting them for something.
Can't trust anything like this online.
Can’t trust anything online
Wouldn't thinking so be the default for the HN crowd? I'd have thought any hacker would assume any text you type in a random website would be used however the website administrator wanted. (Not that the general public would think so.)
Just deployed a fix for this; removed latest and capped pagination.
It is still possible to download all entries via the api and sort them by the timestamp. Removing the cursor data would be one way to mitigate this.
Currently there are a bit more than 43000 entries. As far as I have seen, only the results are stored. When I entered a random name, only a similar name was found, and that similar name result was stored, but not the original input.
update to my previous comment:
All the data is still public. There are more than 104000 entries now.
The original name, that was searched, is also stored in the data (in another field; somehow I missed that before).
@tourtlesoup: Why don't you restrict the access and why don't you put a warning on your page?
That's the first thing I looked to see HN-ers real names :-) and thought "hey, that's a pretty clever to get everyones' names".
Apparently it's fixed now. Surely you'll trust a random website...
I have a unique last name (maybe that's why), but pretty much nailed it:
"George McBay"
> Llama 3.2 1B says
> American actor, best known for his roles in films such as 'The Big Lebowski' and 'The Big Lebowski 2'.
Nailed it! /s
But even the entries that aren't marked as likely hallucinations are wrong for me on this site.
> George McBay
> African American chemist and educator
No, that's Henry Cecil McBay (no direct relation that I'm aware of).
Google Search's AI mode does match actual me, but the information it spits out is all mixed up with information on another person who has my same name (also no relation that I'm aware of) and is also a software developer.
Aye... right now the clusterer does the classification of whether it thinks it is a hallucination or not (it is biased against only small model support) but I tried to optimize for recall over precision. The query is essentially "Who is <name>" so a lot of the hallucinations are just the LLMs their usual mysterious way of thinking - usually some relation but loose.
It looks like something perfect, what is its purpose?
No purpose, just a fun hack and science experiment. Glad to see it getting a good reception!