njhill

@njhill ↗ first byline July 2026

Every morning, RepoJournal's newsroom reads what shipped across the open-source projects developers depend on — and writes the wire. Your commits kept making the news. This is your clippings file: the stories where your work was the story.

stories filed

daily wires

$ grep -rn "author: njhill" wires/ | sort -r

Local LLMs

wire 2026-10-31

Deterministic top-p under TP, KV constants stay on the GPU

[Config] Move V2-only checks into V1 model runner unsupported features vllm-project/vllm

Local LLMs

wire 2026-09-12

Ollama drops its CLI agent, llama.cpp cuts build times, vLLM fixes crash under DP>1

[MRV1] Scope breakable cudagraphs to the piecewise path only vllm-project/vllm

Local LLMs

wire 2026-08-26

OLLAMA'S MLX RUNNER NOW SPEAKS STRUCTURED OUTPUT

[Bugfix][MRV2] Run cudagraph memory profiling in a throwaway graph pool vllm-project/vllm

Local LLMs

wire 2026-08-16

LLAMA.CPP OVERHAULS INFERENCE FLAGS, VLLM PATCHES DEEPSEEK SPARSE MLA

[Bugfix] Pick the DeepSeek V4 eager cudagraph region per model runner vllm-project/vllm

Local LLMs

wire 2026-08-15

QWEN3.8 LANDS IN OLLAMA WITH DEVELOPER INSTRUCTION SUPPORT

[Perf] Avoid more GPU<->CPU syncs in multimodal encoders vllm-project/vllm

Local LLMs

wire 2026-08-05

OLLAMA FIXES RACE CONDITIONS AS LLAMA.CPP SHIPS BREAKING TTS CHANGES

[Core] Explicitly manage torch CPU threads in workers vllm-project/vllm

Local LLMs

wire 2026-07-29

OLLAMA RIPS OUT IMAGE GENERATION, LLAMA.CPP SHIPS SPECULATIVE DECODING

[KV Connector] Support NIXL heterogeneous P/D block sizes for hybrid models vllm-project/vllm