Verify our claims yourself.

The verification stack.

Three independent layers, each answering a different question. The first two ship in the public package; the third is on the roadmap as a paid service. If a claim doesn't survive all three, we don't ship it as a customer-facing number.

Reproduce a published row in 3 commands

Today (shape contract):

# 1. install
pip install ultracompress

# 2. pull a published artifact
hf download SipsaLabs/qwen3-1.7b-base-uc-v3-bpw5 --local-dir ./pack

# 3. verify SHA-256 reconstruction is bit-identical
uc verify ./pack
# → VERIFY: PASS — bit-identical reconstruction guaranteed.

This week (quality contract, v0.6.9):

# 4. reproduce the published PPL ratio on your hardware
uc bench-ppl ./pack --against-bf16
# → published: 1.00401×  measured: 1.0041×  match: PASS

Every published number in BENCHMARKS_2026_05_10.json traces to a source eval JSON in scripts/overlay/artifacts/ in the public repo. We don't ship round numbers without the underlying receipts.

The honest negative side.

Most projects publish the wins and quietly archive the failures. We catalogue both at the same level of detail, and the ratio is meant to be read as a feature: a research lab that hides nothing should expect a refutation count larger than its publication count.

The README's "What doesn't work yet" section is the same idea, applied to capabilities a reader might assume work because the rest of the package does. Both surfaces stay fresh.

What customers should ask us.

If you're evaluating a 5-bit lossless codec for production, these are the right questions to bring to the call. Our answers are the same in writing as they are on a sales call.

• "Can I reproduce your headline number on my own hardware?"

  Yes. Both contracts ship today in v0.6.9. The shape contract reproduces via uc verify; the quality contract (PPL ratio against the bf16 baseline pulled from HuggingFace) reproduces via uc bench-ppl --against-bf16. Same JSON schema as our published numbers.

  / self-serve · no Sipsa-side claims to take on faith

• "Is there an independent third-party audit?"

  Coming as a productized service in Q3 2026 as part of the Compression-as-a-Service catalog. Until then: every published number is fully self-reproducible by the verification stack above, and the source eval JSONs are in the public repo for any external auditor to re-derive.

  / scoped on the public roadmap · not a marketing future-tense

• "What about regulatory-grade verification (SR 11-7, FDA, defense)?"

  The on-prem MSA tier includes audit-trail logs, reproducibility manifests, and the bit-identical reconstruction contract that maps onto SR 11-7 model validation, FDA-regulated healthcare AI dev/deploy equivalence, and defense deploy-bit-exactness. See /enterprise for the compliance architecture and the three engagement tiers.

  / same SHA-256 contract, formalised inside your security boundary

• "What if the architecture I care about isn't in your verified matrix?"

  That's the Phase 0 POC. Bring a model, we deliver a UC pack inside 5 business days, and you self-verify it with uc verify + benchmark it with uc bench-ppl against your existing AWQ/GPTQ build on your hardware. If we don't materially help, you keep the diagnostic and we don't push Phase 1.

  / contact: founder@sipsalabs.com

For VCs specifically.

We know the verifier-quality gap is the #1 trust objection from technical diligence. Here's the timeline to close it — written in the present and near-future tense, not the aspirational quarter-plus-two.