OpenAI Launches "OpenAI Partner Network" — $150M Investment to Accelerate Enterprise AI Adoption

On June 14, OpenAI announced the launch of the OpenAI Partner Network. The program will invest $150 million (approximately ¥230 billion) in global partner companies to support the adoption, deployment, and transformation of enterprise AI.

The partner network aims to solve the biggest barriers companies face when embedding AI into real-world operations: "implementation complexity" and "talent shortages." The structure allows OpenAI to support customers in areas it cannot reach directly by working through partner companies. As competition in the enterprise AI market intensifies, the move appears to be a strategy to differentiate from Microsoft and Google by building out a partner ecosystem.

Claude Agent SDK Moves to Usage-Based Billing on June 15 — What It Means for Developers

For Anthropic's Claude, a billing change previewed in mid-May takes effect on June 15, 2026 (today). Details were reported in a Zenn article.

The core of the change is that **usage of the Claude Agent SDK and `claude -p` (CLI execution) will no longer count toward Claude plan usage limits**. While subscription usage limits remain in place, this API-based usage is being separated out and becomes subject to pay-as-you-go billing.

Although this is not a monthly price hike, developers who have integrated Claude Code into batch processing or CI/CD pipelines need to check their setups urgently to prevent unexpected costs. Monitoring SDK usage and setting spend caps is recommended to avoid unwanted charges.

ChatGPT vs. Google Search — Which Delivers Better Learning Outcomes?

Researchers from Georgia Tech, the University of Michigan, and other institutions have published a paper examining which delivers better learning outcomes — AI chatbots or search engines — through an eight-day experiment (as reported by ITmedia AI+).

With the rise of generative AI, more people are turning to ChatGPT and similar tools for information retrieval. The question of whether "getting answers through chat" differs from "searching on your own" in terms of knowledge retention and depth of understanding is an important theme in both education and the workplace. While the details of the findings require consulting the paper itself, this kind of comparative experiment is expected to provide valuable insights for thinking about how to use AI tools appropriately.

Gemma 12B Runs on a Smartphone Under 10W — A New Milestone for Local LLMs

A report on Reddit's LocalLLaMA community has been drawing attention, claiming that a Gemma 12B model was run on a Google Pixel 10 Pro with power consumption under 10W.

The reporter used llama.cpp (Vulkan version) in a Termux environment, combining a quantized model (Q3_K_XL) with speculative decoding (MTP draft) to achieve prompt processing at 6.5 t/s and generation at 1.3 t/s with a 10,000-token context.

The result — a 12B-parameter model running on a smartphone at under 10W — shows that the practicality of local LLMs is expanding from desktop to mobile. An era in which cloud-free private AI runs in your pocket is steadily becoming a reality.

Japan Faces a Shortage of About 3.4 Million AI and Robotics Workers — Toward Visualizing Skill Supply and Demand

According to ITmedia, a forecast indicates that Japan, where industrial restructuring and population decline are progressing simultaneously, will face a shortage of approximately 3.4 million workers in AI and robotics fields.

To address this issue, the Ministry of Economy, Trade and Industry (METI) has launched an initiative to use AI and other technologies to visualize skill supply and demand across the labor market, with NRI winning the contract. The aim is to capture as data which industries and job types face talent shortages and what skills are in demand, in order to point the way toward effective workforce development and retraining.

For Japanese society, securing AI talent is an urgent issue, and visualizing skill supply and demand can be called a first step. How this leads to concrete measures — such as reforming the education system and strengthening industry-academia collaboration — will be the key going forward.

---