Why we built it

Track the signals. Understand the moves.

Overseas AI startups are moving into Japan faster and more seriously than many people realize.

OpenAI opened its first Asian office in Tokyo. Sierra opened a Tokyo office, acquired the Japanese AI startup OPERA TECH, and named SoftBank its exclusive partner in Japan. Andreessen Horowitz (a16z) chose Tokyo for its first office outside the United States.

Japan is no longer just a future international expansion opportunity. Companies are already hiring, partnering, building local teams, acquiring businesses, and developing go-to-market strategies.

But these moves are surprisingly difficult to track.

The earliest signals often appear in Japanese-language job postings, localized websites, event announcements, partnerships, customer stories, and other scattered sources—long before they become major international news.

For someone trying to understand the market, monitoring each company manually is time-consuming, and not every change is meaningful.

Japan GTM Radar aims to turn this fragmented information into continuously updated, evidence-backed GTM intelligence.

What it does

Japan GTM Radar is built to scale from five companies today to hundreds of overseas AI startups, revealing how they are entering, investing in, selling into, and expanding across the Japanese market.

It combines two complementary capabilities:

Nimble detects what changed. Jev determines why it matters.

Nimble continuously discovers and monitors relevant Japanese web sources, detecting changes in hiring, localization, events, partnerships, customers, and local presence.

Jev analyzes the evidence behind those changes, distinguishing routine updates from meaningful GTM signals. It evaluates whether a company is merely exploring Japan, actively investing in local operations, executing commercial activities, or expanding an established presence.

The goal is not just to track individual companies, but to build a continuously updated view of the broader market: who is showing early interest, who is committing resources, who is actively selling, and whose Japan strategy is accelerating.

How it works

1. Continuous discovery and monitoring with Nimble

• Nimble Search (country: JP, locale: ja) discovers relevant Japanese-language sources.
• Nimble Extract retrieves registered pages and Ashby job boards.
• Every fetch is stored with its URL, fetch timestamp, and page content.
• New snapshots are compared against previous versions to detect changes.
• Unchanged content does not generate duplicate evidence.

2. Evidence interpretation with Jev by TypeSafe

Jev uses systemOne typed questions to evaluate each piece of evidence:

• Company identity: Is this the target company or an unrelated company with the same name?
• Geographic relevance: Is the activity Japan-specific, APAC-wide, or global?
• GTM signals: Which of nine categories does it represent—jobs, Japanese-language support, events, media/visits, ambassadors, advertising, partnerships, customers, or local presence?
• Execution maturity: Is this merely an intention or plan, an executed activity, or continued/expanded activity?
• Strategic significance: What does the evidence imply about the company’s GTM stage and sales motions?

This makes it possible to identify meaningful commercial developments rather than treating every website update as equally important.

3. Evidence-backed intelligence interface

The application provides:

• A market readout showing GTM activity across monitored companies
• Company detail pages with evidence-backed GTM assessments
• An update feed highlighting newly detected developments
• Monitoring and review screens for tracking sources and resolving ambiguous cases

Every judgment links back to its source URL, supporting excerpt, publication date, and fetch date.

Model confidence and source reliability are displayed separately so users can distinguish the strength of an AI judgment from the quality of the underlying evidence.

Built-in guardrails

Japan GTM Radar is designed to avoid overstating what public information proves.

• Unconfirmed activity is not treated as zero; the system avoids unsupported scores and rankings.
• A job posting is not evidence of a completed hire.
• A translated website is not proof of a local team.
• Partners and investors are not automatically customers.
• A failed fetch is not evidence that an activity has disappeared.
• Reposted content is de-duplicated.
• Low-confidence and borderline company-identity judgments are routed to human review.

Real example: Nimble discovered a Tokyo company also named “CORGI.” Jev estimated a 0.60 probability that it was the target company. Rather than incorporating questionable evidence into Corgi’s GTM assessment, the system held it for human review.

What we verified

The current live implementation monitors five representative companies:

Sierra, Corgi, Braintrust, Cohere, and ElevenLabs.

Verified results:

• All five companies are processed successfully.
• A full monitoring run takes approximately two minutes.
• One run uses 21 Nimble API calls and approximately 130 Jev calls.
• Re-running the system against unchanged pages creates no duplicate evidence, judgments, or feed items.
• Detected evidence can be traced through the interface to its underlying sources.

These five companies are a working demonstration, not the intended limit of the product.

The longer-term vision is to monitor hundreds of companies continuously, using Nimble to surface changes across the market and Jev to identify which changes represent meaningful shifts in GTM execution.

Why it matters

Market entry is a process, not a single announcement.

A new Japanese careers page might be an early signal. A local sales hire may indicate resource commitment. A distribution agreement could reveal a new sales motion. A customer announcement, local acquisition, or office expansion may indicate a more advanced stage of market development.

These signals become far more valuable when tracked over time and interpreted together.

Japan GTM Radar is built around that idea: not just knowing which AI companies are in Japan, but understanding how—and how quickly—they are moving into the market.