🗞️ Perplexity just dropped Computer, a new system that handles complete workflows from start to finish.

2025-08-21 · Source: Rohan's Bytes · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems, Software Development & Engineering · Depth: Intermediate, medium

Summary

Perplexity has launched Perplexity Computer, an advanced system designed to automate complex workflows by breaking down large goals into smaller, manageable tasks. This "digital worker" can research, design, code, deploy, and manage projects end-to-end, leveraging persistent memory, web access, and hundreds of connectors. It orchestrates up to 19 different models, including Gemini, Nano Banana, Veo 3.1, Anthropic Opus 4.6, and GPT 5.2, to perform specialized tasks in parallel. Available to Max subscribers for $200 per month, including 10,000 credits, Perplexity Computer aims to replace expensive tools like the Bloomberg Terminal. Concurrently, Anthropic updated its Cowork system, enabling Claude to function as a financial assistant across applications like Excel and PowerPoint, using plugins and connectors to platforms such as FactSet and MSCI. Anthropic also introduced a Remote Control feature for developers, allowing secure, seamless terminal handoffs between desktop and mobile devices for coding tasks, available for Claude Max tier subscribers. Andrej Karpathy highlights the rapid improvement of AI agents in software development, noting their ability to autonomously complete large programming tasks.

Key takeaway

For CTOs and VPs of Engineering evaluating AI integration, these advancements signal a shift towards highly autonomous, multi-modal agentic systems. Your teams should explore Perplexity Computer for automating complex, multi-stage projects and consider Anthropic's Cowork for specialized financial workflows. Additionally, the Remote Control feature offers significant mobility for developer teams, enabling continuous work across devices without compromising security. Prioritize pilot programs to assess the ROI of these new agentic capabilities against traditional manual or single-model approaches.

Key insights

AI agents are rapidly evolving to automate complex, multi-application workflows and software development tasks.

Principles

• Multi-model orchestration enhances task-specific performance.
• Persistent memory and context are crucial for complex agentic workflows.
• Secure remote access improves developer mobility.

Method

Perplexity Computer uses a manager-agent paradigm, breaking down large goals into sub-tasks and assigning the best of 19 models to each, running sub-agents in isolated environments.

In practice

• Use Perplexity Computer for end-to-end project automation.
• Employ Anthropic's Cowork for financial analysis across apps.
• Utilize Claude's Remote Control for mobile coding sessions.

Topics

• AI Agents
• Workflow Automation
• Financial AI
• Developer Tools
• Agentic Engineering

Best for: CTO, VP of Engineering/Data, Director of AI/ML, AI Engineer, AI Product Manager, Business Analyst

Related on AIssential

• Perplexity Computer is Here to Change the Way we Use AI — Perplexity Computer enables end-to-end AI workflow execution by orchestrating multiple specialized AI models and sub-agents.
• India’s First Air Taxi Factory Goes LIVE, Perplexity AI, NVIDIA & Anthropic Shock — Autonomous AI agents orchestrating multiple models are shifting AI from assistants to proactive digital workers.
• Personal Computer Setup - Perplexity — Perplexity's Personal Computer integrates an agentic AI with macOS for automated local and web tasks.
• Perplexity's 19-model AI ‘Computer’ — Perplexity Computer orchestrates 19 AI models for autonomous, long-running, sandboxed agentic task execution.
• 🗞️ Perplexity Unveils 24/7 “Digital Worker” Concept Powered by Mac mini — AI advancements are driving both specialized autonomous systems and an increased demand for human developers.
• Inside Perplexity Computer’s agent platform — AI agent orchestration, memory portability, and agent minimization are evolving, challenging established competitive dynamics and technical assumptions.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Rohan's Bytes.