plainOldCode

# Qwen3.8 Mia AI ↔ hibrid48 benchmark

## Executive summary

- The cluster deployment was faster in all 16 matched throughput cells, although the speed sweep is a **deployment comparison**, not an isolated checkpoint/quantization comparison. The old full Mia throughput sweep used MTP3/server cap 8; hibrid48 used MTP5/server cap 64.
- On the synthetic quality suite, hibrid48 passed 36/60 short coding attempts vs Mia/NVIDIA 40/60. The paired task-bootstrap interval for the −6.7 percentage-point difference is −18.3 to +5.0 pp, so this small suite does not establish a reliable coding-quality regression or equivalence.
- Long-context quality was 22/24 for hibrid48 vs 20/24 for Mia/NVIDIA. The paired interval is −8.3 to +20.8 pp (only 8 unique tasks); do not treat the point difference as a general capability claim.
- Both models truncated many coding attempts at the default 8,192-token budget. The README-recommended paired 16,384-token coding supplement is underway; hibrid48 is graded at 47/60, while the preserved 

plainOldCode

# 2026-09-27 Qwen 서빙 레시피 업데이트·벤치마크 종합 보고서

- 작성일: 2026-09-27, Asia/Seoul(KST)
- 범위: 기존 레시피 기준 측정 → vLLM 0.30 FP8 KV → BF16 KV → C4 고정 MTP3/MTP2 A/B → 설정·vocab·upstream 조사
- 최종 결정: **현 설정 유지 — vLLM 0.30.0 / NVFP4 가중치 / BF16 KV / C4 / MTP2**
- 이 보고서 작성 과정에서는 서버 설정을 변경하거나 재시작하지 않았다.

## 1. 요약

1. vLLM 0.30은 정상 적용됐다. 실행 중 핵심 소스 5개가 공식 v0.30.0 태그와 바이트 단위로 일치했고, v0.30용 MTP 패치와 47k vocab이 양 노드에서 적용된 것을 확인했다.

plainOldCode

# Three.js Logarithmic Solar System Demo — Implementation Prompt

Build an interactive 3D Solar System demo that runs in a web browser using Three.js.

The demo must visualize the Solar System from the Sun through Pluto, including major moons. It should use real astronomical data as its source of truth while applying logarithmic and visibility-enhancing display scales so that the complete system can fit within a single screen and still remain understandable.

## 1. Goal

Create a browser-based 3D visualization that allows users to inspect the Solar System from the Sun to Pluto at a glance.

plainOldCode

# Performance Take-Home 최적화 리포트

## 1. 최종 결과

| 지표 | 값 |
|---|---|
| **최종 cycle count** | **1,229** |
| **Baseline 대비 speedup** | **120.2x** (147,734 → 1,229) |
| **테스트 통과** | 9/9 (모든 correctness + 모든 speed threshold) |
| **`tests/` 변경** | 없음 (anti-cheating 준수) |

plainOldCode

#!/bin/bash

# This script interacts with the Gemini CLI and stores its output in Redis.

# Check if gemini CLI is available
if ! command -v gemini &> /dev/null
then
    echo "Error: Gemini CLI could not be found. Please install it."
    exit 1
fi

plainOldCode

-- Pull in the wezterm API
local wezterm = require 'wezterm'

-- This table will hold the configuration.
local config = {}

-- In newer versions of wezterm, use the config_builder which will
-- help provide clearer error messages
if wezterm.config_builder then
  config = wezterm.config_builder()

plainOldCode

{}

plainOldCode

🌞 Morning    36 commits  ██▎░░░░░░░░░░░░░░░░░░  11.1%
🌆 Daytime   158 commits  ██████████▏░░░░░░░░░░  48.8%
🌃 Evening   112 commits  ███████▎░░░░░░░░░░░░░  34.6%
🌙 Night      18 commits  █▏░░░░░░░░░░░░░░░░░░░   5.6%

plainOldCode

set nocompatible              " be iMproved, required
filetype off                  " required
set clipboard=unnamed

" set the runtime path to include Vundle and initialize
set rtp+=~/.vim/bundle/Vundle.vim

call vundle#begin()
" let Vundle manage Vundle, required
Plugin 'gmarik/vundle'

plainOldCode

set-option -g default-terminal "screen-256color"

set-option -g status-utf8 on
set-option -g status on
set-option -g status-interval 2
set-option -g status-justify "left"
set-option -g status-left-length 60
set-option -g status-right-length 90
set-option -g status-left "#(~/git/open-source/tmux-powerline/status-left.sh)"
set-option -g status-right "#(~/git/open-source/tmux-powerline/status-right.sh)"