News archive
Recruiting, HR & AI news: November 12, 2026
Thursday, November 12, 2026. 79 stories, each linked to its source.
Ranked by importance within each day. Pick a topic to see it newest first.
Thursday, November 12
From GDPR and Do Not Call rules to local job titles, EMEA data is harder to get right.
Get the news by email
Free newsletter
Get the Work Insiders weekly
One email a week: the recruiting and HR tech news that mattered, the tools we just added to the directory, and one practical guide. No spam, and you can unsubscribe in one click.
Hiring Lab analyzes the latest Indeed data each quarter to provide labor market insights for the retail sector.
Mirendil, a San Francisco AI lab launched in June 2026 by researchers from Anthropic, xAI, Google DeepMind and OpenAI, raised a $1 billion Series A led by Kleiner Perkins and Clearlake Capital, with Battery Ventures, Doerr Capital, Gil & Co, Google, M13, Pear VC and S32 participating. The company says one researcher working with agents pretrained Mirendil-Base, a competitive coding model, from scratch in 3.5 months for under $10 million in compute, and that it then post-trained it into Fern, which holds up to 10 million tokens of context. Its Lor platform, now in private beta, runs hundreds of agents that write code and run experiments for days at a time, and Mirendil says it uses Lor internally every day. The round follows a $200 million seed led by Andreessen Horowitz and Kleiner Perkins.
Plurai, a New York startup founded by former Nexar AI leads Dr. Ilan Kadar and Dr. Elad Levi, raised a $13 million seed round led by Team8 and made its platform generally available, free to start for AI engineers. Plurai builds a digital twin of a company's AI agent from its own knowledge base, tools and policies, then runs thousands of realistic and adversarial simulated users against it to catch hallucinations, policy violations, broken tool calls and data leaks, and generates fixes for what it finds. It also uses that simulation data to post-train small models tailored to each agent, which can run the agent more cheaply or act as always-on guardrails in production. The company says design partners cut development cycles from months to hours and reduced failures reaching real users by 82%.
Retell, a voice AI agent platform for contact centers, says it has passed $100 million in annualized run rate less than three years after launch, growing from $10 million to $100 million in 17 months while staying profitable and raising only $4.6 million in total. CEO Bing Wu says the bet is that the people running a contact center should build, run and improve their own AI phone agents without waiting on a vendor for every change. The platform serves more than 4,000 businesses, and OpenAI reported in October that Retell's monthly call volume grew from 30 million in early 2026 to more than 100 million by September. Alongside the milestone, Retell is launching Tell-1, a model built for live contact-center calls with sub-100ms responses, and a Dialer tool for outbound voice and SMS campaigns.
What an email verification API really costs at 10K, 100K and 1M calls: list prices from ZeroBounce, Emailable, Bouncer, MillionVerifier, Hunter and Tomba, and the billing rules behind them.
New York startup millionways launched Thorsten-4, which it calls a Large Psychology Model: it scores people on 50 behavioral and psychological traits from the conversations companies already have, including interviews, meetings, emails, support calls and earnings calls. The model was trained on roughly 7,000 structured interviews run by 50+ psychologists through a German nonprofit research program, with 35M+ words hand-tagged, and the company claims 90% predictive accuracy against established psychological assessments and participants' later behavior. Co-founders Martin Cordsmeier and Max Weidemann told Semafor they see executive search and sales as key uses, such as gauging from an interview transcript how a candidate behaves under pressure, and called it "a decision-making support tool rather than the ultimate decision-maker." millionways also disclosed $6M in US funding, with backers including NFL players Jason Kelce and DK Metcalf; Thorsten-4 is available via the millionways Platform, an API, MCP or self-hosted deployment.
The new brand Insightful builds on the strong legacy of Workpuls, reflects the company’s future vision and comes on the back of impressive 2.5x YOY growth.
Built primarily on OpenAI models, Amplitude’s proactive product agent, Wave, identifies what to build, helps teams ship changes with coding agents, and measures the impact.
Some teams are now paying over $1,100 a month just to maintain the enrichment they had with Clearbit.
Today, we’re introducing Gemini 4 Hydrogen, our daily driver for agentic workflows.
40 interviews a week and a $60 seat: Metaview pricing agent by agent, the free plan caps, what an AI-screened candidate costs in credits, the 24-hour vs 30-day notice gap and where Metaview…
Build faster with Gemini 4 Hydrogen, Google’s new model for cost-efficient agentic coding and debugging.
We are bringing Gemini 4 Hydrogen to Google Search globally to provide you with more helpful, visual, and intuitive responses.
Vercel has released UI SDK in public beta, an open source TypeScript library of more than 50 interface behaviors that developers or AI agents combine into custom components for React, Vue, Svelte, Solid and Web Components. In a Vercel test, an agent using the librarys bridge finished a settings task in 16 seconds for 0.032 dollars, versus 41 seconds and 0.073 dollars with ordinary browser tools.
Get quick answers with Gemini in Chrome. Summarize articles, compare products, and explain complex topics directly in your side panel.
