-
The Crystallization of Transformer Architectures (2017-2025)
A dataset-driven analysis of transformer architecture choices and their convergence over eight years.
-
Building a Fast BPE Tokenizer from Scratch
Incrementally optimizing a BPE tokenizer with complexity analysis and benchmarks.
-
Systematic Pessimism
A new paradigm for scaling quality engineering with AI — automated discovery of edge cases or potential failure modes at every commit.
-
Beyond Automation — The Case for AI Augmentation
The really transformative interfaces won't be the ones that make us more productive; they'll be the ones that make us more thoughtful, more creative, more aware of our own cognitive patterns. Like mirrors for our minds, showing us our blind spots and suggesting perspectives we habitually miss.
-
Rethinking Generation & Reasoning Evaluation in Dialogue AI Systems
As we rely further on (and reap the benefits of) LLMs’ reasoning abilities in AI systems and products, how can we still grasp a sense of how LLMs “think”? Where steerability is concerned — users or developers may desire to add in custom handling logic and instructions — how can ensure that these models continue to follow and reason from these instructions towards a desirable output?
