Perplexity Integrates REA Reverse Engineering Tool into Computer
Perplexity announced the integration of the open-source REA (Reverse Engineer Anything) developed by morluto as a skill into Computer, allowing users to instruct the system to analyze application operation mechanisms and build similar functional versions without source code.
Tasks using the /reverse-engineer command are free until October 31, with each user eligible for up to 10,000 credits, which CEO Aravind Srinivas described as a guiding credit balance for users to attempt reverse engineering any application.
The official demonstration showcased five web versions generated by this skill, including design tools supporting layers, auto-layout, multi-user collaboration, and cross-device synchronization, spreadsheets with formula and chart interactivity, and image editors supporting layer masks and color adjustments.
Two additional tools are aimed at legal scenarios: a contract retrieval and version comparison tool (including original text citation markers) and an audio editor supporting audio clipping, mixing, fading, and WAV export.
REA itself is an MIT open-source MCP server that can connect to analysis tools like Ghidra and Hopper, handling native binaries, JavaScript, Electron applications, and web pages, returning evidence and limitation descriptions, but does not guarantee the restoration of original source code.
From a market perspective, this feature lowers the barrier to application reproduction, potentially enabling independent developers and startup teams to rebuild functionalities at low cost, with funding and attention shifting towards AI agents and open-source tools, while commercial software vendors face pressure from rapid functional approximations, benefiting open-source projects that provide analytical infrastructure.
Additional information shows that REA has accumulated over 300,000 stars on GitHub, with the analysis process running locally and accompanied by evidence records, and users can also independently install it for use in other agent environments.
Source: Public Information
ABAB AI Insight
morluto encapsulated existing reverse engineering tools (like Ghidra) into an MCP interface through open-source REA for AI agents to call, initially focusing on local analysis of binaries and JavaScript applications, quickly gaining significant attention on GitHub, transitioning from personal tools to standardized skills that can be orchestrated by agents.
Perplexity's capital path reflects embedding open-source skills directly into its Computer product, attracting users with limited-time free credits while showcasing replication cases to validate capabilities, expanding resources from model inference to application behavior analysis and reconstruction, forming entry control for the agent tool ecosystem.
Similar cases include the integration of early open-source decompilers and automated testing tools into IDEs, as well as attempts by code generation agents to approximate existing software functionalities; current AI application development is transitioning from model dialogue to executable skills and toolchain integration.
Essentially, this represents a technological substitution: AI agents replace some manual reverse engineering and prototyping processes through behavioral observation and tool invocation, relying on the open-source analysis layer to provide evidence, which is then used by generative models to construct runnable versions, thereby shortening the cycle from observation to implementation, but limited by the inability to fully restore proprietary logic and legal boundaries.
ABAB News · Cognitive Laws
Behavior is observable, but source code may not be restorable.
After open-source tools are integrated into agents, reproduction costs decrease rapidly.
Free credits serve as an entry point; long-term pricing depends on usage frequency.