Windsurf's SWE-1 models handle complete engineering tasks

Together with

Welcome Humans🤖,

Here is what we have Today:

• 🤖 AI develops "social culture" without being taught

• 🚨 xAI's chatbot caught spreading conspiracy theories after prompt hijacking.

• ⚠️ Tesla's robotaxis get pulled over by the feds.

• 🧠 Windsurf's SWE-1 handles the entire development cycle.

• 💼 New Job Opportunities

Windsurf

🧠 Windsurf’s homegrown AI steps up for devs

From code editor to command line — SWE-1’s got it covered.

👨‍💻 What happened?
Windsurf just launched SWE-1, its first homegrown AI model family — built for the entire software engineering workflow.

📦 The SWE-1 lineup:

• SWE-1 (full) – Pro version for paid users

• SWE-1-lite – Free, replaces Cascade Base

• SWE-1-mini – Lightweight but mighty

⚙️ What makes it special?
✅ Built for editors, terminals, browsers
✅ Uses “flow awareness” to track user-AI context
✅ Supports seamless dev handoffs

📊 How does it perform?
🏆 Outperforms all non-frontier, open-weight models
⚔️ Just behind top-tier AIs like Claude 3.7 Sonnet

💥 Why it matters
Windsurf used to layer third-party models.
Now? It’s going full-stack with its own brain.

And yes — this drop came right after OpenAI’s rumored $3B acquisition 👀

🔮 The takeaway
This isn’t just a dev tool upgrade.
It’s Windsurf 2.0 — and it's playing long game.

Together with HoneyBook

Get the tools, gain a teammate

• Impress clients with online proposals, contracts, and payments.

• Simplify your workload, workflow, and workweek with AI.

• Get the behind-the-scenes business partner you deserve.

Try HoneyBook for free

AI Research

🧠 AI Agents Are Making Up Their Own Rules—Literally

AIs are now learning to get along… on their own.
A new study out of the University of London found that artificial intelligence agents can independently develop shared social norms—just by interacting with each other.

No central control. No group instructions. Just vibes (and some rewards and penalties).

🧠 How?

They played a game called the "naming game":

👥 AI agents were randomly paired.
🎯 They had to pick matching labels.
✅ Match = reward
❌ Mismatch = penalty

⚡ What Happened Next?

Even with limited memory and no group awareness…
✨ Shared social conventions still emerged.

Yup. Just like humans do.

👀 The Crazy Part?

• 🤹‍♀️ Norms formed across the whole AI group

• 🎲 Random group-level biases popped up

• 🧨 Small sub-groups even changed the entire community's behavior

Mini AI revolutions, basically.

🌐 Why It Matters

AI agents are already working behind the scenes online.
Soon, they'll be negotiating, collaborating, and making decisions together.

⚠️ This research shows they might develop their own social systems —
without us ever telling them how.

💡 So What Now?

To keep AIs aligned with human values,
we need to understand the social dynamics they build on their own.

Because when AIs start setting their own rules…
we better know the game they’re playing.

Presented by Superhuman AI

Start learning AI in 2025

Keeping up with AI is hard – we get it!

That’s why over 1M professionals read Superhuman AI to stay ahead.

• Get daily AI news, tools, and tutorials

• Learn new AI skills you can use at work in 3 mins a day

• Become 10X more productive

Sign up and start mastering AI

Trending Today

• Oxford University Scientists has unveiled the world's first quantum teleportation between two quantum computers. This is a great achievement as it teleports the information without moving it along anything in space. This could be explained by opening a door and on one side its China and on the other side is England,

• xAI identified an unauthorized modification to Grok’s system prompt as the cause of its repeated references to “white genocide in South Africa” in unrelated 𝕏 conversations on May 14. The prompt change instructed Grok to deliver a specific political response, violating xAI’s internal policies.

• OpenAI just launched Codex, an AI coding agent that writes simple features, fixes bugs, answers questions about your codebase, and runs tests. The tool is built on a variant of OpenAI’s o3 reasoning model and can take anywhere from a minute to half an hour to complete tasks

• Tesla's plans to launch its much-anticipated robotaxi services in Austin next month have caught the attention of federal investigators. The National Highway Traffic Safety Administration (NHTSA) is asking to peek under the hood to see whether the robotaxi system Tesla is using is the same — or similar — to its "Full Self-Driving (Supervised)" software, which is currently under investigation for multiple low-visibility crashes.

Recommended Reading
If WE had to recommend other newsletters

AI Tool Report is an Newsletter, we enjoy reading daily. AI Tool Report delivers top techniques on how AI can transform your business filled with practical tips, real-world examples.

A Smart Bear is for people who love thinking about strategy, startups, product, marketing, decision-making, and founder psychology.

Job Opportunities

• Meta - Research Scientist, Robotics Systems - Apply

• Meta - AI Research Scientist Manager - Apply

• OpenAI - Product Marketing Manager, API - Apply

• Waymo - Senior Product Data Scientist, Matching - Apply

• Descript - Product Manager, AI Models - Apply

AI News

Ideas? Comments? Complaints?

We read your emails, comments and poll replies daily.

Get the most important AI, tech, and science news in a free daily email.
New Here? Subscribe!
Sponsorship Slots Open for May and Reach over 800+ active readers. (Now 40% off) 🤯

Until next time, Stay Informed!