AI Apps — Try Them Live

Try these to see the agent use multiple tools at once:
"What day is it, calculate 18% of 5000, and tell me about AI jobs in Delhi"
"Add a high priority task: review my RAG app. Then show me all tasks."
"Convert $150 to rupees and tell me about Azure AI certifications"

Real use case: this is how AI decisions get an audit trail instead of a black box — one agent proposes, another challenges it, a third finalizes, and every step stays visible. It's the trust gap currently holding back autonomous agents in finance, legal, and compliance.

Try: "Should a startup build its own auth system or use a third-party provider?" · "Is remote work good for productivity?"

Real use case: an always-ready voice assistant pattern with zero per-minute transcription cost — speech-to-text runs free in your browser, GPT-4o-mini streams its answer back token by token instead of making you wait for the full response, and the conversation remembers context until you reset it. The pattern behind in-app voice support and hands-free assistants.

Works in Chrome or Edge · Try: "What is the capital of France?" then "What's its population?" — it remembers

Real use case: AI that works where the cloud can't or shouldn't be used: private documents, factory floors, spotty connections. The model is 2.7 GB and the questions never leave the machine. In the recording below I turn Wi-Fi off mid-demo and it keeps answering.

Honest numbers (laptop CPU, no GPU): first word in about 8–10 s once warm (about 45 s on the very first question while the model loads), roughly 3 tokens/s. It's slow and small, and it can get facts wrong. The trade-off is that it's free and private.

Want to run it yourself? Code + setup steps on GitHub →