OpenAI Launches O3-Pro, Wikipedia Slams the Brakes on AI

Pivot 5
June 12, 2025

Pivot 5 is the must-read daily AI briefing for 500,000+ CEOs and business leaders who need signal, not noise.

5 headlines. 5 minutes. 5 days a week.

OpenAI Launches O3-Pro As Its Most Capable Reasoning Model Yet

OpenAI released O3-pro, a higher-performing version of its O3 reasoning model, and made it available to ChatGPT Pro and Team subscribers as well as via the developer API today. The model replaces O1-pro for these users and will reach Enterprise and Edu customers next week.
API pricing is set at $20 per million input tokens and $80 per million output tokens. Reviewers in OpenAI’s evaluations favored O3-pro for clarity, comprehensiveness, instruction-following, and accuracy, but the company notes it responds slower than O1-pro and currently lacks image generation, Canvas support, and temporary chats.
On internal tests, O3-pro beats Google’s Gemini 2.5 Pro on the AIME 2024 math benchmark and tops Anthropic’s Claude 4 Opus on the GPQA Diamond science exam. These results reinforce OpenAI’s assertion that O3-pro is its most capable model to date.

Read more here.

Chatbots are replacing Google’s traditional search results, sidelining the need for users to click on blue links. News publishers are seeing the referral traffic they depended on collapse as a result.
Google’s AI answers now surface directly in the search experience, satisfying queries without directing readers to external sites. The shift removes a pipeline that long fed audiences to news outlets, undercutting their visibility and reach.
Publishers who built business models around search-driven advertising suddenly face a dramatic shortfall in page views. The article calls the situation an “AI armageddon,” underscoring how generative answers threaten the financial stability of online news.

Read more here.

Doctors at Columbia University Fertility Center achieved the first pregnancy using their STAR AI sperm detection platform. The couple had endured 19 years and 15 failed IVF cycles before the March procedure succeeded.
The platform combines image recognition with a microfluidic chip to isolate rare sperm in azoospermia samples. It scans eight million images an hour and retrieved 44 sperm from one sample that human embryologists had deemed empty after two days of searching.
This milestone addresses a condition that causes about 10% of male-related infertility cases, offering an option beyond donor sperm or invasive chemical methods. The team also freezes multiple batches located by STAR so patients have backup sperm for IVF without extra procedures.

Read more here.

Close CRM simplifies your sales process by unifying all communication in one platform. With email, calling, and video integrated, your team can focus on selling and closing deals faster.
Tailor Close CRM to fit your unique sales workflow. Customize pipelines, automate tasks, and track performance metrics that matter to your business, empowering informed decisions and growth.
Trusted by over 6,000 businesses worldwide, including innovative startups and established enterprises. Proven to boost sales productivity and enhance customer relationships.

Anysphere, the company behind the AI-powered IDE Cursor, closed a $900 million Series C that values it at $9.9 billion. The raise coincides with the release of Cursor 1.0, which adds AI code review, background agents, and memory features.
Cursor’s annual revenue run rate has already exceeded $500 million just two years after launch. The tool is used by more than half of the 500 largest tech companies on the Fortune 500 and now handles up to 1 million transactions per second.
The article argues Cursor may be setting a record for the fastest revenue climb in developer tools. Reuters reports GitHub Copilot earned about $500 million in 2024, and the article says Cursor is on track to generate the same in 2025.

Read more here.

Wikipedia halted its pilot of AI-generated article summaries following immediate backlash from its editor community. The trial had only just begun earlier this month.
The feature showed a collapsible AI summary with a yellow “unverified” label at the top of every page for extension users who opted in. Editors warned the summaries could contain hallucination errors, noting that similar issues forced Bloomberg to publish corrections and scale back its own tests.
The pilot was paused just days after its announcement, showing how credibility concerns can quickly derail AI add-ons on open platforms. Wikipedia says it still hopes to use AI summaries to improve accessibility, so the idea is paused rather than abandoned.

Read more here.