Stories by n-e-w

The Trans-America Trail Guide

Dispersion loss counteracts embedding condensation in small language models

Show HN: Coding agent that compiles intent into deterministic DAG before running

Architecture as an executable constraint instead of documentation

Meshtryoshka: Differentiable Mesh Rendering for Unbounded Scenes

Simplified Sparse Attention via Gist Tokens

Continuity-Enhancing Degree Elevation and Splits

Forward Self Models

Tapered Language Models

Everything is logarithms

Deactivated Slack Profile Generator

Explaining Attention with Program Synthesis

FFI in Miri at 8000 segfaults per second [video]

Show HN: LocksBet – a price comparison tool for prediction markets

Sparsely gated tiny linear experts

Robotics Has a Stiffness Problem

Show HN: Typedframes – Pandas/polars column name checking at lint time

Training-Free Single-Image Diffusion Models

Simple input–output dependencies explain neuronal activity

Self Teaching Autoencoder

1 2 3 4 5 6 7 8 9 10