Local Stack

Debugging a home LLM rig the hard way: skills that wouldn't load, config edits that silently did nothing, and the failure modes that ate whole days before the real cause surfaced.

1. Zero of Twenty-Five Skills Were Loading (2026-08-06)
2. 64% of Our TTS Compute Synthesized Audio We Threw Away (2026-08-18)
3. The Fine-Tune Was Dead Weight. The Reference Clip Was Doing the Work. (2026-08-19)
4. 22% of Our Training Set Was Invented by Another Model (2026-08-18)
5. max_tokens=900 Yielded Zero Pairs. 3000 Yielded 59. (2026-08-19)
6. We Chased an Audio Bug for Months. The Instrument Was Broken. (2026-08-23)
7. Our Compression Logged Success and Compressed Nothing (2026-08-29)
8. Four Reasons Your Config Edit Didn't Take (2026-08-29)
9. The 429 That Was Really a 400 (2026-08-30)
10. Five Parallel Agents, Zero Commits (2026-09-06)