Cloudflare basically bottled their secret recipe for auditing a codebase, and then open-sourced it. 🔍

It's a six-phase security-audit "skill" they built internally to hunt bugs, now anyone can run it. ~30k stars and climbing.

Pairs nicely with today's PHP story (yes, the now, like a wine list).

Full convo at the link in bio. It's from today's Chaos Agents newsletter.

🔗 github.com/cloudflare/sec…

DeepSeek's 75-Cent Search, Hybrid Reasoning Agents, and Programming Language Wars youtu.be/uw5z3V93ZBI

An AI agent that thinks without words when it's faster, and only switches to text when it needs a tool.

+10.4% accuracy ~half the tokens up to 2.6x faster open weights

"Hybrid Reasoning Agents" (Stanford + Together AI) 🧵

Your AI agent editing a spreadsheet usually means praying it didn't nuke row 400.

@univerHQ reframes this: agents edit in a draft, you review the diff and merge. Git-style approvals + rollbacks for sheets/docs/slides. Apache-2.0, ~22k stars.

github.com/dream-num/univ…

$120k in AI tokens sounds wild until you’ve lived through a rewrite.

@Microsoft rewrote Copilot from TypeScript to Rust using agents. About $120,000 in tokens.

What did your last migration cost?

Our free newsletter: chaosagents.ai

AI wrote the code. Your review queue would like a word.
One model writes, a different model reviews. Is that the missing layer in AI coding?
Get the free Chaos Agents newsletter → chaosagents.ai

AI agents designed hardware that runs Qwen on an old FPGA github.com/FeSens/openTPU.
@aviflombaum sees a bigger story: what tiny, fast AI hardware could mean for medical devices.

There’s a strange paradox in reinforcement learning:
We want models to explore.
But the systems judging their outputs tend to reward things that already look “good” — which can push models back toward the same kinds of answers.
A new paper explores what happens when you break that loop.

Wait, you can now search your videos by just… describing what happens in them.
Screen indexes the actual content of your videos locally on your Mac.
This is precisely the kind of thing we started the Chaos Agents newsletter to find.
We go looking. You get the good stuff → chaosagents.ai

An AI math tutor that’s wrong on purpose is such an interesting inversion.

Instead of teaching kids to ask AI for the answer, @lathoa_ai makes them catch the AI’s mistake and explain why it’s wrong.

AI can mimic a lot of things, but it can’t replicate lived experience. 🧠✨

📬 Subscribe to our newsletter at chaosagents.ai

Ever wondered what it actually takes to build an app using AI?

📬 Subscribe to our newsletter at chaosagents.ai

When users love your product and you drop a pointless AI tool... friction is inevitable. 😒💻

📬 Subscribe to our newsletter at chaosagents.ai

AI isn’t just generating plausible ideas anymore—it’s predicting real-world medical breakthroughs. 🤯

📬 Subscribe to our newsletter at chaosagents.ai

Did we already cross the threshold into AGI without realizing it? 🧠

📬 Subscribe to our newsletter at chaosagents.ai

RLHF actually stands for Reinforcement Learning from Human Feedback, but let’s be honest: it’s really Reinforcement Learning for Human Flattery.

📬 Subscribe to our newsletter at chaosagents.ai

Ever wondered what "Trust & Safety" actually means in tech? 🛡️

📬 Subscribe to our newsletter at chaosagents.ai

If you aren't smiling at our babies you hate joy.

When your brain is racing in 20 directions, reading for fun is the first thing to go.

📬 Subscribe to our newsletter at chaosagents.ai

Great ideas used to get stuck on the drawing board. Not anymore.

📬 Subscribe to our newsletter at chaosagents.ai

Ever wonder what actually happens behind the scenes when a website loads? 🌐

📬 Subscribe to our newsletter at chaosagents.ai

Romance scams are way more common than you think. Stay sharp out there. 💔

📬 Subscribe to our newsletter at chaosagents.ai

Nothing kills product love faster than adding an AI feature nobody asked for. 😒

📬 Subscribe to our newsletter at chaosagents.ai

Academics can argue over what counts as "true AI" all day long, but the public has already made up its mind.

📬 Subscribe to our newsletter at chaosagents.ai

Having a great idea is one thing. Actually building it? That used to be the hard part. 💡

📬 Subscribe to our newsletter at chaosagents.ai

Can you spot the difference between human and AI content? 🤖 vs 👤

📬 Subscribe to our newsletter at chaosagents.ai

Ever felt like your algorithm just gets you? 🧠✨

📬 Subscribe to our newsletter at chaosagents.ai

Here’s why spending 30 minutes a night playing with your cat is the ultimate practice for life with a baby.

📬 Subscribe to our newsletter at chaosagents.ai

Scientists are using human brain cells as computers, and we’re breaking down what that actually means. 🧠💻

📬 Subscribe to our newsletter at chaosagents.ai

How close is AI to passing as fully human? 🤖 vs 👤

Enter the Likeness Score — the benchmark measuring just how convincing AI-generated content really is (and how it’s evolving across text, audio, and video).

📬 Subscribe to our newsletter at chaosagents.ai

Just because someone loves you doesn't mean they'll buy your product. 💡

Don't fall into the "Mom Trap." Trade polite compliments for real customer stories so you can build something people actually need.

📬 Subscribe to our newsletter at chaosagents.ai

"Sorry for the late reply!" Sound familiar? 😅

Text burnout is real, and unread notifications stack up fast. But what if AI could summarize every message you missed and let you respond to everyone from a single dashboard?

