Today's Headlines

  • Anthropic reports how AI changed its workforce: junior hires fell from 27% to 9% of technical hiring
  • Replit's CEO says the company is no longer working with Anthropic, and OpenAI is now the default for its agent
  • DeepSeek releases an updated V4.1-Flash under the MIT license, edging past Claude Opus 5 on some of its own benchmarks
  • Meta drops the training language from its revised commercial terms

AI reshaping who does the work, inside a company and across its partnerships, leads the day. Anthropic, where AI agents now log six times the hours of its employees, published figures on how its hiring and day-to-day work have shifted. Outside its walls, Replit's CEO said the AI app-building company had switched the default model behind its main agent from Anthropic to OpenAI.

Today's Top Three

Anthropic reports how AI changed its workforce: junior hires fell from 27% to 9% of technical hiring

Anthropic published a report on October 16, "How employment is changing at Anthropic," describing how AI use has reshaped its own hiring and work.

According to the report, junior-level employees made up 21% of Anthropic's technical departments in early 2025 and 14% now. Junior positions fell even faster as a share of technical hires, from 27% to 9%. Technical departments' share of the whole workforce dropped from 61% to 56% over the past year.

Over the same period, agent work expanded quickly. By August, interactive AI agent hours outnumbered employee hours six to one. The report calls human attention the bottleneck on agent output, estimating that even non-technical employees would spend half their work week reading everything their Claude agents produce.

Calendar data show the median employee spending 20% more time in meetings than a year ago, and technical individual contributors 16% more. Anthropic says it has set up an early warning system to track jobs facing AI-driven change and is prioritizing internal redeployment.

Replit's CEO says the company is no longer working with Anthropic, and OpenAI is now the default for its agent

Amjad Masad, co-founder and CEO of the AI app-building platform Replit, said onstage at TechCrunch Disrupt on October 15 that Replit is no longer working with Anthropic.

Masad cited cost, partnership and performance, and said, "I think ChatGPT is better now." Replit announced a partnership with OpenAI in August, and Masad said Replit had been one of Anthropic's top five customers until recently.

He later told TechCrunch that users can still choose Anthropic's Claude and other supported models, and that OpenAI is now Replit's preferred partner for powering Replit Agent. TechCrunch noted that because most customers stay with the default, the change could sharply reduce Replit's spending with Anthropic. TechCrunch also reported that a person familiar with the partnership said it remains in place.

DeepSeek releases an updated V4.1-Flash under the MIT license, edging past Claude Opus 5 on some of its own benchmarks

China's DeepSeek published the weights of DeepSeek-V4.1-Flash-1017, an update to DeepSeek-V4.1-Flash, on Hugging Face on October 17 under the MIT license.

DeepSeek says the update brings substantial gains in reasoning, coding, agentic tasks and visual benchmarks, while cutting task execution time and output length. Its recommended settings list a 1-million-token context window.

In the comparison table on DeepSeek's model card, the model scores 77.0 on the DeepSWE v1.1 software engineering benchmark, above Anthropic's Claude Opus 5 at 74.0. It also scores 60.5 on Automation-Bench v1.0.6, against 50.3 for Opus 5.

Opus 5 stays ahead elsewhere in the same table: 51.8 to 42.3 on Terminal-Bench 4.0, and 75.3 to 68.2 on NL2Repo-Bench.

Other Developments

Policy

  • In a follow-up, Meta removed language from its revised commercial terms, due to take effect October 30, that would have let it use crawled content to train AI and for other purposes. The clause authorizing crawling of businesses' websites remains, now limited to providing and improving advertising and business products. A Meta spokesperson told The Verge that site owners can still use robots.txt to signal their preferences. Meta removes language that may have allowed it to scrape news sites (The Verge)

Products

  • Users of AI agents including Meta's Muse, xAI's Grok and OpenAI's Dots have reported in recent weeks that the agents are blocked from completing tasks on major websites, The Verge reports. JPMorgan Chase and Charles Schwab confirmed they are blocking AI agents from accessing accounts over security concerns, and Delta Air Lines is preventing agents from booking tickets for users. Websites are closing their doors to AI agents (The Verge)

Business

Source: Selected by the editors from the AI news inbox (collected October 18, 2026: 11 items, 2 primary and 9 secondary).