The lab

Notes, published with their data.

ENGINEERING NOTESUPDATED AUG 2026

Aelius runs Qwen3.6-27B, an open-weights hybrid-attention model, on a single machine we own. These notes document how — including the configurations that failed, the measurements that decided, and the traps that cost us time. Everything here is reproducible on consumer hardware.

NOTE 001 · JUL 2026 · INFERENCE

Serving 84K context on a 24 GB card

The context ceiling wasn't VRAM — it was a 1 GiB safety margin inside llama.cpp's auto-fit. Finding it bought 31% more context and gave up nothing.

Read the note →
NOTE 002 · JUL 2026 · SPECULATIVE DECODING

Speculation has a depth limit

Multi-token prediction at depth 2 runs 65% faster than depth 5 — deeper drafts evict a layer to CPU and halve their own acceptance rate. Measured, with seeds.

Read the note →
NOTE 003 · AUG 2026 · ARCHITECTURE

The economics of hybrid attention

Only 16 of Qwen3.6's 64 blocks pay per-token memory. That single design choice is why a 200-page conversation fits on one consumer GPU — and why Aelius can be free.

Read the note →