The New Interface: Why Jarvis Mode Changes How You Work

In this conversation, Jordan Wilson maps the rapid growth of agentic AI features from OpenAI, Anthropic, Google, and Microsoft. The core idea is that we have moved past simple chatbot interfaces into a Jarvis-like paradigm where AI agents interact directly with our operating systems, health data, and workflows. The hidden consequence of these updates is not just increased productivity, but a fundamental shift in the human-computer interface. For business leaders and operators, the advantage lies in moving from using tools to orchestrating agents that possess persistent, cross-application context. This shift rewards those who can define complex, multi-step workflows rather than those who simply prompt for single-turn text outputs.

The Shift from Interaction to Orchestration

The most significant development is the integration of voice-controlled, agentic workflows that operate across local applications. Wilson notes that OpenAI’s new desktop voice mode, powered by GPT Live, allows users to coordinate tasks across Slack, GitHub, and Notion without manual intervention. The non-obvious dynamic here is the App Shot functionality: the system captures not just a visual screenshot, but the entire programmatic context of open files, even those hidden from view.

"The cool thing is with this new, I kind of wish we could have got a fun name for it... you can have this app shot feature and functionality while you are using chat GPT live voice on desktop. It brings in all that other context without having me to always tell it, oh somewhere in this document, I think that I outlined something about that certain KPI."

-- Jordan Wilson

This creates a system where the AI acts as a persistent layer over your machine. The downstream effect is a reduction in the micromanagement of AI. Users no longer need to manually copy-paste or feed context into a chat window because the agent maintains a persistent awareness of the desktop environment.

The Hidden Cost of Fast Solutions

Wilson notes a recurring pattern in how frontier labs release features: they often prioritize speed and latency, sometimes at the expense of capability. For instance, Google’s release of Gemini 3.6 Flash and 3.5 Flash Lite focuses on token efficiency and low latency for search and document processing. While this provides an immediate benefit for high-throughput, agentic tasks, Wilson points out that these models are not smarter in a qualitative sense; they are simply more economical.

The systems-thinking implication is clear: developers must distinguish between models optimized for reasoning (the Pro series) and those optimized for execution (the Flash series). Relying on the latter for complex, multi-step workflows may lead to a compounding of errors if the model lacks the reasoning depth to handle edge cases, even if it performs the task faster and cheaper.

Where Immediate Pain Creates Lasting Moats

The introduction of Anthropic’s Record a Skill feature in Claude Co-Work reveals a competitive advantage hidden in the mundane. By allowing users to record screen activity, cursor movements, and voice narration to create a reusable skill, Anthropic has lowered the barrier to building institutional knowledge.

"Until now creating a skill meant either chatting with Claud and having it create one for you... The difference here with the teach quad a skill is well, you do not have to work through it. You just click start recording, you do your work, you dictate through it."

-- Jordan Wilson

This creates a feedback loop where the effort required to build a skill is significantly reduced. Over time, teams that consistently document their workflows into these shareable files build a skill library that acts as a moat. This is an investment that requires patience, as most users will skip the recording step, but it creates a durable, repeatable operational system that compounds in value as the team grows.

Key Action Items

• Audit your daily repetitive workflows: Identify three tasks you perform across multiple apps (e.g., Slack to Notion to Email). Over the next week, use Anthropic’s Record a Skill to automate these into reusable files.
• Transition to Jarvis mode: If you are a paid user, begin using the desktop voice interface for your daily planning. This pays off in 12 to 18 months as you become proficient in voice-driven orchestration rather than manual clicking.
• Implement App Shot hygiene: Start keeping your relevant project files open during your AI sessions. Learn to trust the agent to pull context from non-visible windows, reducing the time you spend manually gathering information.
• Optimize model selection: For the next quarter, strictly separate your tasks: use Flash/Lite models for high-volume, low-complexity document processing and reserve Pro models for high-stakes reasoning tasks.
• Centralize health data: If you have Apple Health, connect it to the new ChatGPT health feature to begin tracking trends. This provides a baseline of personal data that will become increasingly useful as the AI gains more sophisticated analytical capabilities over the next 6 to 12 months.