OpenAI DevDay: Dots, Agents & $100B Opportunities
Greg Isenberg · 2026-09-29
💡 Quick Take
1. Use Dots (personal agent platform) to build mini-app plugins and capture a slice of upcoming transaction volume.
2. Play with the Decisions API in limited preview for real-time decision-making and classification.
3. Use the Agents API with computer use to build vertical AI-native applications and managed harnesses.
4. Use Sign in with ChatGPT to eliminate core app inference costs, grow top-of-funnel reach, and upsell advanced features.
5. Build real-world work APIs to connect AI agents to physical-world human specialists for last-mile task execution.
6. Build analytics and optimization layers for agent discovery to track plugin visibility, activation share, and prompt conversion.
📊 Detailed Explanation
OpenAI announced Dots at DevDay as their personal agent platform and direct answer to Meta's Muse and Instinct. Functioning as a major front door with access to 1.2 billion weekly active users, Dots relies on plugins—effectively mini-apps that ChatGPT selects when users ask for specific tasks, such as fixing a permit or finding a restaurant. While early plugin iterations received mixed reviews, OpenAI is doubling down on making ChatGPT an app ecosystem, creating massive commercial opportunities for founders who build plugins that successfully complete user requests.
The Decisions API was introduced as OpenAI's answer to Jev, enabling real-time decision-making by focusing intelligence on specific user-defined questions with finite, predefined answers. Developers supply context using text or images to classify content, route requests, or choose an agent's next action. Currently available in limited preview, it opens up powerful use cases across UI design, conversion rate optimization for app store screenshots, and direct response websites.
The Agents API now supports computer use, allowing developers to build agents that interact directly with software to complete complex tasks. It integrates Codex's multi-agent capabilities, tool search, tool calling, and context compaction while OpenAI runs the underlying infrastructure. This enables teams to build vertical AI-native applications and managed harnesses, provided they have proprietary data, deep vertical understanding, or strong brand trust.
A low-key yet massive announcement was the launch of Sign in with ChatGPT, which allows users to log into third-party websites and apps using their ChatGPT credentials and token allowances. Mirroring the historic growth impact of Facebook's early login integration, this feature allows developers to offer core application functionality for free without worrying about underlying inference costs. Businesses can build wider top-of-funnel acquisition, acquire user contact info, and upsell advanced features like team sharing, enterprise controls, or deeper data capabilities.
These launches unlock a wave of micro-categories and specialized workflow products—such as CAD file cleaners, scientific field researchers, Shopify catalog cleanup tools, and franchise contract review apps—that were previously economically unviable. Founders should structure their businesses around a four-step framework: trigger, decision, action, and feedback, focusing on owning the trigger, action, and feedback layers while selling specialized workflows rather than raw tokens.
🎯 Tech Expert Opinion
The strategic pivot by OpenAI to transition ChatGPT from a general-purpose chat interface into a fully fledged application ecosystem via Dots and plugin support is the most consequential platform shift for software founders since the rise of mobile app stores. By positioning personal agents as the primary interface for over a billion weekly active users, OpenAI is effectively creating a new distribution channel where traditional SEO and web-based acquisition take a back seat to agent discovery and intent-based tool calling. Founders who understand how to package proprietary workflows into reliable plugin mini-apps are positioned to capture unprecedented transaction volume.
The introduction of the Sign in with ChatGPT capability solves the single biggest economic hurdle facing early-stage AI application developers: the high cost of inference. Historically, offering a freemium AI app was financially punishing because every free user query directly incurred server costs. By allowing developers to leverage existing user token allowances and authentication graphs, OpenAI has effectively subsidized the top of the funnel, enabling a return to classic software-as-a-service playbooks where core utilities are free and enterprise-grade collaboration or advanced automation features are monetized.
At the same time, building businesses on top of rapidly evolving frontier APIs carries inherent platform risk, as underlying models frequently absorb niche features into native platform capabilities. To survive and thrive in this agentic era, technical builders must avoid wrapping commodity prompts and instead focus on proprietary data moats, complex multi-step execution workflows, and real-world API bridges that connect digital agent decisions to physical human labor. Those who invest early in building agent analytics, discovery optimization, and vertical harnesses will define the infrastructure of the next technology wave.
Kanal: Greg Isenberg