$ git log --author="bbuf" --all --oneline

On the record

BBuf

@bbuf ↗ 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: bbuf" wires/ | sort -r

Local LLMs

wire 2026-11-06

Ollama stops silently truncating Responses inputs; llama.cpp cuts checkpoint copies

[Refactor] kernels/diffusion/tests: Prune new dead code and make silently skipped tests run sgl-project/sglang

Local LLMs

wire 2026-09-20

Metal MoE/SSM fusion lands, MiMo-V2.5 fp8 sharding fixed

[Cleanup] Deduplicate kernel tests, diffusion fixtures and benchmark helpers sgl-project/sglang

Local LLMs

wire 2026-09-06

vLLM closes scale-out multimodal handoff hole

[diffusion] fuse LingBot MoE group-limited top-k index selection sgl-project/sglang

Local LLMs

wire 2026-09-02

Ollama honors model generation defaults; llama.cpp and vLLM ship MoE performance fixes

[Kernel] Add tuned LFM2.5 Triton MoE configs on B300 sgl-project/sglang

Local LLMs

wire 2026-09-01

vLLM bounds validation errors, AutoRound FP8 arrives

[Diffusion] Fuse Qwen-Image residual norm and NVFP4 quantization sgl-project/sglang
[Diffusion] Fuse FLUX.2 gated residual normalization on Blackwell sgl-project/sglang
[Diffusion] Fuse Qwen-Image FP8 QKV projection and Blackwell epilogue sgl-project/sglang

Local LLMs

wire 2026-08-31

Vulkan top-k radix sort, HIP Q2_0 optimizations, and vLLM CUDA-graph memory fix

[diffusion] perf: absorb Qwen-Image output projection biases sgl-project/sglang

Local LLMs

wire 2026-08-29

OpenVINO backend lands Qwen3.5 on NPU, Whisper.cpp support

[diffusion] Keep Cosmos3 Nano resident on 96 GB GPUs sgl-project/sglang

Local LLMs

wire 2026-08-24

VLLM'S P2P WEIGHT SYNC CUTS RL MOE TRANSFER COSTS

[diffusion] Fuse LongCat-Image QKNorm and interleaved RoPE sgl-project/sglang

Local LLMs

wire 2026-08-14

OLLAMA LAUNCHES DEEPSEEK AND MUSE INTEGRATIONS WHILE VLLM SHIPS HARDWARE-AGNOSTIC MODELING

[diffusion] Wan2.2-TI2V: fuse per-token adaLN table add into contiguous slices + hoist rope cache (denoise -13.1% H100 / -12.6% H200, bit-exact; eager beats compile) sgl-project/sglang

Local LLMs

wire 2026-08-06

OLLAMA FIXES KV CACHE CORRUPTION, LLAMA.CPP HARDENS FILE SEARCH, VLLM KILLS MOE LEGACY CODE

[diffusion] Prefer cuDNN SDPA over FA4 for dense attention on sm_100 (B200) sgl-project/sglang

Local LLMs

wire 2026-08-05

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

[diffusion] FLUX.2 VAE decoder fast path behind quality=high (H200: 1024^2 97.6->29.2 ms, 2048^2 437.2->168.5 ms) sgl-project/sglang

Local LLMs

wire 2026-08-04

OLLAMA REWIRES OPENAI COMPATIBILITY, LLAMA.CPP HARDENS SAMPLING

[diffusion] Fix component accuracy topology reuse sgl-project/sglang

Local LLMs

wire 2026-07-29

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

Enable multimodal prefill BCG for VL and audio models sgl-project/sglang
[diffusion] refactor: remove stale kernels and dead code sgl-project/sglang