Good morning, AI enthusiasts. Claude Code is having its "ChatGPT moment" — and it's not just developers paying attention.

From CEOs scrapping engineering hires to hobbyists building custom apps, Anthropic's tool is fueling a wave of "selfware" that puts software creation in the hands of anyone.

P.S. — The Rundown is hiring for several new roles across marketing, community, social, and more! Learn more and apply here.

In today’s AI rundown:

• Claude Code's virality rattles software stocks

• Ex-OpenAI policy lead’s new AI safety nonprofit

• Optimize prompting with this Markdown strategy

• Anthropic’s first Economic Index of 2026

• 4 new AI tools, community workflows, and more

LATEST DEVELOPMENTS

ANTHROPIC

📉 Claude Code's virality rattles software stocks

Image source: Anthropic

The Rundown: Anthropic's Claude Code is having its "ChatGPT moment" among both developers and hobbyists, with the viral excitement fueling a selloff in traditional software stocks and investors fearing a new era of AI-built 'selfware.'

The details:

• A Morgan Stanley SaaS index is down 15% YTD, with stocks from major software companies like Intuit, Adobe, and Salesforce dropping double digits.

• Vercel’s CTO used Claude to finish a year-long project in a week; another CEO scrapped plans to hire engineers after the tool made him ‘5x more productive’.

• Others went viral for building apps entirely with Claude Code, from custom MRI viewers to handling self-sustaining tomato growth.

• One financial analyst said investors have largely abandoned software stocks regardless of valuation amid existential AI fears.

Why it matters: Software stocks have held sky-high valuations thanks to predictable recurring revenue, logic that suddenly breaks down when AI can build custom tools on demand. Whether this is the actual beginning of a "selfware" era or just a hype-fueled blip, traditional SaaS faces major disruption from tools that turn everyone into a builder.

TOGETHER WITH YOU.COM

🧠 Tinkering with prompts can only get you so far

The Rundown: Most companies get stuck tinkering with prompts and wonder why their agents fail to deliver dependable results. This guide from You.com breaks down the evolution of agent management, revealing the five stages for building a successful AI agent and why most organizations haven’t gotten there yet.

In this guide, you’ll learn:

• Why prompts alone aren’t enough and how context and metadata unlock reliable agent automation

• Four essential ways to calculate ROI, plus when and how to use each metric

• Real-world challenges at each stage of agent management and how to avoid them

If you're ready to go beyond the prompt, this is the playbook for you.

AI SAFETY

🔍 Ex-OpenAI policy lead’s new AI safety nonprofit

Image source: Lovart / The Rundown

The Rundown: Former OpenAI policy lead Miles Brundage just launched AVERI, a new nonprofit pushing for independent third-party audits of frontier AI models instead of letting labs self-certify their own safety claims.

The details:

• Brundage left in October 2024 during OAI’s cut of its ‘AGI Readiness’ team, saying no frontier labs were ready on the safety front for advanced systems.

• AVERI will focus on establishing audit standards and policies rather than conducting evaluations itself, acting as a ‘think tank’ over an actual auditor.

• The institute received donations from employees at leading AI labs — people Brundage says "know where the bodies are buried."

• The org published a framework of "AI Assurance Levels" — ranging from today's self-testing (Level 1) to treaty-grade international verification (Level 4).

Why it matters: “Where the bodies are buried” is an ominous quote, but the AI industry has largely relied on self-reported safety testing thus far — with trust placed on whatever labs choose to publish. AVERI's insider-backed push for external scrutiny could mark the beginning of a shift toward real third-party accountability.

AI TRAINING

📈 Optimize prompting with this Markdown strategy

The Rundown: In this tutorial, you will learn how to extract any completed AI task into reusable markdown instructions — making valuable tasks portable and repeatable across any AI tool.

Step-by-step:

1. Create a Notebook in NotebookLM and complete your task (e.g., upload email to create variations), specifying word count, tone, rules, target, brand details

2. Prompt: "Take the task and extract the context into instructions for AI in markdown format. The goal is to drop these into any AI with different variables and receive polished output with one prompt.”

