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

On the record

dhiltgen

@dhiltgen ↗ 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: dhiltgen" wires/ | sort -r

Local LLMs

wire 2026-11-07

llama.cpp swaps in a smarter expert cache, Ollama ships a leaner Qwen3.8 Flash Next

MLX: make Qwen3.8 Flash Next memory efficient ollama/ollama

Local LLMs

wire 2026-11-05

Ollama honors YaRN in Qwen3, llama.cpp fixes quantized KV state copies

mlx: reject unsupported CUDA devices ollama/ollama

Local LLMs

wire 2026-11-04

vLLM splits prefill and decode graph sizes, GLM MTP norm aligned

mlx: load bundled CUDA libraries in-process on Linux ollama/ollama

Local LLMs

wire 2026-11-01

Ollama stops overriding model sampler defaults on /v1

openai: don't override model sampler defaults ollama/ollama

Local LLMs

wire 2026-10-30

Faster Metal MXFP4 decode, a GLM-4.7 streaming parser stall fixed

mlx: sort long top-k rows in chunks ollama/ollama

Local LLMs

wire 2026-10-27

Ollama fixes OpenCode v2 launch, llama.cpp rewrites CPU concat as one kernel

mlx: add DFlash2 support ollama/ollama

Local LLMs

wire 2026-10-26

Metal grid truncation fixed, vLLM patches GLM-5.3 weight loss

[Metal] Add sdpa_split kernels for mid rows for speculative decoding ml-explore/mlx

Local LLMs

wire 2026-10-23

Ollama fixes cloud metadata for every launcher, MLX speeds 6-16 row decode ~3x

[Metal] Add qmv_mma/qmv_nax kernels for 6-16 rows with large N/K ml-explore/mlx

Local LLMs

wire 2026-10-22

Ollama streams real errors, llama.cpp forwards sampling params on /v1/messages

[Metal] Add fast route for dequantizing fp4 to half in fp_qmm_t ml-explore/mlx

Local LLMs

wire 2026-10-18

vLLM v0.32.0 lands system-one endpoint; Ollama exposes template_source

api: populate show renderer/parser fields and add template_source ollama/ollama

Local LLMs

wire 2026-10-16

llama.cpp 0.7.0 ships MoE host cache, ollama reaps orphaned Windows runners

win: prevent orphaned runners ollama/ollama

Local LLMs

wire 2026-10-15

llama.cpp lands GDN chunked kernels and SYCL IQ3 reorder

MLX: version bump ollama/ollama

Local LLMs

wire 2026-10-14

llama.cpp server stack overflow, Ollama brings Metal kernels to CUDA

mlx: port the Metal custom kernels to CUDA ollama/ollama

Local LLMs

wire 2026-10-09

vLLM drops compile-based sequence parallelism, llama.cpp fixes slot checkpoints and a silent CUDA fallback

server: hide duplicate and downgrade guards from list (#18874) ollama/ollama

Local LLMs

wire 2026-10-08

Ollama drops the CLI account step, MLX fixes foreign buffer ownership

manifest: avoid symlinks on Windows ollama/ollama

Local LLMs

wire 2026-10-06

vLLM's request-controlled caches, and Vulkan sparse attention for quantized K/V

mlx: mitigate high latency after GPU idle ollama/ollama

Local LLMs

wire 2026-10-01

MLX ports Metal gather_qmm wins to CUDA, heads-up on a qmm_sm80 corruption bug

[CUDA] Add fp_gather_qmv to optimize gather_qmm ml-explore/mlx
[CUDA] Add gather_qmm_rhs_sm80 ml-explore/mlx
Improve pre-M5 metal memory usage for SDPA D512 ml-explore/mlx

Local LLMs

wire 2026-09-30

Ollama gates System One on declared capabilities, llama.cpp fixes a stale-UI trap

create: support explicit model capabilities ollama/ollama

Local LLMs

wire 2026-09-27

vLLM locks down multimodal kwargs, llama.cpp cleans up failed restores

Improve metal memory usage for SDPA D512 ml-explore/mlx

Local LLMs

wire 2026-09-26

Tiled k-quant matmul lands in llama.cpp, SYCL gets sparse FA

Improve metal memory usage for SDPA D256 ml-explore/mlx

Local LLMs

wire 2026-09-25

Ollama eases off typical_p, llama.cpp lands Vulkan int8 tensor cores

api: deprecate typical_p (#18627) ollama/ollama

Local LLMs

wire 2026-09-23

Ollama folds structured output into one pass, llama.cpp binds multiple addresses

mlx: speed up Qwen 3.8 prompt processing ollama/ollama
mlx: select Gemma 4 image resolution dynamically ollama/ollama

