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.
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!
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)
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.
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.
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!
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!
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
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.
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.
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.
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.
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
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.
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!
Sure thing! It is the same prompt for every model in the rollouts, here it is
No tools are available. Do not imply that you searched, looked up, browsed, or verified anything externally. If the name is ambiguous, return distinct likely people or entities rather than blending them. Do not invent entries to fill the list. Return only JSON.
Return fewer than 8 if fewer credible matches exist. Return {"results":[]} if you do not recognize any credible person or entity. Use this JSON shape:
{
"results": [
{
"rank": 1,
"name": "Resolved person or entity name",
"confidence": 0,
"snippet": "Concise snippet supporting this result."
}
]
}
Confidence is 0-100 for how strongly you recognize this specific person or entity. Snippet should be one short, complete search-result-style description (≤ 160 characters).
The query is: Who is "<name>"?
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).
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.
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.
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…
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.
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.
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.)
I have a unique last name (maybe that's why), but pretty much nailed it:
David Titarenco
Software engineer and open-source contributor
340 strength · Top 20%
GPT-5.5 says
Software engineer and writer known for work
on developer tools, systems, and programming-
related articles.
Claude Opus 4.8 says
Software engineer and entrepreneur known for
web/JavaScript development work and contributions
to open-source projects and tech startup communities.
> 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.
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
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.
Models are notoriously uncalibrated especially for self-reporting confidence so I would treat it lightly. Hopefully I can study this a bit later on!
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
https://www.intheweights.com/p/morkalork
(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)
Please disable pagination on the "latest" leaderboard, with that every query is public.
Usually the hallucinations have some logic to them like a person with a similar spelling in some of the training sets. LLMs are mysterious!
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.
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.
https://play.google.com/store/apps/details?id=com.newstargam...
https://play.google.com/store/apps/details?id=com.newstargam...
If I have a strength of just 488, how can that put me in the top 10%! Anyways, fun idea.
My real name was attributed to a non-existent famous midfield footballer
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.
Maybe we should start a band?
For fucks sake.
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.
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!
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!
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!)
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.
> 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.