Zero-Shot World Models Attack AI’s Data Bottleneck
Most vision models get good by seeing absurd amounts of data. Zero-shot world models are interesting because they try a…
-
-
Qwen3.6-35B-A3B is Unverified: Qwen3.5 is Real
Qwen3.6-35B-A3B is being passed around as a major new open model release: 35 billion total parameters, 3 billion active, Apache…
-
AI Reproducibility Crisis: Why Claims Fail to Verify
A paper reports a new state-of-the-art result. The repo is public. The figures look clean. The conference is top-tier. In…
-
ARC-AGI-3 Human Baseline Now Means Median Not Outlier
ARC Prize didn’t just tweak a benchmark. The ARC-AGI-3 human baseline now uses a median human run per level instead…
-
Spiking Neural Network Hits 1B Parameters, Hints at New Behavior
A spiking neural network allegedly reached 1.088 billion parameters and trained from random initialization to a reported loss of 4.4…
-
Open Models Are Winning code arena rankings by Fitting the Loop
A strange thing happened to code arena rankings. They stopped being just a nerdy scoreboard and started acting like a…
-
Empirical Research in Machine Learning Ended Math’s Monopoly
A theorem-first paper and an ablation-heavy systems paper can now describe the same model class and leave with very different…
-
Speculative Decoding’s Ceiling Just Moved With DFlash
A serving engineer watches tokens arrive in that familiar trickle: fast enough to demo, slow enough to feel like the…
-
Reduce LLM Hallucinations? Why ‘Make-No-Mistakes’ Fails
The first time you see it, it’s kind of perfect: a tiny folder in your Cursor skills called make-no-mistakes. One…
