1 — Submit
Send the intake form on the front page. Nothing is charged at submission — every request goes to a human first.
Learn more
The machine, the process, what lands in your hands — and what happens to a codebase before research even starts.
Purpose-built research infrastructure with a human in command.
A run is a multi-hour to multi-day research campaign across everything your question touches — academic literature, archives, forums, code, and media, including audio and video. The strongest sources are read end to end, not skimmed. Every claim that survives is tied to its sources and its confidence, kept alongside the claims that contradict it — and where a claim can be tested, it is independently re-checked rather than quoted.
Quality is enforced, not promised: a report that has not earned its confidence does not ship. A human operator scopes every request before anything runs and signs off every report before delivery.
For scale: a recent standard run read 96 sources in full and mapped 477 sourced claims before a word of the report was written. Reports support a premise, weaken it, or dismantle it — negative evidence is kept because it prevents wasted direction.
How a run works.
Six steps, two of them human. No payment is ever taken before a person has read your request and agreed the scope with you.
2 — Scope review
We assess feasibility, boundary fit, and the right tier for the question. The scope decision goes to your contact email.
3 — Approval & payment
If approved, you receive a single-use payment link for the agreed tier, and a secure upload space opens in your portal for any repos or files. Nothing charges before approval.
4 — The run
The campaign runs for hours to days depending on tier, across every source type your question touches. The method is ours; the evidence trail it returns is yours.
5 — Human review
An operator verifies the report against the evidence, strips anything that overclaims, and signs it off.
6 — Delivery
Your report lands in your client portal, with status updates visible along the way.
Codebases get understood before they get researched.
Codebase requests begin with a dedicated understanding phase before any research fires. The repository is mapped and audited end to end, and what comes out is an understanding packet in four parts:
Full defect inventory
Every error, vulnerability, dead branch, and code smell found — filed and prioritised as a working TODO, not a wall of linter noise.
Health report
Architecture map, dependency risks, complexity hotspots, and the places where the codebase fights itself.
Build plan
A sequenced plan for fixes and next construction, grounded in the inventory rather than in opinion.
Grounded research seed
The packet becomes the research run's pre-context — discovery starts from your actual system, not from assumptions about it.
The inventory and build plan ship alongside the research report — for many teams the pre-research phase pays for the run on its own. The code package rides on any tier for +$150 NZD (typical repositories; larger codebases are quoted at scope review).
What you receive.
Every report is built from the same spine, scaled to your tier. You always know what the evidence supports, what it undermines, and what would change the conclusion.
Executive synthesis
Direct answer, bottom line, strongest evidence, weakest assumption, and practical research meaning.
Evidence landscape
Source clusters, disagreement zones, field shape, missing coverage, and quality signals.
Claim ledger
Important claims tied to source context, numeric care, approved wording, and source limits.
Mechanism map
Causal pathways, cross-domain bridges, implementation analogies, and where the bridge may fail.
Hypothesis cards
What might be true, why it is plausible, what would support it, and what would kill it.
Validation roadmap
Next studies, experiments, checks, replications, and evidence that would change the conclusion.
Choose your depth.
Tiers differ in how many runs are chained together and how far the strongest lead gets pushed — not in how carefully the work is done.
Tier 1 — Discovery · $150 NZD
1 full stack run, then human review. A first pass on a bounded question: 3–5 hour run, 2 day delivery.
Tier 2 — Deep Study · $250 NZD
A full stack run chained into a second run, concatenated, then reviewed. For questions that deserve a follow-through: 6–8 hour run, 2 day delivery.
Tier 3 — Professional · $600 NZD
Chained runs plus a dedicated deep study into the strongest lead, all research concatenated and reviewed. 12–24 hour run, 2–4 day delivery.
Tier 4 — Full Package · $2500 NZD
Two full stack runs, 2×2 chained, deep studies into the strongest leads, a final consolidating run, then human review. 1–2 week delivery.
Tier 5 — Corporate · $TBC
Corporate-level research and discovery tailored to your needs — scope, cadence, and delivery agreed directly. Email us for options.
What we've pointed it at.
The machine has handled nearly everything we've aimed it at — and where a question needs a capability that doesn't exist yet, we build it. Tuned, not templated.
Cold cases & investigations
Archival sweeps, timeline and entity mapping, and case files that preserve contradictions instead of forcing a story.
Scientific literature landscapes
Neuroscience to biology: what the evidence supports, where it disagrees with itself, and what nobody has measured yet.
Philosophy & foundations
From formal proof to cultural meaning — even "what is 2+2?" runs hundreds of sources deep here.
Codebase intelligence
Health, defect inventory, and build planning — then research on top of a system that has actually been understood.
Technical feasibility & architecture
Mechanism mapping and cross-domain bridges for people building something that doesn't exist yet.
Thesis direction & proposals
Evidence-grounded directions, gaps worth a career, and support for research funding proposals.
History & archives
Primary-source hunting across eras, jurisdictions, and formats — including scanned documents.
Media-sourced questions
Podcasts, talks, and video enter the evidence pool alongside papers — spoken claims get cited too.
Where the line is.
Every report is a speculative, evidence-backed starting point. It maps what the evidence supports, what it undermines, and what would change the conclusion — it does not certify decisions, diagnose, or stand in for qualified professionals. Out-of-scope requests are declined or reframed, high-stakes topics require explicit acknowledgement before approval, and human review gates every release. The full terms live on the Legal page.