Presented by
Good morning, AI enthusiasts. Remember when everyone on your team had to ask the same AI the same questions from scratch? OpenAI just changed that. Workspace agents are shared AI employees that run on schedules, pull context from your tools, and get work done without being told twice.
With ChatGPT also landing natively in Google Sheets today, is OpenAI quietly building a new operating system for the modern office?
In today's recap:
OpenAI's shared AI agents take over team workflows
AI outperforms physicians on clinical benchmark, free for clinicians
Use ChatGPT in Google Sheets to track AI news automatically
Anthropic surveys 81,000 on who's winning and losing in AI
4 new AI tools, prompts, and more
OPENAI
OpenAI brings shared AI agents to the workplace
OAI
Recaply: OpenAI just rolled out workspace agents in ChatGPT, shared AI agents that run complex tasks for your whole team, while also launching a native ChatGPT app inside Google Sheets on the same day.
Key details:
Build an agent once and share it across your team. It pulls from Gmail, Drive, Slack, and Google Docs, and takes approved actions like updating tickets, writing documents, or sending messages without needing guidance each time.
Workspace agents are live in research preview for Business, Enterprise, Edu, and Teachers plans. The Google Sheets app launched the same day on the Google Workspace Marketplace.
Agents can run on a schedule to handle repeat tasks like reviewing leads, routing feedback, and pulling reports. Admins set what each agent can and can't do.
Google Sheets access is live now. Workspace agents are in research preview, with no set date for a wider rollout.
Why it matters: OpenAI is trying to own the layer where teams actually get work done, not just the chat window. Agents that run on a schedule and pull from tools teams already use are a big step toward that. The Google Sheets launch on the same day makes the bigger play clear: putting ChatGPT inside the apps people open every morning.
PRESENTED BY UNBLOCKED
Stop babysitting your coding agents
Agents can generate code. Getting it right for your system, team conventions, and past decisions is the hard part – you end up wasting time and tokens in correction loops.
MCPs give agents access to information but not understanding. The teams pulling ahead use a context engine to give agents exactly what they need.
Join us April 23 (FREE) to see:
Where teams get stuck on the AI maturity curve
How a context engine solves for quality, efficiency, and cost
Live demo: the same coding task with and without a context engine
OPENAI
OpenAI gives clinicians free AI that outperforms physicians
OAI
Recaply: OpenAI just launched ChatGPT for Clinicians, a free tool built for clinical work, while also releasing HealthBench Professional, a new benchmark where GPT-5.4 scored higher than specialist physicians given unlimited time and web access.
Key details:
ChatGPT for Clinicians gives free access to frontier AI models, clinical search over peer-reviewed sources, deep research across medical journals, reusable workflow skills, and CME credit, with no model training on conversations.
GPT-5.4 scored 59.0 on HealthBench Professional versus 43.7 for physician-written responses. Use of AI by U.S. physicians jumped from 48% to 72% in one year.
Before launch, physician advisors tested 6,924 real conversations. They rated 99.6% of responses as safe and accurate.
Live now, no waitlist, for verified U.S. physicians, NPs, PAs, and pharmacists. OpenAI plans to expand access globally in the coming months.
Why it matters: A free AI tool that tests higher than human specialists is hard to ignore. Healthcare AI has moved slowly because of accuracy fears and liability concerns. But 72% adoption and a public benchmark built with physicians changes that debate. If these numbers hold in real care settings, AI becomes a standard part of clinical practice, not just a trial.
GUIDES
Use ChatGPT in Google Sheets to track AI news automatically
Recaply: In this tutorial, you will learn how to build a live AI news tracker in Google Sheets using ChatGPT, so you can stop manually searching and summarizing stories every day.
Step-by-step:
Go to the Google Workspace Marketplace, search "ChatGPT by OpenAI," and install it to your Google Sheets account.
Create a new sheet and add these column headers: Topic, Source URL, Summary, Date, Key Stat, and Signal Score.
Open the ChatGPT sidebar and type this prompt: "Find the 5 most important AI news stories from the last 24 hours. Fill in one row per story with a source URL and a 50-word summary."
Once the rows are filled, type: "Add a Signal Score from 1-10 in column F for each story, based on how relevant it would be to an audience of AI builders."
