#Japan #Tokyo #OpenAI #Microsoft #Anthropic

The New Era of AI Training with Screen Recording

Renewer, based in Shibuya, Tokyo, has recently released a comprehensive free guide titled "A Complete Comparison Guide to Teaching AI Tasks via Screen Recording". This guide, consisting of 62 pages, dives into the recently launched screen recording features from three major players in AI—OpenAI, Anthropic, and Microsoft. From June to July 2026, these companies introduced their innovative solutions that transform how AI learns tasks, marking a significant shift in the landscape of AI training.

Background to an Industry Shift

In a remarkable development, just a month apart, OpenAI, Anthropic, and Microsoft unveiled their similar solutions:

• - June 18, 2026: OpenAI introduced "Record & Replay" via Codex.
• - July 21, 2026: Anthropic launched "Record a skill" through Claude.
• - July 29, 2026: Microsoft released "Skill Recorder" as an open-source software on GitHub.

Each of these features allows users to demonstrate their tasks on screen, which the AI then uses to generate retrievable skills. Unlike traditional robotic process automation (RPA) that records the coordinates of actions, these new capabilities capture the intent behind those actions. This shift from writing prompts to demonstrating tasks exemplifies a fundamental change in how we communicate work to AI systems.

A notable aspect of this transition is that all three functions output skills in a common open standard format, termed "SKILL.md" (Agent Skills). The synchronization of these competitors on such a standard signals a maturing market for AI task training.

Early Outcomes and Key Insights

Real-world applications of these newly revealed functionalities are already starting to surface, particularly from international users. One example highlighted the automation of a social media posting schedule that had been manually managed for two years. After recording a single session, the generated skill did not simply replicate mouse actions but intelligently chose appropriate methods step by step.

In another instance, the process of uploading 47 videos, which previously took 15 hours, was reduced to just 47 minutes with zero errors. The resulting skill allowed for dynamic adjustments during execution, demonstrating the incredible efficiency of these technologies.

Key insights from these reports suggest that simple and repetitive tasks yield the most significant benefits from this technology, while complexities involving conditional branching or creative decision-making should still be managed by humans. When adapting skills to different tools, it's crucial to allow time for necessary adjustments. Establishing these boundaries will guide users towards effective results.

Overview of the Guidebook

This guidebook compiles a wealth of information on the aforementioned capabilities and provides a neutral comparison across all three technologies. Key sections encompass:

1. Understanding Screen Recording for AI Task Training: An exploration of the shared mechanisms across the three technologies and the historical progression of AI’s capabilities in this domain.

2. Comparison of Features: A detailed analysis outlining why all three companies utilize SKILL.md, discussing utilization criteria, generated content, data management, and more with a comparison table across twelve axes.

3. Claude's "Record a Skill": Details on the workflow and limitations of converting demonstrations into SKILL.md format with important practical considerations.

4. Codex's "Record & Replay": Positioning of this first introduced feature and the governance within organizations through its three-stage process (Record → Draft → Replay).

5. Microsoft's "Skill Recorder": Insights into its open-source design philosophy and what elements are gathered in the cloud during analysis.

6. Best Practices for Recording Good Skills: Essential checklists and templates to follow before recording sessions, addressing common pitfalls and proposing practical voice modeling techniques to maintain focus during demonstrations.

The guide stresses the importance of articulating the "why" at every step in the process, a practice that applies universally across all three functionalities. Given the lack of existing Japanese resources summarizing these features, the guide serves a crucial educational purpose.

How to Access the Guide

Interested readers can download the guide for free by filling out the form on the following URL:
Download URL: Complete Comparison Guide

Release Date: August 26, 2026
Format: PDF (62 pages)

Additional Insights from the CEO

A deeper explanation from CEO Ryohei Horiuchi regarding these features can also be read on his note, where he elaborates on the differences between the three technologies and offers guidance on effective skill recording.

• - Note Article: "Teaching AI Tasks through Recording: An Exploration of the Mechanisms and Differences Between Claude, Codex, and Microsoft"

Learn More

Introducing Meboshi: AI for Task Visualization

To streamline the initial phases of using recording-based AI solutions, Renewer also introduces the Meboshi AI agent, designed to aid businesses in visualizing their operational tasks. By automatically documenting PC work and detecting patterns, Meboshi identifies tasks that are suitable for automation, providing data-driven insights into frequency, required time, and branching complexities. This tool generates necessary documents such as workflow diagrams and requirement definitions, establishing a firm basis for decisions on which operations to record and automate.

Renewer provides a no-cost program for businesses to preview Meboshi’s capabilities. Additional resources include:

• - Custom Trial Generation: An illustration of potential deliverables based on existing operational documents, generated within 2-3 business days.
• - Resource Kit: Comprising service materials, deliverable samples, and an investigation guidebook.

Meboshi Service Site: Explore Meboshi

Inquiries and Free Program Applications: Contact Us

About Renewer

Renewer is on a mission to make AI widely adopted across enterprises, positioning itself as a provider of AI products tailored for business needs. This includes the Meboshi tool and the prompt and AI agent assistant "Kono" for enterprise clients. The company’s guidebook series has reached over 1,600 organizations, including several listed on the Nikkei 225.

Recruitment Notice

Renewer is on the lookout for individuals who are excited to tackle the challenge of making AI common. The company values mutual understanding, welcoming part-time or contract engagements to start. Those interested in solving technical challenges or seeking new challenges are encouraged to reach out for a casual discussion with CEO Ryohei Horiuchi.

Schedule a Casual Meeting: Sign Up Here