Private Alpha Tester Contribution Guide
Alpha testers help BIThub improve before public release.
This role is for invited users who are willing to test the platform, use the AI system, report problems, and contribute useful feedback.
Alpha testers are not just early users.
They are signal sources.
Who Alpha Testing Is For
Alpha testing is for users who can test BIThub with a real use case.
Good tester use cases include:
- using AI private messages for research or writing
- testing Constructs with real prompts
- testing Nodes or Workspaces when available
- testing CORE workflows when invited
- finding unclear guide language
- reporting broken links, bad routing, or permission problems
- creating useful topics that can become reusable knowledge
- testing marketplace or category workflows when enabled
Requirements
Alpha tester access may require:
- a BIThub account
- staff invitation or approval
- a specific use case
- willingness to report feedback
- willingness to follow platform rules
- enough participation to understand the basic system
Alpha tester permissions may change during testing.
Access is not permanent unless staff confirms it.
What Alpha Testers Should Test
AI Private Messages
Send private messages to AI Constructs.
Test whether the AI:
- understands the task
- follows instructions
- uses the right tone
- produces useful output
- cites or grounds claims when needed
- asks for clarification when necessary
- avoids overfitting or empty language
Constructs
Test different Constructs on the same task.
Look for differences in reasoning, style, reliability, and usefulness.
Report which Construct handled the task best and why.
Tools
Some agentic Constructs have tools.
Check the registry to see available agents and tool access:
Report tool issues such as:
- wrong tool selection
- missing source links
- bad search results
- weak grounding
- stale information
- failed calls
- confusing output
Guides
Read guides like a new user.
Report:
- stale screenshots
- broken links
- unclear instructions
- repeated sections
- missing next steps
- jargon without definition
- pages that explain the wrong thing
Categories and Permissions
Report access problems.
Examples:
- you can see something you should not see
- you cannot access something you were told to test
- a category description is unclear
- a user path is blocked
- a trust-level rule feels confusing
Nodes, COREs, and Workspaces
When available, test advanced AI surfaces with real tasks.
Do not only test toy prompts.
Useful tests include real research, analysis, writing, comparison, technical review, and workflow execution.
What Useful Feedback Looks Like
Good feedback is specific.
Use this format:
Problem:
What happened?
Expected:
What should have happened?
Where:
Topic, page, Construct, Node, Workspace, model, or link.
Steps:
What did you do before the problem happened?
Evidence:
Screenshot, quote, output, error message, or link.
Impact:
Why does this matter?