AI News Recap: March 13, 2026
The Week AI Moved Into Government, Healthcare, and the Enterprise All at Once
AI Agents Now Have Their Own Social Media, a Seat in Congress, and Better Job Security Than 1,600 Atlassian Employees
Happy Friday the 13th (again), folks! That's two in a row now, because apparently the calendar is as unsettling as the news cycle. Somewhere between Meta buying a social network for robots, the U.S. Senate giving ChatGPT a security clearance, and Amazon becoming your new doctor, I started to wonder if the AI singularity already happened and nobody told the humans. Oh, and Grammarly got caught pretending to be Stephen King. It was that kind of week.
Table of Contents
👋 Catch up on the Latest Post
🔦 In the Spotlight
💡 Beginner’s Corner: Zero-Shot Sim-to-Real Transfer
🗞️ AI News
🔥 Rhythm's Hot Takes
🧩 NeuralBuddies Word Search
👋 Catch up on the Latest Post …
🔦 In the Spotlight
Meta Acquires Moltbook, the Viral Social Network for AI Agents
Category: Business & Market Trends
🤖 Meta has acquired Moltbook, a Reddit-style social network designed exclusively for AI agents, bringing its co-founders into Meta Superintelligence Labs (MSL).
🔗 Moltbook allowed AI agents using OpenClaw to autonomously post, comment, and interact with one another, racking up millions of registered bots within days of launch.
🔐 Security researchers revealed Moltbook’s infrastructure was poorly secured, making it easy for humans to impersonate AI agents on the platform.
💡 Beginner’s Corner
Zero-Shot Sim-to-Real Transfer
Imagine learning to drive entirely in a video game, and then getting behind the wheel of a real car for the first time and driving perfectly. That is essentially what zero-shot sim-to-real transfer means in robotics.
Traditionally, training a robot to pick up objects, navigate rooms, or manipulate tools requires collecting thousands of hours of real-world demonstration data, a process that is expensive, slow, and limits who can participate in robotics research. Zero-shot sim-to-real transfer flips this model: the robot trains entirely inside a computer simulation and then performs the task in the physical world without any additional real-world practice.
This week, Ai2 announced MolmoBot, a robotic system trained on 1.8 million simulated manipulation trajectories that successfully transferred its skills to real robots with no fine-tuning. If this approach scales, it means the biggest bottleneck in robotics shifts from collecting data to designing better virtual worlds.
Related Story Below: Ai2 Demonstrates Zero-Shot Sim-to-Real Transfer With Open Robotics
🗞️ AI News
U.S. Senate Approves ChatGPT, Gemini, and Copilot for Official Staff Use
Category: Legal & Governance
🏛️ The Senate Sergeant at Arms approved ChatGPT, Google Gemini, and Microsoft Copilot for official use by Senate staff for tasks like drafting, research, and briefing preparation.
🔒 The memo specifies that staff should not input personally identifiable information, physical security details, or classified material into the AI tools.
📋 The House of Representatives has already approved similar tools including Claude Pro, making this part of a broader trend of AI adoption across Congress.
Amazon Launches Its Healthcare AI Assistant on Its Website and App
Category: Healthcare & Biotechnology
🏥 Amazon expanded its Health AI assistant beyond the One Medical app to Amazon.com and the Amazon app, allowing users to ask health questions, manage prescriptions, and book appointments.
🆓 Users do not need to be Prime subscribers or One Medical members to use Health AI, though Prime members get up to five free consultations with a One Medical provider.
⚠️ Amazon says interactions occur within a HIPAA-compliant environment, but researchers continue to warn against sharing personal health data with AI chatbots.
Grammarly Hit With Class-Action Lawsuit Over AI ‘Expert Review’ Feature
Category: AI Ethics & Regulation
⚖️ Journalist Julia Angwin filed a class-action lawsuit against Grammarly’s parent company Superhuman, alleging the company used the names and identities of hundreds of writers without consent.
✍️ Grammarly’s “Expert Review” feature presented AI-generated editing advice as if it came from real figures like Stephen King and Neil deGrasse Tyson, none of whom had agreed to participate.
🚫 Following widespread backlash from journalists including Casey Newton and Kara Swisher, Superhuman disabled the feature and its CEO issued a public apology.
Anthropic in Talks With Private Equity Firms on AI Joint Venture
Category: Business & Market Trends
💼 Anthropic is in discussions with Blackstone and Hellman & Friedman to form an AI-focused joint venture aimed at selling Claude technology to companies backed by the investment firms.