Local LLMs

wire 2026-09-15

llama.cpp 0.4.1 lands Maple 20B-A1B, Ollama drops typical_p

create: add server-side MLX imports and drop GGUF conversion ollama/ollama

Local LLMs

wire 2026-09-12

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

llm: keep gemma3n projector off the CPU ollama/ollama
llm: keep gemma3n projector off the CPU (#18376) ollama/ollama

Local LLMs

wire 2026-09-11

Ollama fixes unbounded KV memory growth, vLLM clarifies CVE timing

server: extract GGUF metadata and unify capabilities ollama/ollama

Local LLMs

wire 2026-09-04

Ollama enables speculative decoding under structured output, vLLM adds LoRA and fusion

ci: rebuild MLX macOS test payloads the release can't supply ollama/ollama

Local LLMs

wire 2026-09-03

Error propagation hardened in ollama's MLX bindings, gemma4 gains multimodal support

Report cached prompt tokens ollama/ollama

Local LLMs

wire 2026-09-02

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

Honor model generation defaults ollama/ollama
MLX, MLX-C: version bump ollama/ollama
ci: wire up MLX unit tests for PR runs ollama/ollama

Local LLMs

wire 2026-08-20

OLLAMA FIXES DEADLOCK THAT HUNG CHAT AND GENERATE

server: don't wedge chat and generate on a mid-stream parser error ollama/ollama

Local LLMs

wire 2026-08-16

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

mlx update (#17761) ollama/ollama

Local LLMs

wire 2026-08-15

QWEN3.8 LANDS IN OLLAMA WITH DEVELOPER INSTRUCTION SUPPORT

llm: transcode WebP images for llama-server ollama/ollama
qwen3.8: support developer instructions ollama/ollama
qwen3.8: add renderer and MLX import support ollama/ollama
renderers/qwen: tolerate non-leading system messages (#17757) ollama/ollama

Local LLMs

wire 2026-08-14

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

launch: add Muse Code integration ollama/ollama
model/renderers: match Muse Glimmer reasoning template ollama/ollama

Local LLMs

wire 2026-08-12

OLLAMA CUTS REPEAT PENALTY BLOAT, LLAMA.CPP SHIPS POCKET-TTS SUPPORT

nemotron_h: support the Nemotron 3.5 prompt layout ollama/ollama

Local LLMs

wire 2026-08-11

OLLAMA PATCHES CRITICAL PARSER BUG, NEMOTRON 3 LANDS WITH SPECULATIVE DECODING

parsers: recover boundary tokens fumbled into glimmer ATEM invoke names ollama/ollama
mlx: implement Nemotron 3 ollama/ollama
Release v0.32.7 ollama/ollama

Local LLMs

wire 2026-08-09

OLLAMA CUTS MODEL CONFIG BLOAT 100KB+ PER MODEL, LLAMA.CPP SHIPS TOOL ISOLATION AND CUDA FUSION GAINS

manifests: remove OCI rootfs from the model config (#17619) ollama/ollama

Local LLMs

wire 2026-08-08

OLLAMA SHIPS BLOCK-DIFFUSION SPECULATIVE DECODING, SGLANG LAUNCHES KIMI K3 DAY-0

openai: expand namespace tool declarations in the responses API ollama/ollama

Local LLMs

wire 2026-08-05

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

progress: fix data races on ticker, states, spinner, and bar state ollama/ollama

Local LLMs

wire 2026-07-30

OLLAMA CLEANS UP LINT PIPELINE AS LLAMA.CPP ADDS SPECULATIVE DRAFTING FOR GLM-5.2

CI: enable lint on the whole tree (#17457) ollama/ollama
lint: clean up current tree ollama/ollama

Local LLMs

wire 2026-07-25

OLLAMA LANDS LAGUNA SUPPORT AND CRUSHES MEMORY LEAKS WHILE SGLANG HITS V0.5.16 WITH CONFIDENCE-DRIVEN SPECULATIVE DECODING

model: add Laguna MLX support ollama/ollama
mlx: keep loaded model memory resident ollama/ollama
server: fix ps data race on scheduler loaded map (#17376) ollama/ollama

Local LLMs

wire 2026-07-23

OLLAMA AGENTS NOW IMPORT SKILLS FROM CODEX AND CLAUDE

model/parsers: finalize incomplete GLM tool calls ollama/ollama
model: align Laguna with upstream llama.cpp ollama/ollama