AI + Careers Briefing — October 17, 2026

Today's stories share one question: who does the work, and who controls it, once AI is in the loop. Anthropic's own data shows fewer junior people in its technical teams and more time spent managing Claude. Replit swapped its default AI model, a reminder that the tools you learn can change under you. Meta pulled back contract language that could have let it train on business accounts' web content. And in Philadelphia, a protest against a sidewalk delivery robot ended in an arrest. For job seekers, the durable assets are judgment, portable skills and control over your own work.

Is Anthropic's own workforce getting less junior?

Yes: Anthropic says the share of junior-level employees in its technical departments slipped from 21 percent to 14 percent since early 2025, according to TechCrunch and Anthropic's own report.

The proportion of junior technical hires fell faster, from 27 percent to 9 percent. Technical departments' overall share of employment fell from 61 percent to 56 percent over the last year. TechCrunch reports that Anthropic attributes the broader shift to the "natural expansion of business functions as the company scales" and to coding tools that let non-technical employees build software themselves.

In an April 2026 survey, an open-ended question asked how AI had affected day-to-day work. Anthropic says 27 percent described productivity gains from Claude tools, and 20 percent described their role shifting "from actually doing work to managing Claudes." Based on calendar data, the company says the median employee spends 20 percent more time in meetings than a year ago. TechCrunch reports the increase is 16 percent for technical contributors.

Why it matters for job seekers: Our read: this is one employer's data, not a market-wide measure, and Anthropic ties part of the change to business growth, so it does not prove AI is replacing junior roles. It is still a useful signal because the company building these tools reports fewer junior hires and more people describing their work as managing Claude. If you are early in your career, expect interviewers to probe for judgment, review skills and the ability to coordinate with other people, not only for task execution. Our entry-level hiring playbook covers how to position yourself in that market.

Did Replit just make OpenAI its default AI model over Anthropic?

Yes, according to TechCrunch: Replit CEO Amjad Masad said OpenAI is now the preferred partner for powering Replit Agent, while users can still choose Anthropic's Claude or other supported models.

Masad made the comments on stage, TechCrunch reports, saying "I think ChatGPT is better now." He described the decision this way: "It was a decision based on both cost, partnership, and performance." He also said Replit publishes a public benchmark on which Sol and Astra, the models he named, consistently perform better at building vibe-coded applications from scratch.

Masad noted that Replit had been, until recently, a top-five customer of Anthropic's. In a statement to TechCrunch, he said Replit remains model agnostic.

The story is not fully settled. TechCrunch reports that one person familiar with the matter, who asked not to be named, said the partnership with Anthropic was still in place. The outlet originally published the piece on October 16 and later updated it with more context on Replit's partnerships with Anthropic and OpenAI, and changed its title.

Why it matters for job seekers: Our read: when a major AI coding tool changes its default model, the tool you learned last year may behave differently this year. That argues for skills that travel across vendors, such as writing clear prompts, evaluating outputs against a benchmark, and designing workflows that can swap one model for another. Replit's own statement frames the platform as model agnostic, and candidates can borrow that posture. Our research on the AI skills salary premium looks at which AI-related skills employers are paying for. When you list AI experience on a resume, name the task you accomplished and how you checked the result, not only the product you used.

Did Meta really remove the language that raised AI-training concerns?

Yes: Meta has taken out a provision in its commercial terms of use that legal experts said could let it scrape publisher websites and use the content for AI training, TechCrunch reports.

The removed language said Meta could use content from a business account's domains, web pages or URLs to provide, develop, improve, research, train and maintain the integrity of its products and services. Legal experts interviewed by Nieman Lab called it a "remarkable provision," according to TechCrunch.

After the backlash, Meta said the crawling terms it consolidated only applied to its ads and business products, and that it had clarified this in the terms themselves. Meta also said: "It does not change how Meta uses web crawlers."

TechCrunch notes that Meta won a high-profile copyright lawsuit last year, but other cases against AI companies, Meta included, are still continuing.

Why it matters for job seekers: If your portfolio, writing samples or case studies live on a domain tied to a business account on any large platform, the platform's terms can decide how that work gets reused. Our read: this episode shows terms can change quickly and get reversed just as fast, so it is worth skimming the terms before linking a work domain to a business account. Keep dated copies of your samples, and see our guide to building a portfolio that holds up in interviews.

What does a Philadelphia robot-blocking arrest say about delivery automation?

Police in Philadelphia arrested a retired University of Pennsylvania professor for disorderly conduct after she repeatedly refused to move away from an Uber Eats delivery robot, PhillyVoice reports.

Carolyn Marvin, 78, a professor emeritus, was leaving a flu shot appointment on a Wednesday afternoon when she spotted the robot on a Center City sidewalk, according to PhillyVoice. Police responded to a report of "criminal mischief involving a robot" on the 1200 block of Filbert Street around 1:16 p.m. Officers asked her to step aside so the robot could finish its delivery, and when she repeatedly refused, they arrested her for disorderly conduct. Marvin said she is exercising her First Amendment right to protest. PhillyVoice reports she has made a practice of blocking the robots to protest corporations that displace human delivery workers, and that this was the first time police got involved. Uber declined to comment.

The robots reached Philadelphia streets in March. 6abc reported that Uber Eats launched the service on March 10, using robots operated through Avride in a section of Center City between 10 a.m. and 10 p.m., with more than a dozen partner restaurants.

Why it matters for job seekers: The launch area and hours are limited, and neither article gives figures on couriers affected; 6abc's report does not address the impact on human delivery workers. Our read: this is a pilot-scale deployment, not proof of mass displacement, but the public protest shows delivery automation is now a visible local issue. If you rely on gig delivery income, treat it as a signal to build adjacent skills, such as customer service, dispatch or logistics coordination, rather than a reason to panic.

What to watch

• Whether other employers publish workforce data like Anthropic's. Its report covers one company, and part of the shift reflects business growth.
• Replit and Anthropic. TechCrunch's source says the partnership is still in place, so the split may not be final.
• The copyright cases still pending against AI companies, Meta included, which will shape how published work can be reused.