Independent AI research on agent reliability, evals, memory, and systems. Read more ↓
Cut a trained model at a chosen depth. Let the lower layers read the prompt and hand the upper layers a projected memory instead. The prompt gets 4x cheaper to process. Then we found the failure that …
We built a public benchmark for typed model decisions, tested a frozen 149M encoder with twenty small classifier heads on it, compared the result with TypeSafe's Jev, and fixed two bugs in the classif…
Here's a number that should bother you more than it does: the average npm package quietly trusts 79 other packages and 39 maintainers you've never heard of. Python is no better. Run `pip install flask…
We present the first demonstration of retrospective scientific hypothesis generation using a language model trained exclusively on literature predating a known paradigm shift. A 238M-parameter GPT-sty…
How we connected an M3 Max and an M4 MacBook Pro over Thunderbolt, built a distributed training cluster, and implemented sharded LoRA for tensor-parallel fine-tuning.
We present a method for training interactive, action-conditioned world models directly from mobile gameplay recordings, requiring no game engine access, emulator hooks, or manual annotation. Our appro…
In January 2026, Logical Intelligence announced they had built the world's first commercial Energy-Based Reasoning Model (EBRM). The system, they claimed, could reason about formal constraints in ways…
When you quantize an LLM to run locally on a Mac, the standard approach is uniform 4-bit. Every layer gets the same treatment. We wanted to know: what happens if you measure which layers actually need…
A few days ago, Dimitris Papailiopoulos posed an open challenge: what is the smallest transformer that can solve 10-digit integer addition with at least 99% exact-match accuracy? This kicked off a fun…
A Comparative Analysis of National AI Strategies in UAE, South Korea, France, Singapore, India, and Japan
We present a novel approach to Parkinson's Disease (PD) voice detection using steering vectors derived from wav2vec2 neural audio embeddings. Our method achieves 91.5% AUC for PD classification while …
