Bb5a31f81ClaimOpus 5.5 completed 26% of the machine-readable formalization of Fermat's Last Theorem over 18 hours before halting progressAgreeDisagree
Bb5a31f81ClaimFable 5.2 formalized Fermat's Last Theorem into a machine-readable proof in 30 hours using 1,000 agentsAgreeDisagree
Bb5a31f81ClaimResend built OAuth authentication and role-based access control to deploy a remote Model Context Protocol serverAgreeDisagree
Bb5a31f81ClaimAblating 80% to 90% of neurons in the Qwen 3.6-35B A3B language model improved performance on ARC-AGI benchmark puzzlesAgreeDisagree
Bb5a31f81ClaimNoam Shazeer proposed the SwiGLU gated multi-layer perceptron architecture in a 2020 paperAgreeDisagree
Bb5a31f81ClaimTester Ali K. Miller utilized Fable 5.2 to build a water safety map covering 3,100 United States cities using official government dataAgreeDisagree
Bb5a31f81ClaimThe original open-source Model Context Protocol server for Resend was created by Andreessen Horowitz investor YokoAgreeDisagree
Bb5a31f81ClaimThe Neuron Scientist agent utilizes experimental tools including generating counterfactual sentence pairs and steering activations to refine neuron hypothesesAgreeDisagree
Bb5a31f81ClaimThe prototype operating system designed by Fable 5.2 boots on a laptop with machine drivers, file storage, a window manager, a login screen, an installer, and integrated Claude functionalityAgreeDisagree
Bb5a31f81ClaimResend expanded its Model Context Protocol server implementation to include over 80 individual toolsAgreeDisagree
Bb5a31f81ClaimSpecific reasoning neurons in language models fire to modulate confidence, induce delays, trigger self-checking, or pivot problem-solving directionsAgreeDisagree
Bb5a31f81ClaimResend deploys `llms.txt` files and Markdown versions of its marketing and documentation pages to facilitate parsing by AI modelsAgreeDisagree
Bb5a31f81ClaimKurt Tigges and Nick Cammarata co-authored the research paper Neurons as an Interpretable BasisAgreeDisagree
Bb5a31f81ClaimAblating approximately 90% of individual neurons by retaining only extreme activations causes minimal performance degradation on standard language model benchmarksAgreeDisagree
Bb5a31f81ClaimHypotheses generated by the Neuron Scientist agent are evaluated by prompting an independent language model to predict token-by-token activation levels on unseen textAgreeDisagree
Bb5a31f81ClaimIndividual neurons in language model multi-layer perceptrons can express a compact set of distinct meanings rather than hundreds of unrelated meaningsAgreeDisagree
Bb5a31f81ClaimSelecting the top 64 activating neurons per expert in Qwen 3.6-35B A3B ablates approximately 88% of all neuron activationsAgreeDisagree
Bb5a31f81ClaimResend added explicit text data to its pricing page because web-crawling LLM agents fail to execute interactive JavaScript pricing widgetsAgreeDisagree
Bb5a31f81ClaimFable 5.2 coordinated over 1,000 sub-agents over eight days to design and compile a prototype operating system from an empty directoryAgreeDisagree