Back to Flash News

Anthony Pompliano: Software is about to shift from reactive to proactive execution

Crypto and tech investor Anthony Pompliano (@APompliano) posted on November 17, 2026, that software has been mainly reactive over the past half century and will in the future shift to proactive execution in ways that are currently unimaginable.

This post is an independent expression of opinion and did not include a specific product launch, technical white paper, or data support. Pompliano has recently continued to publicly discuss the impact of AI agents on software interaction and workflows, including his personal use of AI tools to build internal systems (such as a CEO operating system, daily briefings, and cross-tool integration), as well as the trend of agents directly calling the internet and various products on behalf of users.

Original source:

  1. Title: Software has been reactive for half a century. It is about to become proactive in ways we can’t imagine.
    Publisher: Anthony Pompliano (@APompliano)
    Publication date: 2026-11-17
    URL: x.com/APompliano/status/2122849208940322970
    Nature of source: personal public opinion post
    Verification status: full original post read, content and timestamp confirmed
    Supporting content: direct statement about the history of software being reactive and expectations for proactivity

ABAB AI Insight

Pompliano’s statement corresponds to his long-observed path of AI agent evolution. Traditional software relies on user-initiated commands (clicking, typing, triggering rules), while current large models and tool-calling capabilities allow agents to continuously monitor context, proactively initiate actions, and complete cross-system tasks. The personal systems he has recently publicly built (including calendar integration, to-do tracking, and cross-tool synchronization) are early examples of this type of practice; the core change is that agents can autonomously advance workflows without the user explicitly issuing instructions.

This view is consistent with industry discussions of “Internet 3.0” or agent-first interaction: users no longer directly operate multiple interfaces, but instead communicate with agents through natural language, with agents handling retrieval, decision-making, and execution. Pompliano has previously emphasized that people without a technical background can quickly build internal tools with existing AI tools and significantly improve execution speed, while also noting that companies remain cautious about fully delegating authority to agents.

From a business perspective, proactive software may compress the value of traditional SaaS seat-subscription models, while increasing demand for reliable tool calling, permission control, and observability. Pompliano’s observations point more to a directional judgment than to already-deployed large-scale results; most companies are still in the “human in the loop” stage, and agent permission boundaries and error costs are the main constraints on practical deployment.

ABAB News · Cognitive Law

  1. This opinion post reflects the poster’s judgment about the capability boundaries of AI agents and cannot be directly equated with the product state or revenue impact already achieved by the industry.

  2. The actual effectiveness of “proactive” software depends on permission control, error handling, and user supervision mechanisms, not simply on improvements in model capability.