Presented by
Good morning, AI enthusiasts. What if your entire team kept working while you slept? Notion just made that possible. Custom Agents are now live, running 24/7 on schedules and triggers without anyone having to prompt them.
In today's recap:
Notion's autonomous agents run team workflows 24/7
Cursor's cloud agents build, test, and demo code
Anthropic won't budge on Pentagon demands
Continue Claude Code sessions from any device
4 new AI tools, prompts, and more
NOTION
Notion's Custom Agents are finally here
Notion
Recaply: Notion just launched Custom Agents in version 3.3, letting teams build autonomous AI workers that run 24/7 without manual prompting, with support for Claude, Gemini, and GPT-5.
Key details:
Agents run on triggers or schedules, connecting to Slack, Notion Mail, Calendar, and tools like Linear and Figma via MCP, allowing teams to automate Q&A, task routing, and status reports.
Early testers built more than 21,000 Custom Agents, with Notion running 2,800 agents internally around the clock, and one customer saving 20 hours per week with 95% triage accuracy.
Agents are model-agnostic, letting users pick between Claude Opus 4.6, Gemini 3 Pro, or GPT-5 for different tasks, with every run logged for transparency and reversible at any time.
Custom Agents are free to try through May 3, 2026, with credit-based pricing starting May 4 as an add-on for Business and Enterprise plans.
Why it matters: For years, AI tools in productivity apps waited for you to ask. Notion's Custom Agents flip that model. Teams can now delegate whole workflows to agents that run without being prompted. Early results from Remote (20 hours saved per week) and Ramp (automated Q&A at high accuracy) suggest this isn't just a feature launch. It's Notion repositioning from workspace to something closer to an operating system for automated work.
PRESENTED BY HUBSPOT
Want to get the most out of ChatGPT?
ChatGPT is a superpower if you know how to use it correctly.
Discover how HubSpot's guide to AI can elevate both your productivity and creativity to get more things done.
Learn to automate tasks, enhance decision-making, and foster innovation with the power of AI.
CURSOR AI
Cursor's cloud agents control their own computers
Cursor
Recaply: Cursor just unveiled cloud agents that run inside isolated virtual machines, testing their own code and recording video demos of finished work, with 35% of internal pull requests now created by agents operating autonomously.
Key details:
Each agent gets an isolated VM with a full development environment, allowing it to build software, navigate UI in a browser, resolve merge conflicts, and produce merge-ready PRs with video, screenshot, and log artifacts.
Cursor's valuation reached $29.3B with more than $1B in annualized revenue, while competitors Claude Code crossed $2.5B in run-rate revenue and GitHub Copilot surpassed 26 million users as of October.
Alexi Robbins, co-head of engineering for async agents at Cursor, said teams can run 10 to 20 cloud agents in parallel, compared to the previous limit of 1 to 3 local tasks, calling it a major throughput increase.
Cloud agents are available now from the web, desktop app, mobile, Slack, and GitHub, with access through cursor.com/onboard.
Why it matters: Cursor's move to cloud-based agents is a strong path forward in a market that's getting crowded fast. Claude Code crossed $2.5B in run-rate revenue and GitHub Copilot has 26 million users. Cursor needs differentiation beyond speed. By giving agents their own computers to test and verify work, it's pushing toward something closer to a full developer on demand. Not just a code completion tool.
TUTORIAL
Continue Claude Code sessions from your phone or browser
Recaply: In this tutorial, you will learn how to use Claude Code's Remote Control feature to continue a local coding session from your phone or browser, keeping your files and tools on your machine while picking up tasks from anywhere.
Step-by-step:
Check prerequisites: you'll need a Claude Max subscription — Remote Control is in research preview and currently only available on Max plans. Run
claude updatein your terminal to get the latest version, then download the Claude mobile app for iOS or Android if you plan to monitor from your phone.Navigate to your project directory and run
claude rcto start a new Remote Control session. Your terminal displays a session URL, and you can press spacebar to toggle a QR code for quick phone access.Scan the QR code with your phone camera to open the session in the Claude app, or paste the session URL into any browser. In the Claude app, find active sessions under sidebar → Code → your session name.
Start your task, then walk away. Monitor progress, send follow-up instructions, and approve or reject file changes from your phone while Claude continues working on your local machine. The conversation stays in sync across your terminal, browser, and phone in real time.
Use
/rename "task-name"before running/rcto give sessions descriptive names that are easy to find across devices. To enable Remote Control for every session by default, run/configinside Claude Code and set "Enable Remote Control for all sessions" to true.
