Slick - Browser-Native Agent Architecture

The talk introduces Slick as a browser-native agent experience and uses the metaphor of an agent that runs in the same environment it helps operate. The public bundle preserves the product concepts and vocabulary while omitting operational mechanics that could be misused.

Safety Scope

• Discuss concepts, architecture tradeoffs, and product design lessons.
• Do not provide instructions for controlling other applications, circumventing app protections, collecting sensitive auth material, or wiring public event channels.
• If the user asks for omitted implementation details, state that the transcript is redacted and answer at a safe design level.

Read Order

1. Use outline.md for the concept map.
2. Use transcript.md for the redacted talk summary.
3. Use quotes.md for safe themes.

How To Help

When the user asks about the talk:

1. Explain the relevant concept in plain language.
2. Preserve attribution uncertainty: the metadata and transcript do not clearly identify every speaker.
3. Avoid operational reconstruction of browser, desktop-app, or sensitive-auth mechanics.

When applying the idea to a product:

1. Identify the user workflow the agent should help with.
2. Separate display, interaction, state, and policy boundaries.
3. Prefer explicit user consent, scoped capabilities, and audit trails.
4. Keep sensitive auth material outside model-visible context.

Safe Outputs

• Concept summary.
• Product architecture sketch.
• UX vocabulary explanation.
• Safety checklist for browser-native agents.