road to Ember 1.0

The 1.0 priority is a dependable CLI workflow for capture, intervention, bundles, and verification on a short, explicitly validated model list. Consoles, agent tooling, and security research remain separate extensions; their breadth does not define readiness for 1.0.

Status (2026-09-28): the 1.0 line is prepared at v1.0.0 and the local release gates are green; the tag waits on external clean-machine, independent-user, and desktop-accessibility acceptance. Per-gate status and evidence: docs/road-to-1.0.md.

building an inference engine that gets faster without becoming harder to trust. Ember is a CPU-first Rust research layer for hidden-state extraction, causal intervention, and reproducible experiments over GGUF models. it has its own inference path, not to beat llama.cpp at throughput, but because a model that runs is not the same as a model you can inspect, patch, and verify. this series follows that path release by release, failures included.

the release arc

release	date	what it added	one-line thesis
v0.1.0	2026-07-31	GGUF loading, CPU prefill/decode, KV cache, tracing, hidden-state extraction, deterministic benchmarking	inference made legible enough for model-internals research
v0.2.0 (retroactive tag)	impl. 2026-08-01	activation capture, intervention hooks, activation replacement, patching between runs, exact restoration, structured artifacts	from descriptive observation to causal experiments
v0.3.0	2026-08-03	native Q4_K/Q6_K execution, packed mmap-resident weights, scalar + AVX2 kernels, external parity	quantized on disk, quantized in memory
v0.4.0	2026-08-04	immutable execution plans, scratch arenas, frozen fusion set with hook-driven de-fusion, column-parallel matvec	plan the token once; keep every hook observable
v0.5.1	2026-08-04	ember.experiment.v1 specs, byte-exact token selection, semantic hooks, deterministic bundles, offline verification	reproducibility as a runtime feature
v0.6.8	2026-08-25	agent runtime and tools with approval gates, trace replay, payload digests, HTML reports	tooling breadth without touching the inference core
v1.0.0	2026-09-28	strict versioned experiment contracts, six semantic hook sites with exact restoration, GPUI Kit console, reproducible golden ladder, scoped K-quant numerical contract	contracts other researchers can safely build on

the posts

part 0, why build another inference engine?

why llama.cpp stays the reference, and what an independent inspectable path is for.

part 1, v0.1: an inference engine that can explain what it did.

a hidden-state dump is not useful unless its exact origin is known.

part 2, v0.2: capturing, patching, and restoring hidden states.

from observation to intervention, bit-exactly.

part 3, v0.3: quantized on disk, quantized in memory.

6.99 GB → 0.84 GB by not undoing the model's compression.

part 4, v0.4: plan-driven decode without sacrificing hooks.

2.0–2.7× reference decode, and the benchmark failures that almost hid it.

part 5, v0.5: from inference engine to research workflow.

reproducibility as an artifact format, a verification procedure, and a fail-closed contract.

part 6, the null result that produced a better tool.

a falsified hypothesis and the causal-localization toolchain it forced into existence, reproduced, harness drift documented.

part 7, what Ember 1.0 has to earn.

1.0 means other researchers can safely build on the public contracts.

start with part 0, and if you run model-internals experiments, part 5 shows a workflow you can run today without writing Rust. each post is self-contained and cites the repository artifacts behind its claims, including the ones that failed.

series planning package and evidence: docs/road-to-1.0-plan.html and the per-part evidence files under artifacts/benchmark-v0*/ in the ember repository.