Loading Runway...
New research, role-level analysis, and AI market intelligence. No spam.
The signals from the past week share a single thread: organisations are no longer asking whether AI can do the job. They are acting as though it already can.
ClickUp announced a mass layoff, explicitly replacing hundreds of employees with thousands of AI agents across internal operations and customer-facing roles. This is not a restructuring dressed up in AI language — the company named the substitution directly.
Epsilon India reported productivity gains with no headcount increase, specifically citing AI acceleration across software development, technical support, and customer onboarding. The framing is softer than ClickUp's, but the arithmetic is the same: more output, no new hires.
Debt collection is now being automated by AI systems handling calls and payment processing autonomously, displacing human collectors entirely in what Wired describes as one of the most undesirable roles in the workforce. The fact that this role was undesirable did not protect it.
Novo Nordisk is using AI to cut drug launch timelines by up to two-thirds, with its Bengaluru centre absorbing more global responsibility. That compression does not leave the same number of roles intact.
Virgin Atlantic used OpenAI's Codex to achieve near-total unit test coverage and zero critical defects against a fixed deadline in mobile app development — a concrete delivery result, not a pilot metric.
The DeepSWE benchmark placed GPT-5.5 significantly ahead of competing models on autonomous code generation, breaking what had been a performance plateau. Alibaba's Qwen3.7-Max can now run autonomously for 35 hours with memory persistence and external tool integration, including Claude Code.
Anthropic's Mythos Preview model identified over 10,000 critical bugs across more than 1,000 open-source projects in limited trials with 50 partners. That is the security audit and code review workload of a large team, automated.
OpenAI's Codex was named a Leader in Gartner's 2026 Magic Quadrant for Enterprise AI Coding Agents — a classification that signals procurement decisions, not just experimentation, across large organisations.
An AI-generated short story was submitted to the Commonwealth Short Story Prize — one of the most prestigious literary competitions in English. The literary world, as The Verge reports it, was not prepared for this.
China's AI-generated short drama market is expanding rapidly as studios use generative AI to compress production costs and increase output volume. Writing, editing, and production roles in short-form content are being automated at scale.
Google's shift to AI-generated search results triggered a 30% increase in DuckDuckGo instals as users rejected the change. This does not reverse the direction of travel — it confirms that AI is now the default output layer for one of the internet's most trafficked surfaces, which reshapes every business model built on search visibility.
Document your AI tool use with output data, not just activity. Virgin Atlantic's Codex case was credible because it cited test coverage and defect rates against a deadline. "I use AI in my workflow" is not enough. Record what shipped, how fast, and with what error rate.
If your role involves reviewing, summarising, or quality-checking output at volume, identify which part of that work AI cannot yet do reliably. Anthropic's bug-detection model found 10,000 critical issues autonomously. The residual value in security and review roles is in judgment calls the model escalates, not the detection pass itself.
Do not treat creative or clinical roles as inherently protected. The Commonwealth Short Story Prize submission and the radiation oncology LLM entering routine clinical practice are from opposite ends of the occupational spectrum. The common factor is that both fields assumed human authorship or judgment was the entry requirement. Neither assumption held.
Runway tracks AI career signals so you can act on evidence, not noise. careerrunway.ai