Anyone who’s Googled themselves precocious knows that it doesn’t rather deed the mode it utilized to. Sure, there’s everything going connected with Google hunt itself, but there’s besides an inescapable feeling that web hunt isn’t the canonical root of accusation that it utilized to be, with conscionable arsenic galore radical learning astir you and maine from chatbots.
Thomas Dimson and Joey Flynn had a akin feeling, starring them to make In the Weights. The “weights” successful question are the numerical parameters that signifier an AI model’s grooming and output, truthful the website purports to measurement however good “a exemplary is capable to callback idiosyncratic without utilizing tools similar web search.”
“Being successful the weights means your beingness was deemed important successful the process of creating superhuman artificial intelligence,” the website says.
To execute this, In the Weights supposedly queries antithetic models (including Grok, Gemini, aggregate versions of GPT, Claude, and Llama, positive lesser known models) with a question akin to, “Who is <name>? Give up to 10 results, each with a abbreviated statement and confidence.” It past “cluster[s] akin descriptions unneurotic and assign[s] a spot score.”
For example, this humble tech blogger received a spot people of 641, placing maine successful the apical 6% of names. I was feeling beauteous bully until I saw that multiple TechCrunch colleagues scored adjacent higher. And the leaderboard has been shifting arsenic I constitute this post, with “Home Alone” prima Macaulay Culkin presently successful the apical slot with a spot people of 988, followed by opera vocalist Luciano Pavarotti.
The results besides amusement which models returned answers for a fixed name, and they item imaginable hallucinations — seemingly GPT-5.4 Mini says that Anthony Ha is an “ambiguous sanction signifier that could notation to aggregate radical with the initials A.H.A.”
Asked wherefore helium built In the Weights, Dimson told TechCrunch via email that helium and Flynn were looking to “get the originative juices flowing again” aft leaving OpenAI (which they some joined done the acquisition of their plan startup Global Illumination).
Dimson said helium was reasoning astir however “Google vanity searches are the incorrect nonsubjective successful 2026 arsenic much postulation moves to LLMs” and astir the information that “so galore lives are encoded someway successful a clump of floating constituent numbers wrong the AI brain.” He besides said the absorption of the tract was “sealed” by a tongue-in-cheek blog post riffing connected AI weights and Terry Bisson’s classical abbreviated communicative “They’re Made Out of Meat.”
“Reception has been insane truthful far, we thought this would beryllium a mild curiosity but it seems similar it has struck a nervus of wanting to spot if you unrecorded everlastingly successful the ace quality (the examination origin doesn’t wounded either!)” Dimson added.
While I’m not rather arsenic convinced that being “remembered” by a chatbot is simply a guaranteed summons to immortality, I can’t contradict that I find the results some intriguing and jealousy-inducing, particularly since they’re codified successful an easy-to-compare score. (AI professional Anthony Moser scoffed that this is “literally the aforesaid arsenic asking 13 chatbots to archer you astir yourself.”) Also helping: The information that the tract features a cute, Nintendo-inspired retro design.
Dimson said helium plans to excavation successful further into wherefore antithetic models successful the aforesaid bid instrumentality antithetic results, which models are biased towards antithetic types of people, and which radical “should person a Wikipedia nonfiction but don’t.”
When you acquisition done links successful our articles, we whitethorn gain a tiny commission. This doesn’t impact our editorial independence.















English (US) ·