Estimated duration
~1h
Deadline
Thu, Aug 27, 2026
Location
📍 Remote only
USD
per approved result
One-time QA test run for https://siift.ai/.
What to do:
Test siift as a real user from login through active project use. The goal is to uncover obvious functional, UX, AI, and reliability issues across the full experience rather than test individual features in isolation.
Approach the test naturally: use a real or hypothetical business idea and try to make meaningful progress with it.
Work through at least the first several relevant phases of the project.
Do not follow a rigid script. Explore the product the way a genuine user would.
Use the main interactive functionality you encounter, including:
Check that every interaction produces the expected result and does not leave the UI in an unclear state.
Examples include:
Also treat poor AI behavior as a QA issue, including:
Flag anything that makes you think:
"I don't know what happened, what I'm supposed to do next, or whether this worked."
This includes confusing states, unclear feedback, unusually slow responses, broken formatting, unexpected redirects, disappearing information, or workflows that technically work but feel obviously broken.
For every issue, capture:
Title: Short description of the problem Severity: Critical / High / Medium / Low Location: Page, phase, feature, or workflow Steps to reproduce: What you did immediately before the issue Expected: What should have happened Actual:
Deliverable:
Tester requirements:
# siift End-to-End QA Test ## Objective Test siift as a real user from login through active project use. The goal is to uncover **obvious functional, UX, AI, and reliability issues** across the full experience rather than test individual features in isolation. Approach the test naturally: use a real or hypothetical business idea and try to make meaningful progress with it. ## Test Flow ### 1. Login & Project Setup * [ ] Log into siift from the website/app * [ ] Create a new project * [ ] Complete the initial project setup/onboarding * [ ] Add enough context about the business/project for siift to understand what you are working on * [ ] Confirm the project loads correctly and information persists ### 2. Work Through the Project Work through **at least the first several relevant phases** of the project. * [ ] Add and develop business ideas * [ ] Ask siift questions and have several back-and-forth conversations * [ ] Follow recommendations or next-step suggestions * [ ] Navigate between project phases * [ ] Add, edit, and review information in the structured workspace * [ ] Add ideas and research/evidence to the relevant tables * [ ] Revisit previously entered information and confirm it remains accurate Do not follow a rigid script. Explore the product the way a genuine user would. ### 3. Files, Data & Connections * [ ] Upload one or more files * [ ] Confirm siift correctly recognizes and uses information from those files * [ ] Connect at least one available external tool/data source, where possible * [ ] Ask questions that require siift to use the connected information * [ ] Confirm connected information appears to be incorporated correctly ### 4. Actions & Interactive Features Use the main interactive functionality you encounter, including: * [ ] Suggestions/recommended prompts * [ ] Action buttons / action runs * [ ] Add-to-table or similar structured-data actions * [ ] Research-related actions * [ ] Buttons generated within conversations * [ ] Navigation and phase controls * [ ] Any edit/delete/update controls encountered Check that every interaction produces the expected result and does not leave the UI in an unclear state. ### 5. Action Board * [ ] Open the Action Board * [ ] Create several tasks * [ ] Edit/update tasks * [ ] Change task status where supported * [ ] Confirm tasks persist correctly * [ ] Navigate away and return to confirm the board remains accurate ## What to Look For ### Functional Bugs Examples include: * Buttons do nothing * Actions return errors * Information fails to save * Pages or components fail to load * Incorrect information appears * Duplicate information is created * UI freezes or becomes unresponsive * Navigation breaks * Files fail to upload/process * Integrations fail * Actions produce unexpected results ### AI / Agent Problems Also treat poor AI behavior as a QA issue, including: * AI ignores instructions or important project context * AI appears unaware of information already provided * AI repeatedly asks for information it already has * Previous chat messages/context appear to be lost or improperly managed * Responses contradict known project information * Suggestions are unrelated to the user's current goal * Agent starts an action but never completes it * Agent hangs, stalls, or remains in a loading state * Tool/action runs silently fail * AI claims an action occurred when it did not * Information added through chat does not correctly propagate to the structured workspace * Changes in the structured workspace are not reflected in later AI responses ### UX / Reliability Issues Flag anything that makes you think: > "I don't know what happened, what I'm supposed to do next, or whether this worked." This includes confusing states, unclear feedback, unusually slow responses, broken formatting, unexpected redirects, disappearing information, or workflows that technically work but feel obviously broken. ## Bug Reporting For every issue, capture: **Title:** Short description of the problem **Severity:** Critical / High / Medium / Low **Location:** Page, phase, feature, or workflow **Steps to reproduce:** What you did immediately before the issue **Expected:** What should have happened **Actual:** What happened instead **Evidence:** Screenshot or screen recording where useful **Reproducible?:** Always / Sometimes / Once Pay particular attention to issues that can be reproduced consistently. ## Definition of Done The test is complete once you have: * Created and actively used a new project * Progressed through several phases/workflows * Had meaningful multi-turn conversations with the AI * Added structured ideas/research * Uploaded files * Tested available integrations * Used suggestions and agent/action runs * Used the Action Board and created tasks * Returned to earlier parts of the project to test persistence/context * Documented every meaningful bug, failure, stall, or clearly incorrect AI behavior encountered The goal is to find bugs.
Video assignments require spoken microphone narration throughout. Captions, written narration, and on-screen text are not substitutes.
Access to a device that can open the target URL
Report every issue with a severity and exact reproduction steps
Only submit evidence you captured yourself during this run
25-45 year old, web-native, familiar with AI tools, has a business idea
USD
per approved result