Google launched the Gemini Decisions API in Google AI Studio, powered by a new model called Gemini Decision 1 that makes closed-set choices instead of generating text token by token. It can score up to 1,000 questions at once against text, images, audio, video and PDFs, with three decision types: yes/no, multiple choice with up to 255 options, and rubric scoring. Pricing is $0.10 per million input tokens with nothing billed for output, which Sundar Pichai says is about 5x cheaper than Flash-Lite, and Google says the model leads several benchmarks with the lowest calibration error. For teams screening documents, photos or applications at scale, it's a cheap way to run large batches of fixed checks; it has a 256K context window and can be tried free in the API.
OpenAI has acquired Odessia, an AI travel agent that planned and booked trips, and founder Francis Davidson, who previously founded Sonder, is joining OpenAI to focus on the future of travel. Davidson says he started Odessia less than a year ago and that tens of thousands of early adopters used it to plan more than $60 million a year of travel over the summer; the company was backed by Sequoia and angel investors. Odessia's standalone product has shut down, and terms of the deal were not disclosed. In his announcement, Davidson said the small team ran almost every function, from engineering to legal, finance and support, through AI agents, and his post drew more than 700,000 views on X.
Fireflies.ai launched the new Fred on November 11, expanding its meeting notetaker into an AI teammate that also sorts email, summarizes Slack, schedules meetings, records screens and searches across 240+ connected apps. The release adds a free scheduler, free unlimited voice typing (Fireflies Talk), free screen recording (Fireflies Clips), "FredBots" that take on ongoing tasks, and Voice Agents that can run candidate screenings, sales discovery and support calls with summaries and scorecards for review. Fireflies also gave a first look at Fireflies Fly, a small device that clips to the back of a phone to take notes on calls and in-person conversations. The company says it is used by more than 1 million organizations and teams at 75% of the Fortune 500; paid plans start at $10 per user per month billed annually.
n8n Labs, a new experimental team inside workflow automation company n8n, released Nod as a research preview: a standalone personal agent that runs on your Mac, Windows PC or own server with its own sandboxed computer and browser. Nod can prep calls, chase invoices, pull statements and cancel trials across connected apps like Gmail, Calendar, Drive and Slack, and on websites that have no integration at all. It is free and open source under Apache 2.0 and works with the AI provider you choose, including OpenAI, Anthropic, Gemini or a local model, so users pay only their model provider. n8n says a separate check confirms you asked before Nod sends, pays, posts or deletes anything, and that an email or web page can't give it orders.
Rox, which builds AI agents that research accounts, run outreach and grow revenue for large sales teams, raised a $100 million Series B co-led by General Catalyst and March Capital at a $1.3 billion valuation, with Sequoia, Databricks Ventures and others joining. CEO Ishan Mukherjee says the company reached that valuation only 12 months after launch and that customers have closed more than $7.5 billion in revenue since partnering with Rox. Its core customers are Global 2000 companies in medical devices, manufacturing and financial services, and Mukherjee cites 10x ARR growth since launch and 700%+ net revenue retention across enterprise accounts. Rox says it will use the round to double down on its applied AI team.
Alibaba's Qwen team has opened a preview of Qwen4-Max, which its documentation calls the next-generation flagship multimodal model in the Qwen family, built for coding, office productivity, front-end design and complex agent tasks. The model handles text, images and video with deep reasoning, web search, tool calling and structured outputs, and works with OpenAI- and Anthropic-compatible APIs. Access is through Alibaba's Token Plan subscription only, with no pay-as-you-go option, and early access is also offered on QwenWork, Qoder, the Qwen app and QwenStudio. The preview runs from November 11 to November 25, 2026 (UTC+8), after which Alibaba says this preview model will be retired; no benchmark table or parameter count was published with it.
Stripe has completed its acquisition of OpenRouter, the AI model gateway that lets developers reach more than 400 models from one API, closing a deal the two companies first announced on August 19. Stripe says it is "building the economic infrastructure for AI" and that, together with OpenRouter, it will help businesses explore new models and optimize token routing and usage. OpenRouter says more than 10 million developers and companies use it and that its commitment to neutrality across model providers stays the same. Stripe CEO Patrick Collison said OpenRouter is now serving about 125% more tokens per week than when the deal was announced; neither company disclosed the price, which Bloomberg had reported at more than $7 billion in August.
Vorflux, founded by Rippling co-founder Prasanna Sankar, has launched Vorflux 2.0, built around AutoBots: AI agents you hire for an ongoing role rather than a single task, such as on-call triage, cloud cost, bug hunting, project management or turning customer feedback into fixes. Each AutoBot gets its own Slack channel, where it takes requests, starts work sessions, sets its own schedules and brings decisions to people, all within a daily budget and approval limits the team controls. Teams can hire from a directory of ready-made AutoBots or create one with a goal, and can bring their existing Codex or Claude Code subscriptions. The company also said it recently closed a $15 million seed round backed by Y Combinator, Peak XV and Alliance DAO.
With Codex, Holiday Extras is turning customer expertise into better apps, higher booking conversion and new possibilities for growth.
No stories on this topic yet.





