Tag
This paper investigates how instruction-tuned Transformer models (LLaMA and Mistral) encode the discourse relations of causation and antithesis using interpretability techniques on next-token prediction.
A detailed blog post from Antithesis describing how their software testing tool was used to explore and achieve 'rebirth' in Tetris by reaching level 255, overcoming challenges and discovering glitches along the way.
A deep dive on Antithesis, a multiverse debugger for large distributed systems that offers deterministic replay and fault injection, now available as a free article.
Hillel Wayne announces version 0.14 of his book 'Logic for Programmers' and that he will join Antithesis as a developer educator, with plans to continue his newsletter but with a shifted focus and possibly lower frequency.