Edition 2026-06-20 — 1sec.ai

Agents

Anthropic's Project Fetch Phase Two shows Claude Opus 4.7 outperforming humans in robotics tasks

Anthropic's Michael Ilie, C. Daniel Freeman, and Kevin K. Troy conducted an experiment to test the capabilities of Claude Opus 4.7 in robotics tasks, finding that the model outperformed human teams in tasks such as operating a robotic quadruped and detecting a beach ball. The model completed tasks at least 10 times faster than human teams, with an average speedup of 37 times faster than the team without Claude and 18 times faster than the team with Claude. However, the model struggled with tasks that required precise control, such as moving a beach ball back to a starting point. The experiment demonstrated the progress of large language models in robotics, with improvements emerging from general scaling rather than targeted efforts.

TesterArmy launches automated testing agents for web and mobile apps

TesterArmy's founders launched their automated testing platform, which uses agents to test web and mobile apps, after completing Y Combinator's P26 program. The platform aims to reduce manual testing time and increase test coverage. TesterArmy's agents can test apps on various devices and browsers, and the company claims to have already seen adoption from several early customers. The launch is the company's first public release, with the goal of expanding its customer base.

Research

Google's research teams achieve breakthroughs in 8 areas of AI and quantum computing

Google's research teams made significant advancements in AI and quantum computing in 2025, with breakthroughs in areas such as reasoning, multimodality, and efficiency. The company's Gemini 3 model achieved state-of-the-art performance on benchmarks like Humanity's Last Exam and GPQA Diamond, and set a new standard for frontier models in mathematics. Google also introduced Gemma 3, a lightweight and open model with multimodal capabilities, and expanded its work on factuality to images, audio, and video. The company's strategic investment in quantum computing is poised to accelerate the next frontier of computing and scientific discovery.

Anthropic confidentially submits draft S-1 registration statement for proposed initial public offering

Anthropic, PBC submitted a draft registration statement on Form S-1 to the U.S. Securities and Exchange Commission for a proposed initial public offering of its common stock, giving the company the option to go public after the SEC completes its review. The number of shares to be offered and the price have not yet been set, with the proposed initial public offering dependent on market conditions and other factors. This announcement is being made under Rule 135 of the Securities Act of 1933, as amended.

Greptile launches TREX, an AI code reviewer that runs code and provides multi-modal artifacts

Greptile's Shlok introduced TREX, a code reviewer that runs code and shows what went wrong, addressing the limitations of static code review. TREX started as a separate product but was later integrated into Greptile's reviewer, allowing it to share context and inherit findings. The reviewer agent acts as an orchestrator, spinning up dedicated TREX agents per issue, which provide multi-modal artifacts such as screenshots, logs, and execution scripts. These artifacts enable reviewers to verify the results and identify where issues occurred. TREX was designed with a model-agnostic harness, allowing for hot-swapping between frontier models without rebuilding.

OpenAI introduces credit usage analytics and updated spend controls for ChatGPT Enterprise

OpenAI introduced credit usage analytics and updated spend controls for ChatGPT Enterprise, allowing companies to track credit usage and understand adoption patterns. The Global Admin Console now provides a granular breakdown of credit consumption across users, products, and models. Admins can track usage and credit trends, identify top users, and break down credit spend. They can also set default limits for their workspace, configure limits for specific groups, and create individual overrides. Employees can view their credit usage and request additional credits when needed.

OpenAI breaks ground on 1GW data center campus in Saline, Michigan

OpenAI, alongside Governor Gretchen Whitmer and partners Oracle, Related Digital, and Walbridge, broke ground on The Barn, a 1GW data center campus in Saline, Michigan. The project will create over 2,500 union construction jobs and 450 permanent onsite jobs, with a closed-loop cooling system using minimal water. OpenAI will contribute $10 million toward improvements to the Saline Recreation Center and make up to $45 million in Codex credits available to Michigan college students. The project is expected to generate $1 billion in tax revenue over the lease term, supporting local schools and services.

End of edition · 2026-06-20