Added:
- "Show HN: Llmbuffer – Python library for cache-optimized LLM conversation history" (4 points, by scottmp10)
- "Show HN: Magenta Real-Time Music Generation Locally on iPhone, Without the GPU" (6 points, by MediaSquirrel)
- "Show HN: Meadow Notes – extract and publish microsites from your Markdown graphs" (3 points, by gmccreight2)

Changed:
- **"Show HN: Extend UI"** moved from #2 to #1 and points increased from 81 to 123.
- **"Show HN: HelixDB"** moved from #1 to #2 and points increased from 68 to 81.
- **"Show HN: Atlasphere"** moved from #4 to #3 and points increased from 10 to 19.
- **"Show HN: Artie"** moved from #3 to #4 and points increased from 12 to 19.
- **"Show HN: Kctx"** moved from #5 to #10 and points increased from 3 to 5.
- **"Show HN: Ustps"** moved from #6 to #8 and points increased from 8 to 10.
- **"Show HN: macOS menu bar gauges for your Claude Code quota"** moved from #7 to #9 and points increased from 56 to 58.
- **"Show HN: NBSDgames"** moved from #12 to #12 (same rank) and points increased from 7 to 10.
- **"Show HN: GentleOS"** moved from #9 to #13 and points increased from 121 to 122.
- **"Show HN: Resonate"** moved from #14 to #14 (same rank) and points increased from 43 to 44.
- **"Show HN: Performative-UI"** moved from #15 to #15 (same rank) and points increased from 1157 to 1160.
- **"Show HN: Nucleus"** moved from #16 to #18 and points increased from 33 to 36.
- **"Show HN: Gitdot"** moved from #26 to #24 and points increased from 320 to 321.
- Several other items shifted positions without point changes, including: **Gravity** (8→11), **Learn while you wait** (13→19), **Drift** (17→20), **Social network** (18→21), **A curated collection** (19→25), **Construct SQL** (20→26), **Camel Mono** (21→27), **A Bluesky client** (24→28), **SNItch** (25→29), and **Tempis** (27→30). Items that remained in the same rank with no point change include **Cost.dev** (#22) and **Learn from 30 historical figures** (#23).

Removed:
- "Show HN: RiskKernel, kill -9 an AI agent and resume it without paying twice" (5 points, by prashar32)
- "Show HN: A 150M model that extracts verbatim evidence spans for RAG, no LLM call" (6 points, by justacoolname)
- "Show HN: SoulOS open-source replacing system prompts with stable state machines" (3 points, by ls1911)