Working with Upstream Zero
This is a company, and this page is its commercial wing — structured as the seven questions a serious buyer asks, each answered by inspectable objects rather than copy. The register never changes: we measure and diagnose; we do not promise evaluator behavior.
1 · What do you study?
Commercial evaluation — how organizations are assessed, recommended, validated, and eliminated — observed today through AI evaluators. The research program lives at /research.
2 · What have you observed?
Honestly: nothing yet. The observatory is at First Light — zero published observations, and the front page prints that number. What exists is the instrument and its rules. Buyers deserve that answer straight.
3 · How might this affect my company?
If AI evaluators mediate what buyers in your category see (the open bridge hypothesis, H-1), then unverifiable claims and invisible capabilities cost you consideration you never knew you lost. Each research object carries a Commercial Relevance block translating it into buyer terms — without borrowing the object's evidence tier.
4 · What can you measure for us today?
- experimental
- experimental
Every capability is currently experimental — the build itself forbids marking a capability operational until it derives from published method. That enforcement is not a promise; it is a compiler error.
5 · What would a paid engagement produce?
- ENG-1 · provisional
Organizations that want to know how AI evaluators currently assess, recommend, validate, or eliminate them
- ENG-2 · provisional
Organizations whose capabilities are real but unverifiable by a machine reader
- ENG-3 · provisional
Organizations making decisions based on single-shot AI evaluation results
- ENG-4 · provisional
Organizations that need to know when evaluator behavior about them changes
- ENG-5 · provisional
Organizations building the structured, verifiable account of themselves that machine evaluators consume
- ENG-6 · provisional
Organizations willing to contribute anonymized evaluation data to the research program in exchange for early access to findings
Deliverable specifications are inspectable before buying: Evaluation Observation Report (specification draft). Engagements promise artifacts, never outcomes — the object model has no field an outcome-promise could live in.
6 · What is still experimental?
Everything commercial, currently — see the maturity labels above. And the measured-outcome register is empty:
No measured outcomes yet. Engagement results will publish here as tiered evidence, with client consent, as they are measured. There are no testimonials or logo walls on this site — this register is what replaces them. Slower, and worth more.
7 · How do we start?
A calm, specific path: what to bring and what a first conversation covers →