Stories by lairv

MTEB v2: Evaluation of embedding and retrieval systems for more than just text

RTFM: A Real-Time Frame Model

Figure 03, our 3rd generation humanoid robot

xLSTM Scaling Laws: Competitive Performance with Linear Time-Complexity

Writing high-performance matrix multiplication kernels for Blackwell

IBM Granite 4.0: hyper-efficient, high performance hybrid models for enterprise

Failing to Understand the Exponential, Again

Claude can sometimes prove it

DINOv3

OxCaml - a set of extensions to the OCaml programming language.

Deprecate TF + JAX

Llama 4: Did Meta just push the panic button?

CheerpX 1.0: high performance x86 virtualization in the browser via WebAssembly

Madrona: A GPU-Accelerated Game Engine for Batch Simulation

James Cameron Joins Stability AI Board of Directors

The future of Deep Learning frameworks

Kyutai Moshi: a realtime audio LLM

Teardown Engine Technical Dive [video]

1 2