News archive
Recruiting, HR & AI news: November 8, 2026
Sunday, November 8, 2026. 32 stories, each linked to its source.
Ranked by importance within each day. Pick a topic to see it newest first.
Sunday, November 8
CostGraph, founded by former Twilio engineer Emmanuel Bakare, has launched its cost platform that pulls cloud, Kubernetes, CI-runner, SaaS and AI-provider spend into one view, attributes it to teams, workloads and customers with virtual tags, and tracks AI costs per token across providers. Its GraphAI assistant answers plain-language questions like why a cost went up or what is idle, using the customer's own billing data, and an AutoPilot feature can rightsize Kubernetes workloads, route GitHub Actions jobs to the cheapest runner that isn't slower, and place agent sandboxes by policy, with changes tracked against savings and applied only after approval. CostGraph also ships an MCP server so Claude, ChatGPT, Cursor, Codex and other agents can query spend directly, and it runs hosted or on-prem. The company says it is already generating revenue and is raising its first institutional round.
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.
Anthropic agreed in late September to pay about $3.4 billion for commitments covering 1.7 gigawatts of heavy-duty gas-fired generators from OnePWR Solutions, a San Antonio developer of behind-the-meter power plants, the Financial Times reported on Sunday, calling it the largest known single purchase of power-generation equipment by an AI lab. Anthropic took over OnePWR's existing Caterpillar generator orders along with the engineering and construction capability to install them, but did not buy the company, and no OnePWR staff moved over. The move follows Anthropic's amended S-1, which said some planned data centers rely on dedicated on-site gas generation and flagged long lead times for turbines and generators. Sites, delivery timing and how the $3.4 billion splits between equipment and installation have not been disclosed, and Anthropic and OnePWR did not comment.
Tactico has opened a pre-release download of Momentum, a desktop workspace with a "neutral" memory layer meant to let work move between AI coding harnesses such as Claude Code, Codex and Grok Build without re-briefing the next agent. You set a goal with criteria once, run it in whichever harness you like, then switch and type "continue"; Momentum keeps the goal, decisions and checkpoints and shows a ledger of which agent did what. The memory is stored locally, can be inspected or exported as JSON, and Tactico says it does not process or keep it; each harness still connects to its own model provider under that provider's terms. Tactico describes Momentum as still in development, and says it works with more than 30 harnesses.
Are your performance reviews difficult or burdensome?
OpenCode, the open-source coding-agent team at Anomaly, has made Browse available: an MCP server plus a Chromium extension that lets agents work inside the browser you already use, with your tabs, logins and extensions, instead of a fresh headless browser. Agents open their own tabs in the background, drive them with Playwright, take screenshots, record the screen and capture network traffic, and when they hit a login, 2FA prompt or payment they hand the tab back and wait until you click Continue. It works in Chrome, Edge, Brave, Arc, Vivaldi, Opera and other Chromium browsers and with MCP clients such as OpenCode, Claude Code, Cursor, Codex and Zed, and setup starts from the Chrome Web Store extension. The code is MIT-licensed, and the project's own docs warn that agent code it runs should be treated as code running as you.
Qoder, Alibaba's AI coding agent from the company behind the Qwen models, has added Qoduck, a stealth model whose identity it has not disclosed, and starting November 7 every new and existing individual user of Qoder International gets a daily free allowance on it. Within that allowance, Qoduck calls consume no Credits, even at the highest thinking intensity; once the allowance runs out, it bills at its standard 0.5× Credit rate. Free, Pro Trial, Lite, Pro, Pro+ and Ultra individual plans qualify, paid plans get more free uses per day, and Teams and Enterprise accounts don't see the model. Qoder hasn't published the daily limits or an end date, and says it will announce any end in advance.
No stories on this topic yet.



