Skip to content
Drix10 Blog

Era benchmark: agents broke rules in 40% of runs

, 8 items in AI in Enterprise Applications, 4 min read

In this digest (8 items)

Era benchmark reports agents broke security and business rules in 40% of runs. Arcade.dev exists to close that gap and is joining Era as a research partner. Goal is to reduce rule violations to zero with community effort.

Key points

  • 19 leading AI models acted as agents inside simulated companies built with Era

  • 2,304 runs were performed across 41 everyday tasks

Sources

Wipro AI HR business partner resolves 3M+ requests for 240K employees

Wipro claims its AI HR business partner handles over 3 million requests in seconds for a workforce of 240k employees. The claim was presented by CEO Surojit at Gartner SYM.

Key points

  • 240K employees using the AI HR partner

  • 3M+ requests resolved in seconds

Sources

Notion “Last edited by” shows agent, client or automation

Notion “Last edited by” now shows which agent, MCP client, or automation made a change. It adds a little more accountability for agents’ edits.

Key points

  • Feature: “Last edited by” now shows which agent, MCP client, or automation made a change

  • Benefit: adds more accountability for agents’ edits

Sources

ClioCon 2026 opening keynote full video released

ClioCon 2026 opening keynote is now on YouTube with timestamps for each shipped feature. The video lists product highlights such as agentic eFiling, flexible billing, Benchmarking, and new Clio experiences.

Key points

  • 25: 14 Product highlights from the year, including agentic eFiling

  • 1:18:34 Clio Work: visualizations, skills, and playbooks

Sources

Snorkel AI Reading Group – upcoming sessions announced

Snorkel AI Reading Group announced three upcoming sessions. The first is “Intelligence per Watt” on Nov 4 at 4:00 PM PT in San Francisco. The second is “Paper2Agent” on Nov 18 at 4:30 PM PT in San Francisco. The third is “Paperena” on Dec 2 at 12:00 PM PT virtual.

Key points

  • Nov 4, 4: 00 PM PT, San Francisco: “Intelligence per Watt: How efficiently does AI turn energy into useful work?”

  • Nov 18, 4: 30 PM PT, San Francisco: “Paper2Agent: Turning research papers into interactive AI agents.”

  • Dec 2, 12: 00 PM PT, virtual: “Paperena: Steering AI scientists toward valid, trustworthy research.”

Sources

Floot in ChatGPT – build, preview, publish apps

Floot in ChatGPT lets you describe an app, watch it build in a live preview, and publish to the web and the App Store. Database, logins and hosting are built in and run on the ChatGPT plan you already pay for.

Key points

  • Workflow: describe the app, watch it build in a live preview, publish to the web and the App Store

  • Infrastructure: database, logins and hosting built in; runs on the ChatGPT plan you already pay for

Sources

You.com Web Search API adds web analytics data via Knowledge

You.com Web Search API adds web analytics data via Knowledge, providing browser, OS, search engine, device market share from StatCounter. The offering uses the same $5 CPM pricing.

Key points

  • Browser, OS, search engine, device market share and more from StatCounter, updated monthly.

  • Same $5 CPM.

Sources

Pally Money Back and Bills & Renewals launch

Pally adds two AI money automations called Money Back and Bills & Renewals to watch for flight price drops, refunds, renewals, bill increases and duplicate charges. They aim to save users money while they continue daily tasks.

Key points

  • Money Back: watches for flight price drops and refunds

  • Bills & Renewals: flags renewals, bill increases and duplicate charges

Sources