Hasei Koki of Matsuo Lab to Speak at Offers Webinar on 'Agent Harness' for Engineers, July 9

Hasei Koki of Matsuo Lab to Speak at Offers Webinar on 'Agent Harness' for Engineers, July 9

Overflow Co., Ltd. (Headquarters: Minato-ku, Tokyo; CEO: Yuto Suzuki; hereinafter 'Overflow'), which operates the AI-era engineer recruitment platform 'Offers,' will host an online webinar on Thursday, July 9, 2026, as part of its engineer-focused webinar series '#Offers_DeepDive.' The event, titled 'How to Ensure and Continuously Improve AI Agent Accuracy? Matsuo Lab Explains 'Agent Harness' in Simple Terms,' will feature Hasei Koki, Director of AI Development at Matsuo Laboratory Co., Ltd. Overflow's Director and CTO, Tabito Ohtani, will moderate a deep dive into how to ensure AI agent accuracy and sustain an improvement cycle. Attendance is free.

▶ Registration (free): https://offers-jp.connpass.com/event/396197/

This event is recommended for:

Engineers who have heard the term 'agent harness' but want to clarify what it actually means

Those facing challenges in ensuring AI agent accuracy or building autonomous improvement cycles

Those who wish to reconfirm the fundamentals and differences between agent harness and context engineering

Event Overview

Title How to Ensure and Continuously Improve AI Agent Accuracy? Matsuo Lab Explains 'Agent Harness' in Simple Terms

Date and Time July 9, 2026 (Thu) 19:00–20:00 (Live Stream)

Format Online (Zoom)

Fee Free

Target Audience Engineers involved in or interested in AI agent development

Speakers Hasei Koki (Director, AI Development Division, Matsuo Laboratory Co., Ltd.)

Moderator Tabito Ohtani (Director and CTO, Overflow Co., Ltd.)

Organizer Overflow Co., Ltd. (Offers / #Offers_DeepDive)

Registration https://offers-jp.connpass.com/event/396197/

Background: Preventing AI Agents from Being 'Built, Run, and Forgotten'

As LLMs become increasingly integrated into society, products and workflows using AI agents are rapidly growing. However, few teams have a clear answer to the question:

How can we ensure the accuracy of an AI agent after initial deployment, and how can we build a system to continuously improve it?

Recently, the term 'agent harness' has emerged, but many still seek clarification on its definition and how it differs from context engineering.

At this event, Hasei Koki from Matsuo Lab—on the front lines of AI research—will clearly explain the origins and rising importance of agent harness, along with practical applications and future directions.

The first half (lightning talk) will clarify what agent harness is and why it has become critical alongside the evolution of LLMs.

In the second half discussion, the moderator and speaker will delve deeper into the topic, incorporating real-world implementation examples from Matsuo Lab. They will unpack the mindset and practical approaches needed to move beyond one-time tuning and achieve continuous improvement in agent performance.

What You’ll Learn

Background on what agent harness is and why it’s gaining attention

Differences from context engineering and how the concept of harness evolves with LLM advancements

Concrete use cases at Matsuo Lab and future development directions

Speaker Profiles

Hasei Koki (Director, AI Development Division, Matsuo Laboratory Co., Ltd.)

After working in R&D at manufacturing firms and an AI education startup, Hasei joined Matsuo Laboratory in April 2024. As a senior data scientist, he worked on generative AI-based solutions. He now leads AI solution projects and team management.

Tabito Ohtani (Director and CTO, Overflow Co., Ltd. | Moderator)

After researching and developing route search engines, Ohtani joined CyberAgent, Inc. in 2010, serving as system development and operations lead for the Ameba division and contributing to organizational management as a division board member. In 2013, he joined M-Tech Inc., where he led infrastructure development for payment systems and AI analytics platforms and served as system development lead, experiencing the company’s IPO in 2015. He co-founded Overflow Co., Ltd. as CTO.

Related: Offers AI Harness Open-Sourced

Offers AI Harness Released as Open Source

On June 10, 2026, Offers released 'Offers AI Harness,' a skill set enabling engineers to directly access Offers’ job search, salary benchmarking, and profile optimization features from their everyday AI coding agents (e.g., Claude Code, Codex), on GitHub under the Apache 2.0 license. By simply asking their local AI agent, 'Find me suitable job openings,' users can receive matched job listings, skill gaps, and estimated salaries based on their registered profile.

▶ GitHub Repository (Apache 2.0): https://github.com/overflow-tm/offers-ai-harness

▶ Service Details & Implementation Guide (Offers AI Harness): https://service.offers.jp/worker/ai-harness/

▶ Press Release: https://prtimes.jp/main/html/rd/p/000000294.000053307.html

About Offers

Offers: AI-Era Engineer Recruitment Platform

'Offers' is an AI-era engineer recruitment platform. Its AI analyzes technical skills, experience, and aspirations from multiple angles, visualizing market value and next career steps. Users can set the latest job types and skills in their profiles and receive offers from companies that resonate with them. Since its official launch in September 2019, over 35,000 engineers nationwide have registered, and the platform has supported recruitment for more than 1,000 companies.

Offers: https://offers.jp/

About Overflow Co., Ltd.

Company Name Overflow Co., Ltd.

Headquarters 5F, Shin-Toranomon Jitsugyokan Building, 1-1-21 Toranomon, Minato-ku, Tokyo 105-0001

Representative Director Yuto Suzuki

Founded June 9, 2017

Capital 208 million yen

URL https://overflow.co.jp/

Media Contact

Public Relations, Overflow Co., Ltd. TEL: 03-6555-2239 E-mail: pr@overflow.co.jp