ML Times

Jan 2, 2026

Daily

Weekly

AI Futures Model: Dec 2025 Update
The AI Futures Model predicts a 3-5 year delay in achieving full coding automation compared to previous models, primarily due to a more cautious assessment of AI R&D speedups and the dynamics of automation.
Dell's version of the DGX Spark fixes pain points
Dell's GB10 mini workstation addresses key issues of the DGX Spark, featuring a 280W power supply and improved thermal design for enhanced performance and quieter operation, despite being priced higher at $4,000+.
Open sourced Loop Attention for Qwen3-0.6B: two-pass global + local attention with a learnable gate (code + weights + training script)
Loop Attention enhances Qwen3-0.6B by implementing a two-pass attention mechanism that combines global and local contexts through a learnable gate, allowing for adaptive focus during processing.
Eigenvalues as models - scaling, robustness and interpretability
Matrix eigenvalues serve as a novel approach to model nonlinearity, enhancing scaling, robustness, and interpretability in machine learning frameworks.
LEMMA: A Rust-based Neural-Guided Theorem Prover with 220+ Mathematical Rules
LEMMA is a Rust-based theorem prover that integrates 220+ mathematical rules with a neural-guided approach, enhancing the efficiency of symbolic computation through Monte Carlo Tree Search.
I built a drop-in Scikit-Learn replacement for SVD/PCA that automatically selects the optimal rank (Gavish-Donoho)
The randomized-svd library introduces auto-rank selection using Gavish-Donoho hard thresholding, eliminating the need for costly cross-validation in SVD/PCA applications.
Vibe Coding for CTOs: The Real Cost of 100 Lines of Code
Vibe coding represents a shift from traditional programming to orchestrating AI agents, allowing developers to focus on high-level objectives rather than line-by-line coding, fundamentally changing the software development landscape.
Announcing OpenAI Grove Cohort 2
New paper by DeepSeek: mHC: Manifold-Constrained Hyper-Connections