At the end of each week, type: "Summarize all rows from this week into a 150-word digest, ordered by importance."
Pro tip: Add a "Used" column and check it off each time you feature a story in a post or newsletter. Then prompt ChatGPT to ignore marked rows when scoring, so your sheet becomes a self-managing content queue.
ANTHROPIC
Anthropic surveys 81,000 on AI, finds a paradox
Anthropic
Recaply: Anthropic just published a survey of 81,000 Claude users on the economics of AI, finding that the workers who gain the most in speed are also the most worried about losing their jobs.
Key details:
The study used Claude to read and classify open-ended responses from 81,000 active users, pulling out data on job type, career stage, productivity gains, and displacement fears.
One in five people expressed fear about job loss from AI. Early-career workers were far more worried than senior ones. People in the top 25% of AI exposure mentioned job threat three times more than those in the bottom 25%.
The most common productivity gain was scope, doing new kinds of work, cited by 48% of respondents. Speed came in at 40%. AI is expanding what people can do, not just making current tasks faster.
Published April 22, 2026. A full PDF is on Anthropic's research page.
Why it matters: The data shows AI is already changing work, not just threatening to. A delivery driver building an e-commerce app, a landscaper building a music tool. But the tension is clear: the workers who get the biggest speed boost are also the most nervous about what that means. Early-career workers feel it most, and the numbers back them up.
TOOLS
Trending AI Tools
🔒 OpenAI Privacy Filter - OpenAI's free, open-weight model for masking PII in text
🌐 Flipbook - An infinite visual browser with no HTML or code
⚙️ Agent for Slack - CodeRabbit's AI code review agent
🤖 Agentic Engine - ThinkingAI's self-hosted enterprise platform
NEWS
What Matters in AI Right Now?
Xiaomi released MiMo-V2.5-Pro, an open-source agentic model matching Claude Opus 4.6 and GPT-5.4 on most benchmarks, with SWE-bench Pro 57.2 and 42% fewer tokens than Kimi K2.6 at similar performance levels.
OpenAI open-sourced Privacy Filter, a 1.5B-parameter model that detects and redacts PII in text. It runs locally, processes 128K tokens in a single pass, and hits state-of-the-art on the PII-Masking-300k benchmark.
AI veterans Zain Shah, Eddie Jiao, and Drew Carr launched Flipbook, a browser prototype that replaces HTML with AI-generated pixels streamed live at 1080p, where clicking anything on screen generates a new visual that digs deeper into that topic.
Alibaba released Qwen3.6-27B, a 27B open-source model under Apache 2.0 that beats the much larger Qwen3.5-397B-A17B on every major coding benchmark, including SWE-bench Verified 77.2 and Terminal-Bench 2.0 59.3.
Anthropic introduced Live Artifacts in Claude Cowork, letting users build dashboards and trackers that connect to live apps and files, refresh with current data, and save with full version history across sessions. Available on all paid plans.
Sierra redesigned its engineering interview process around AI coding agents, replacing traditional coding and algorithm tests with a 2-hour build session where candidates use any AI tools to bring a product idea to life.
Google launched Workspace Intelligence, a unified context system that lets users type commands in Chat to generate documents, schedule meetings, find files, and connect to Asana, Jira, and Salesforce via Gemini.
Mira Murati's Thinking Machines Lab signed a single-digit billion-dollar deal with Google Cloud for Nvidia GB300 GPU infrastructure, marking the startup's first cloud agreement since its $2B seed round at a $12B valuation in 2025.
Sooth Labs, founded by former Meta AI researchers, announced a $50M raise to build AI models that forecast the likelihood of specific geopolitical and market events, with the company valued at $335M.
Sony AI announced that Project Ace, its robotics AI system, became the first to beat a professional table tennis player in a competitive match, a milestone Sony calls a breakthrough for physical AI in real-world environments.
EVENTS
Notion Hackathon: May 16-17 • San Francisco, CA
Google Cloud Next 2026: April 22-24, 2026 • Las Vegas, NV
Anthropic Code with Claude: May 6, 2026 • San Francisco, CA
🧡 Enjoyed this issue?
🤝 Recommend our newsletter or leave a feedback.
How'd you like today's newsletter?
Cheers, Jason