3. Create a new project in NotebookLM and paste the markdown instructions as text — now you've turned hours of prompting into a versatile project template

4. Start a new chat with custom input variables to achieve the same outputs as your original chat

Pro tip: After creating markdown instructions, save them to a database like Notion.

PRESENTED BY UNWRAP

⚡ Powerful insights for powerful brands

The Rundown: Unwrap aggregates all your customer feedback (surveys, reviews, support tickets, social, sales calls, etc.) into a single AI-powered view, helping product, support, and CX teams at Southwest Airlines, Stripe, lululemon, and DoorDash turn every customer signal into actionable insights.

With Unwrap, you get:

• All customer feedback auto-categorized into a single view

• Natural language queries to explore feedback instantly

• Real-time alerts, custom reporting, and clear sentiment tracking

Grab time directly with the team to talk through how Unwrap can automate your customer feedback analysis.

AI RESEARCH

📊 Anthropic’s first Economic Index of 2026

Image source: Anthropic

The Rundown: Anthropic published its fourth Economic Index, providing a deep dive into 2M Claude chats — revealing that most AI usage still looks more like collaboration than replacement, with humans in the driver's seat for the majority of tasks.

The details:

• AI handles around 1/4 of tasks in almost 50% of all jobs, up from 36% last year, but full role replacement is happening at under 10% of firms.

• Tasks requiring high-school skills are sped up 9x faster with AI, and 12x for college-level tasks, with Claude’s gains increasing for difficult use cases.

• Claude’s ability to handle longer tasks continues to rise, with Anthropic’s data showing a 50% success rate for tasks as long as 19 hours.

• Coding tasks still dominate Claude usage, though augmentation (learning, feedback, iteration) overtook pure automation-related tasks.

Why it matters: Productivity gains are happening fast, but the panic of AI completely taking over jobs doesn’t match the data (yet). The issue is for junior employees — if AI handles the grunt work usually used to train and provide experience, an entire generation of young talent might be stuck in a completely new economic reality.

QUICK HITS

🛠️ Trending AI Tools

• 🤖 GLM-4.7 Flash - Z AI’s fast, efficient variant of its SOTA open-source model

• 🗣️ TranslateGemma - Google’s new family of open-source translation models

• ⚙️ Claude Code - Anthropic’s deep-context AI coding assistant

• 🗣️ Scribe V2 - ElevenLabs’ SOTA transcription model

📰 Everything else in AI today

Z AI released GLM-4.7-Flash, a 30B, speed and efficiency-focused variant of its powerful open-source model that leads its size class across a series of benchmarks.

Business insights firm Gartner projected that worldwide AI spending will reach $2.52T in 2026, a 44% jump driven largely by massive infrastructure buildouts.

South Korea is currently holding an AI competition to elevate the country’s top homegrown AI models, with LG, SK, and startup Upstage remaining in contention.

Anthropic published research showing AI can slip into playing other characters during long chats, and introduced a fix that halved harmful responses.

xAI engineer Sulaiman Ghori announced that he left the company, coming on the heels of a revealing viral interview detailing the startup’s inner workings.

COMMUNITY

🤝 Community AI workflows

Every newsletter, we showcase how a reader is using AI to work smarter, save time, or make life easier.

Today’s workflow comes from reader Prateek M. in India:

"I recently went on a trip… and was looking for a new way to remember our moments. When I came across music/song creation AI tools, I realized a song would be a great way to cherish those memories. I went to ChatGPT and explained our trip. I uploaded both the inputs (lyrics & music description) onto Suno and… got the perfect trip song!"

How do you use AI? Tell us here.

🎓 Highlights: News, Guides & Events

• Read our last AI newsletter: Ads are officially coming to ChatGPT

• Read our last Tech newsletter: Wikipedia inks deals with Amazon, Meta

• Read our last Robotics newsletter: 1X now has a world model

• Today’s AI tool guide: Optimize prompting with this Markdown strategy

• RSVP to next workshop @ 4PM EST Friday: AI Foundations Bootcamp pt. 3

See you soon,

Rowan, Joey, Zach, Shubham, and Jennifer — the humans behind The Rundown