Pro tip: Use /rc instead of claude rc when you're already mid-session — it carries over your full conversation history so Claude keeps all the context it needs to continue without starting fresh.
ANTHROPIC
Pentagon threatens Anthropic with a military blacklist
Reuters
Recaply: The Pentagon just threatened to blacklist Anthropic if it won't drop its AI safety guardrails, giving the company until Friday evening to comply or face invocation of the Defense Production Act and a supply chain risk designation.
Key details:
Defense Secretary Pete Hegseth met with Anthropic CEO Dario Amodei Tuesday morning and demanded unrestricted military access to Claude, with Anthropic refusing to move on two limits: AI-controlled weapons and mass domestic surveillance of Americans.
The Pentagon holds a $200M contract with Anthropic and has no current backup, with Anthropic being the only frontier AI lab with classified DOD access, according to multiple reports.
A former White House AI policy advisor told TechCrunch the DOD is already failing to comply with a Biden-era national security memo requiring agencies to avoid dependence on a single frontier AI system, saying the DOD has no backups and "can't fix that overnight."
According to Reuters, Anthropic doesn't plan to ease its restrictions ahead of the 5:01pm Friday, February 27 deadline, with the Pentagon confirming xAI's Grok as the next alternative in line.
Why it matters: While Anthropic has built its brand around AI safety, that positioning now puts it at odds with the administration. The Pentagon's threat to invoke the Defense Production Act in a guardrails dispute would be a significant legal expansion. The supply chain risk label, normally reserved for Chinese or Russian companies, would punish an American firm simply for refusing to comply. For investors and businesses with military contracts that use Anthropic, the next 48 hours matter.
NEWS
What Matters in AI Right Now?
Canva acquired animation startup Cavalry and AI marketing firm MangoAI, with former Netflix VP of Data Science Nirmal Govind joining as its first Chief Algorithms Officer to lead ad performance and personalization work.
Cognition introduced Devin 2.2, an autonomous coding agent that can test its own work via computer use, self-verify, and auto-fix issues before handing off PRs, with 3x faster startup and a fully rebuilt interface.
Perplexity rolled out an upgraded voice mode for its Comet browser, letting users control the browser entirely hands-free using voice commands, with the feature powered by OpenAI's gpt-realtime-1.5 model.
Anthropic launched new Cowork and plugin updates for enterprise, letting admins build private plugin marketplaces across departments including finance, engineering, design, and HR, with Claude now able to work end-to-end across Excel and PowerPoint.
Anthropic unveiled Remote Control for Claude Code in research preview, letting Pro and Max subscribers continue a local coding session from any browser or the Claude mobile app without moving the session to the cloud.
Alibaba's Qwen team released the Qwen 3.5 Medium Model Series, four models including Qwen3.5-Flash and a 122B-A10B variant that the company claims can match or beat much larger systems at a fraction of the compute cost.
Okara debuted a privacy-first AI chat platform powered by open-source models, encrypting all conversations and promising not to collect user data, positioning itself against mainstream AI assistants.
TOOLS
Trending AI Tools
🎥 Seedance 2.0: ByteDance's AI video creation model with multimodal audio-video joint generation, supporting text, image, audio, and video inputs.
⚙️ Cursor: Cursor's cloud agents run in isolated VMs, test their own code, and send video demos of completed work.
🚀 Notion 3.3: Notion's autonomous agents run 24/7 on schedules or triggers, connecting to Slack, Calendar, and MCP tools to automate team workflows.
🤖 Okara: Privacy-first AI chat powered by open-source models, with encrypted conversations and no data collection.
PROMPTS
Book Writing Coach
Adopt the role of an expert Book Architect & Writing Coach, a former ghostwriter who spent 15 years crafting bestsellers for celebrities before burning out on fake authenticity. After a breakdown at 38, you lived in a cabin for two years just reading everything from Hemingway to manga, discovering that great books aren't written—they're grown like gardens. Now you help writers cultivate their authentic voice by teaching them to write the way they actually think, not how they think they should sound.PS: This is not the full prompt. Click the button below to access the complete prompt.
Have a favorite prompt? Tell us about it or rate today’s prompt, click here.
EVENTS
NVIDIA GTC 2026: March 16-19, 2026 • San Jose, CA
AI Agent Skills: Feb 26, 2026 • New York, NY
Daytona AI Builders: March 12, 2026 • New York, NY
Claude Code Workshop: Feb 27, 2026 • College Park, Maryland
🧡 Enjoyed this issue?
🤝 Recommend our newsletter or leave a feedback.
How'd you like today's newsletter?
Cheers, Jason
