SuperWizard

Inspiration

Most AI tools can talk, but they can’t execute. We wanted an AI that actually does the work in the browser: research, navigation, automation, and deliverables.

What it does

SuperWizard is an autonomous AI browser agent. Give it a goal, and it can research multiple sources, verify relevance, interact with websites, fill forms, and generate exportable outputs (PDF/DOCX/XLSX). It also shows live task progress so users can track every step.

How we built it

We built SuperWizard as a Chrome Extension (Manifest V3) with Vanilla JavaScript, HTML, and CSS. Core logic lives in a background orchestration layer with: 1)action routing, 2)autonomous workflow execution, 3)deterministic fallbacks, 4)source verification, 5)multi-format report export. Gemini powers reasoning/summarization, while custom browser automation ensures practical execution reliability.

Challenges we ran into

1)Dynamic and inconsistent web UIs 2)False-success behavior (opened a page but didn’t complete task) 3)Dead links / low-quality sources 4)Balancing autonomous power with safety and transparency

Accomplishments that we're proud of

1)Built a working AI operator (not just a chatbot) 2)Added multi-source verified research flow 3)Implemented dead-link detection + PDF fallback 4)Added DOCX/XLSX export 5)Built live task progress panel (“what AI is doing now”)

What we learned

LLM quality alone is not enough. Real product reliability comes from orchestration, guardrails, fallback logic, and clear user-facing progress states.

What's next for SuperWizard

1)Safer execution for high-risk actions (approval gates) 2)Better long-term memory/personalization 3)Deeper multi-app workflows (Docs, Sheets, CRM, marketplaces) 4)Stronger team collaboration and task history

Built With

• css3
• html5
• javascript