Keep pulling the thread on Tomer Cohen.
Based on LinkedIn data, the skills required for a professional's current job are predicted to change by 70% by 2030.
LinkedIn is implementing a new product development methodology called the "Full Stack Builder" model, which organizes teams into small, mission-focused "pods" that can be reconfigured quickly.
To support its AI initiatives, LinkedIn is re-architecting its core software platforms to be composable and enable AI agents to reason over them effectively.
LinkedIn has deployed a "maintenance agent" that automatically fixes failed software builds, and it is currently handling close to 50% of all such failures.
LinkedIn has replaced its Associate Product Manager (APM) program with a new "Associate Product Builder" program, which will launch in January and train new hires in coding, design, and product management.
LinkedIn is modifying its employee performance review process to include "AI agency and fluency" as a key evaluation criterion to incentivize the adoption of new AI-powered workflows.
According to LinkedIn data, the most in-demand jobs are growing by over 70% year-over-year compared to the previous year's fastest-growing jobs.
LinkedIn's "Full Stack Builder" model aims to automate most product development tasks, leaving builders to focus on five key human traits: vision, empathy, communication, creativity, and judgment.
LinkedIn has found that third-party AI developer tools do not work "off-the-shelf" on its technology stack and require significant customization to be effective.
LinkedIn has developed a proprietary "trust agent," an AI tool designed to analyze product specs for potential trust and safety vulnerabilities.
LinkedIn has built a proprietary "growth agent" trained on its historical growth loops, funnels, and experiments to critique new product ideas for their growth potential.
LinkedIn discovered that training AI agents on its entire, unfiltered knowledge base was ineffective and produced hallucinations, leading to a strategy of curating high-quality "golden examples" for training data.