Work with me
Three fixed-scope engagements, each ending in a written result.
The research on this site is preregistered and reported even when it
reverses. You can commission the same kind of work. Each offer below has a
fixed scope, a fixed price range and a fixed deliverable. Nothing is
open-ended, and no result is chosen in advance.
OFFER 01
Claim audit
Test one result before you build on it.
- Who it is for
- A team with a model, analysis or benchmark result it plans to rely on.
- What I do
- Rerun the claim from your artifacts, add the adversarial controls it needs, and state what survives.
- You receive
- A written audit with every control's result, a reproducible script, and a plain verdict.
- Scope
- One claim, one system, up to four weeks.
$8,000 – $15,000
OFFER 02
Preregistered study
Find out what an intervention does before you ship it.
- Who it is for
- A team evaluating a steering, patching, ablation or protocol change.
- What I do
- Write and register the hypothesis and analysis plan first, then run the study exactly as registered.
- You receive
- The registered protocol, run logs, and a results report with controls and limitations, including any reversal.
- Scope
- One hypothesis, one comparator set, up to eight weeks.
$20,000 – $35,000
OFFER 03
Measurement instrument
A tool that shows how much of the picture survived.
- Who it is for
- A team whose system dynamics are described but never directly observed.
- What I do
- Build an instrument on your data that runs offline, with a fidelity figure reported on every view.
- You receive
- A working tool, tests, a fidelity note for each view, and handover documentation.
- Scope
- One system, one instrument, up to six weeks.
$25,000 – $45,000
How it works
Send a paragraph describing the claim, system or question. I reply within
a week saying which offer fits, or that none does. Prices are fixed ranges
for the scope above, confirmed in writing before any work starts. Expenses
and taxes are not included.
I do not guarantee a result, and I do not take on work for surveillance,
targeting, or security testing without authorisation. The
terms apply to everything produced here.