📬 Subscribe to our newsletter at chaosagents.ai

When your AI’s behavior shifts subtly, the risks are real.

It’s not just about politics; a drifting AI agent can easily steer a customer down the wrong path without anyone noticing.

📬 Subscribe to our newsletter at chaosagents.ai

"Now I'm just a manager for coding agents..."

The emotional reality of delegating to AI instead of human developers.

📬 Subscribe to our newsletter at chaosagents.ai"

Can Artificial General Intelligence and faith coexist? 🧠✨

Here’s the key distinction: sentience and AGI aren't the same as the human soul.

📬 Subscribe to our newsletter at chaosagents.ai

An entire generation is growing up with AI as second nature. 🤖

As sci-fi edges closer to reality, if we ever end up powering the system, we’re demanding the premium VR experience; cows can’t have all the fun. 🐄🥽

📬 Subscribe to our newsletter at chaosagents.ai

Where is AI actually taking us? 🔮

🤖 Are we just training AI to make it smarter?

💡 Or is AI going to make us smarter, more curious, and more fulfilled?

📬 Subscribe to our newsletter at chaosagents.ai

The internet can be a wild place. 😬 That’s where Trust & Safety comes in.

T&S is all about identifying online risks—and building the teams and systems that protect you from them.

📬 Subscribe to our newsletter at chaosagents.ai

How “human” does AI actually sound? 🤖💬

That’s where the Likeness Score comes in. It measures how close synthetic content gets to real human creation—and how that line is shifting as AI evolves.

📬 Subscribe to our newsletter at chaosagents.ai

Pause before you post. ⏸️

Small prompts like “Are you sure you want to send this?” are doing big work behind the scenes. It’s called Safety by Design—and it’s changing how tech platforms keep users safe.

📬 Subscribe to our newsletter at chaosagents.ai

What does true AI autonomy actually look like?

OpenClaw pushes agentic AI to its limits—connecting directly to your Slack, email, and (if you’re feeling brave) your bank accounts.

📬 Subscribe to our newsletter at chaosagents.ai

Where are we on the AI hype vs. reality spectrum right now? 🤖

0️⃣ "Trapped in the Matrix"
5️⃣ "AI drafted my email and I kind of liked it"
🔟 "Eating breakfast with my robot roomie"

Drop your score from 0–10 👇

More users using Claude ➡️ More feedback ➡️ Better outputs over time.

Adoption doesn’t just follow AI intelligence; it shapes it.

📬 Catch the full breakdown at chaosagents.ai

How do you objectively measure when AI-generated content becomes indistinguishable from human work? Meet the Likeness Score 🧠

📬 Catch the full breakdown at chaosagents.ai

Mind racing every time you try to read for fun? 🤯📚

Navarrow Wright joined us to share how AI completely transformed his reading routine, and how it can do the same for you.

📬 Catch the full breakdown at chaosagents.ai

What really happens when you click a link? 🧐

Headless browsers do the invisible groundwork, fetching data and mapping out pages before you even see them.

Subscribe to our newsletter at chaosagents.ai

As AI grows, so does the room for misuse. ⚠️
Openness empowers, but it also exposes. The upside? More eyes on the code = a better shot at keeping AI honest.
Subscribe to our newsletter at chaosagents.ai

Sam Altman just dropped $250M into Merge Labs’ seed round, a direct Neuralink competitor. 🧠💰

Our question: How does an active tech CEO drop that kind of money into a high-stakes rival frontier startup? Strategic vision or governance nightmare?

Subscribe to our newsletter at chaosagents.ai

I had to warn my AI: Don't embarrass me. 😳

Subscribe to our newsletter at chaosagents.ai

Ever feel like the algorithm knows you better than your friends do? 🤖

That’s Algorithmic Transference in action.

Subscribe to our newsletter at chaosagents.ai

Cut through the AI hype. 🧠⚡️
Two technologists. Learning out loud.

Subscribe to our newsletter at chaosagents.ai 🤖

The biggest mistake tech companies are making right now? Forcing AI into workflows that already work.

Slapping an unnecessary AI tool in their face doesn't make the product "modern", it just adds friction and slows them down.

Drop your email and build the future with us chaosagents.ai

What if you didn't need a multi-million-dollar data center to train the next frontier AI model?

Can we train large-scale AI models across slow, cheap, consumer internet connections instead?

Drop your email and build the future with us chaosagents.ai

Walked into a room of students expecting peak AI hype (9/10). They were at a 5 or 6, asking: "How do we reconcile the environmental impact of building this?" 🌍

The next generation is already at real-world consequences.

Drop your email and build the future with us chaosagents.ai

Friendly reminder for founders: Validation is not the same as support.

Falling into the "Mom Trap" relying on the praise of people who love you and want you to succeed is a quick way to build something nobody actually needs.

Drop your email and build the future with us chaosagents.ai

Tired of the AI hype cycle? Same.

We’re two technologists learning out loud, breaking down what actually matters in tech and ethics. No fluff, just the real shift. 🧠🤖

Drop your email and build the future with us chaosagents.ai

Cut through the AI noise. 🔇
Join two technologists learning out loud. 🧠🤖

Subscribe to the Chaos Agents newsletter and learn with us: chaosagents.ai

Generative UI is a classic case of "just because we can, doesn't mean we should."

Human habits don't vary that much—and when you destroy consistency, you destroy usability.

🎧 Want the full breakdown? Tune in → chaosagents.ai