63 long-form pieces on architecture, training, inference, agents and safety.
Every major enterprise vendor now sells a way to build AI agents, and they are not competing on the same axis. The real choice is not which product is best. It is how much control you are willing to trade for how much convenience.
An A100 can execute 312 trillion half-precision operations per second but can only pull about 2 terabytes from memory in that same second. FlashAttention made attention fast not by computing less, but by refusing to touch memory.
A 70B model in FP16 needs 140 GB of memory it spends most of its time waiting to read. Dropping each weight to four bits cuts that to 35 GB, and for years that cut also broke the model. Here is what changed.
A model trained on 4,096 tokens can be coaxed into reading a quarter-million without retraining from scratch. The trick is not bigger attention; it is lying to the model about position, and splitting the sequence across a ring of GPUs.
The dream of automating office work is older than the personal computer. Tracing the line from VisiCalc through RPA to LLM agents reveals a recurring lesson: the tools that record steps break, and the tools that pursue goals are …
Training a model to generate photorealistic images teaches it geometry, semantics, depth, and object relationships. Google DeepMind's Vision Banana proves that a lightweight instruction-tuning pass over an image generator can bea…
Two servers run the same model on the same GPUs. One reports 4,200 tokens per second and is unusable; the other reports 2,600 and feels instant. Throughput is a property of the server, latency is a property of the request, and th…
Pure Mamba was supposed to replace attention. Instead the most efficient open models in 2025 are roughly seven-eighths Mamba and one-eighth attention. The reason is the KV cache, and what each layer can and cannot remember.
For years the field raced to build bigger models. Then a 2022 study showed most of those giants were undertrained, and quietly redrew the map of how to spend a compute budget.
Most agent frameworks treat state as an afterthought, tacking memory onto a chain of LLM calls. LangGraph inverts the design: the graph is the state machine, every node transition is checkpointed, and cycles are first-class citiz…
Chain-of-thought made models reason out loud, one word at a time. A new line of work lets them reason in the silent space between words, trading auditability for compute that does not have to be spelled out.
A 36.7-million-parameter graph network produces a 10-day global forecast in under a minute and beats the world's best physics-based system on 90 percent of verification targets. It also cannot start without that system's data ass…