Meta Develops Consumer AI Assistant Hatch and Launches Instagram Shopping Agent
Meta is developing a consumer AI assistant codenamed "Hatch," similar to OpenClaw, aimed at transforming complex agent tools into ready-to-use products for the general public.
Hatch is currently powered by Anthropic Claude Opus 4.6 and Sonnet 4.6, and will switch to Meta's own Muse Spark model upon release. It has set up real website sandbox tests with DoorDash, Etsy, Reddit, etc., to validate autonomous decision-making and cross-dialog memory, with plans to start internal testing by the end of June.
Meta also plans to integrate an independent shopping agent into Instagram before Q4 this year, allowing users to click on short videos and autonomously navigate to external web pages to complete purchases, directly competing with TikTok Shop. Zuckerberg stated that the current open-source agent experience is too poor, saying, "I can't even trust it for my mom to use."
Source: Public Information
ABAB AI Insight
Meta has been heavily investing in the Llama series and collaborating closely with Anthropic since 2024. Previously, it used the Claude model for internal tool testing, and the Hatch project continues its "full-stack AI" strategy by validating real e-commerce processes in a sandbox environment, similar to Instagram's early iterations from images to shopping features.
In terms of capital strategy, Meta is shifting AI R&D resources from pure language models to the agent execution layer, distributing the shopping agent directly through Instagram's 2 billion monthly active users. The goal is to close the loop between user attention and purchasing behavior, increase e-commerce GMV, enhance advertising monetization efficiency, and reduce reliance on external APIs.
Similar to Google Gemini Agent's integration in Search and Shopping, and TikTok Shop's attempts at agentization, Meta is currently in a critical acceleration phase of transforming social platforms from content distribution to AI-driven transaction platforms.
Essentially, this is a restructuring of the industry chain: through Hatch and the Instagram shopping agent, Meta aims to reconstruct scattered agent tools and e-commerce processes into platform-level native capabilities, shifting capital from open-source model calls to its own Muse Spark for closed-loop control. Mechanically, it replaces manual browsing with AI agent shopping through autonomous decision-making and memory capabilities, seizing the pricing power of the next generation of e-commerce infrastructure.
ABAB News · Cognitive Law
No matter how strong the technology is, if ordinary people are hesitant to let their moms use it, it doesn't count as a real product.
The best agent isn't the smartest one, but the one that can seamlessly integrate into users' existing habits.
The ultimate success of a platform depends on who can turn AI from a toy into a tool that users automatically spend money with every day.