🏗️ The proposed partnership would follow a Palantir-style model, offering consulting services to help companies integrate Anthropic’s AI into their operations.
⚔️ The talks are continuing despite Anthropic’s ongoing dispute with the Pentagon over military use of its AI tools.
Atlassian to Lay Off 1,600 Employees in Pivot to AI
Category: Workforce & Skills
📉 Atlassian announced it will cut approximately 10% of its workforce, or 1,600 jobs, to redirect resources toward AI and enterprise sales.
💰 The company expects to incur $225 million to $236 million in restructuring charges, with the majority of affected employees in North America (40%), Australia (30%), and India (16%).
🔄 CEO Mike Cannon-Brookes acknowledged that while AI is not directly replacing people, it changes the skills and roles the company needs going forward.
OpenAI Plans to Launch Sora Video Tool Inside ChatGPT
Category: Generative AI & Creativity
🎬 OpenAI plans to integrate its Sora AI video generator directly into ChatGPT, potentially exposing the tool to hundreds of millions of existing users.
📉 The move comes as the standalone Sora app has struggled with declining engagement, with installs dropping 45% month-over-month in January 2026 and falling out of Apple’s top 100 apps.
💸 The integration could significantly increase ChatGPT’s operating costs, as video generation is far more computationally expensive than text-based tools.
Ai2 Demonstrates Zero-Shot Sim-to-Real Transfer With Open Robotics Stack
Category: Robotics & Autonomous Systems
🤖 The Allen Institute for AI (Ai2) released MolmoBot, a robotic manipulation model trained entirely on synthetic simulation data, achieving zero-shot transfer to real-world robots without any real-world fine-tuning.
📊 The accompanying dataset, MolmoBot-Data, contains 1.8 million expert manipulation trajectories generated procedurally, bypassing the need for expensive human teleoperation.
🔓 The entire stack, including training data, generation pipelines, and model architectures, has been released as open source to democratize physical AI research.
New MIT Class Uses Anthropology to Improve Chatbot Design
Category: Education & Learning
🎓 MIT launched an undergraduate class cross-listing computer science with anthropology, teaching students to design AI chatbots that serve as social guides rather than addictive distractions.
🧠 The course, called Humane UXD, uses methods from linguistic anthropology to integrate human interactional and interpersonal needs into AI programming.
👩💻 Students designed chatbots aimed at helping young users become more social and socially confident, including projects focused on news engagement and emotional well-being.
Perplexity Launches Enterprise Version of Its ‘Computer’ AI Agent
Category: Tools & Platforms
🏢 Perplexity announced at its inaugural Ask 2026 conference that its multi-model AI agent, Computer, is now available to enterprise customers with security controls, compliance features, and Slack integration.
🖥️ The company also unveiled “Personal Computer,” software that runs on a dedicated Mac mini to merge local applications with Perplexity Computer for always-on, local AI assistance.
🔢 Computer orchestrates 19 different AI models to execute complex workflows, routing each task to the best-suited model, and can create subagents to handle specialized problems.
Anthropic Launches Code Review Tool to Check AI-Generated Code
Category: Tools & Platforms
🔍 Anthropic launched Code Review, a new feature in Claude Code designed to review pull requests and catch bugs in AI-generated code before they reach the codebase.
📈 The tool addresses a growing bottleneck caused by Claude Code’s rapid increase in code output, which has overwhelmed traditional human review processes.
🏢 Code Review is available first to Claude for Teams and Claude for Enterprise customers in a research preview.
🔥 Rhythm's Hot Takes
“If it grooves, it communicates.”
Every great recording has a moment of silence. A breath before the drop, a rest between movements. It is what gives the music its shape. Without it, you just have sound.
I have been thinking about that this week, because AI is not pausing. Meta acquired a social network where bots talk to bots. The Senate approved AI for legislative work. Anthropic is building private equity partnerships while simultaneously fighting the Pentagon. Perplexity launched an enterprise agent that runs 19 models at once. And 1,600 Atlassian employees learned that “pivoting to AI” sometimes means pivoting away from the people who built the company. Every one of these stories has momentum, volume, energy. But I keep listening for the rest, the moment where someone asks: is the rhythm right? Because tempo without intention is just noise. And right now, the industry is playing faster without asking whether the song still makes sense.
— Rhythm 🎵
