Start with a real problem.
Useful software begins with something worth making easier. The technology follows the question.
INDEPENDENT SOFTWARE PROJECT EARLY-STAGE RESEARCH
Software beyond
the obvious.
We’re exploring AI-assisted software testing
for developers and small teams.
Built with curiosity. Guided by evidence.
INDEPENDENT BY DESIGN.CURIOUS BY DEFAULT.
01 / THE THESIS
Less friction.
More possibility.
Developers still spend time manually repeating user journeys after a change. When something breaks, turning it into a clear, reproducible bug report is another task.
We’re investigating a testing assistant that takes a description of a user journey, explores a web app, and gathers evidence of unexpected behavior for a developer to review.
The planned Claude API integration would help interpret test goals, reason about what the agent observes, and draft useful findings. This is a research direction; no integration or product is publicly available yet.
AI as part of the workflow.
Human judgment at the center.
Human judgment at the center.
A PROPOSED TESTING WORKFLOWEXPLORATION 001
01
HUMAN
Describe the journey
A goal, constraints, and expectations.
AGENT LOOP
02 / REASON
Plan checks
from context.
03 / EXPLORE
Follow the
user journey.
04 / VERIFY
Compare with
expectations.
05
HUMAN
Review the evidence
Decide what needs attention.
INTENT → ACTION → EVIDENCE↳ REPEAT
02 / RESEARCH & DEVELOPMENT
One focus.
Room to explore.
Software testing is the primary research focus.
Developer tools and media are longer-term interests.
FIELD NOTES / 01PRIMARY RESEARCH
What if the feedback arrived while the idea was still fresh?
We’re investigating an assistant that follows a user-described journey through a web app, checks what happens, and captures reproducible findings. Claude is a planned reasoning layer for interpreting goals and observations. The integration is not live, and no public product is available.
01Explore a journey02Check the behavior03Capture the evidence
03 / THE APPROACH
A small lab.
A high bar.
Wayward Labs is an independent, founder-led software project. We’re exploring how AI can make building and testing software more useful, starting with the gap between a working feature and the evidence that it works as intended. The work is early-stage research, with no public product available yet.
Build to learn.
Put an idea into working software. Test the assumptions, study the rough edges, and keep iterating.
Keep people in control.
Make actions understandable, results inspectable, and important decisions deliberate.
GOOD THINGS START WITH A CONVERSATION.
A question
worth exploring?
Exploring a similar question?
We’d like to hear from you.