Support and peer review

AI writes the code. ThinkSpark makes it research-grade.

Anyone can generate software now. Building a tool you can trust with real research, real participants and real data is something else. Members get hands-on support at every stage, each one built around a principle research already runs on.

Contact us to join

Illustration

Support at every stage of the research lifecycle

Seven stages, from first question to a tool that outlives the project. Each one is anchored to a research principle, so the support is about rigour, not just code.

  1. 01
    Scope
    Validity

    Measure what you mean to measure.

    Scoping sessions

    Turn a research question into a buildable spec, and agree up front what evidence would show the tool works.

    Prior-art check

    Find out what already exists before you build, so your tool adds something new to the field.

  2. 02
    Build
    Transparency

    Every step documented.

    The core
    The Starter Kit

    Our playbook for building with Claude Code, from first setup to the verification habits that keep AI-built tools honest.

    Member-led tutorials

    Walkthroughs from researchers who have already shipped, in your field's language, not an engineer's.

    Builder buddies

    Every new member is paired with someone who has taken a tool all the way to release.

    Build clinics

    Bring what you're stuck on and work through it live with people who have hit the same wall.

  3. 03
    Validate
    Reliability

    Right answers, every time.

    Validation studio

    Test the tool against known answers and gold-standard data. Code that runs is not the same as code that's right.

    Stress tests

    Edge cases, messy inputs and failure modes, found before your participants find them.

  4. 04
    Protect
    Integrity

    Participants first.

    Ethics and data desk

    Consent flows, data minimisation and guidance on ethics approval, designed in from day one.

    Security review

    People, not just AI, check AI-written code before anything reaches a participant.

  5. 05
    Review
    Peer review

    No tool ships unchecked.

    Lab peer review

    Six stages, modelled on journal peer review, from submission to published reports.
    Cross-disciplinary critiques

    Researchers from other fields try to break your tool and tell you what it misses.

  6. 06
    Publish
    Open science

    FAIR by default: findable, accessible, interoperable, reusable.

    Release and citation

    Open licence, DOI and citation file, so your tool can be found, reused and credited.

    Demo nights

    Show what you've built, get feedback, and meet the people who will use it.

  7. 07
    Sustain
    Reproducibility

    Built to outlast the project.

    Graduation

    A planned handover to a research group, with the documentation to run it for years.

    Maintenance watch

    When AI models or dependencies change, the lab flags what might break, before it breaks.

    Failure library

    Honest write-ups of what went wrong in past builds, so no one repeats them.

Standards

Built like research. Held to it, too.

Open by default

Every tool we release is open source.

Evidence before claims

Tools show what they know, how sure they are, and where it came from.

Reproducible

Anyone can trace how a result was produced, and produce it again.

Private by design

Tools collect only what the research needs, and nothing more.

Secure before release

Every release goes through a security review first.

Built to be handed over

Documented so a research group can own it long after we're done.

Peer review

Every tool is peer-reviewed. Exactly like a paper.

Our review process follows the journal model, step for step. A tool doesn't ship until it's been through all six.

  1. STEP 1
    Submission

    The author submits the tool with a methods note: what it does, how it was built, and what it can't do.

  2. STEP 2
    Editorial screening

    The lab's editor checks scope and completeness. Incomplete submissions go back before review begins.

  3. STEP 3
    Independent reviewers

    Two reviewers from outside the tool's team are assigned. Their identities stay anonymous to the author.

  4. STEP 4
    Reviewer reports

    Each reviewer tests the tool, assesses methods, correctness, security and usability, and recommends: accept, minor revisions, major revisions or reject.

  5. STEP 5
    Response to reviewers

    The author revises the tool and answers every point in a response letter. Reviewers check the revision.

  6. STEP 6
    Decision and publication

    Accepted tools are released open source, with the review reports published alongside and a DOI so the work can be cited.

Got a question worth building?

Tell us your field, your question, and the tool you wish existed. Your idea stays yours: we'll help you build it, or pair you with a lab member to build it alongside you.

Join ThinkSpark Contact us

hello@jointhinkspark.com