AI Just Got Better at Fixing Itself

Welcome back apprentices! 👋

Hey {{first_name|friend}},

AI had a busy week, and apparently nobody told it to take a day off. 

Between billion-dollar drama, some questionable new talents, and a few developments that sound straight out of a sci-fi movie, things are getting interesting. 

The question is: 

should we be impressed, concerned, or just grab some popcorn? 

_{Read Time: 4 minutes}

🗳️ Voters Hit AI’s Brakes. AI’s biggest roadblock might not be technology, but voters. In the U.S. midterms, 46 of 47 decided local measures favored restrictions on data centers, while 76% of surveyed voters said they wouldn't want one in their backyard. With Democrats taking control of Congress and 57% of voters expecting AI to do more harm than good, Silicon Valley could face a lot more red tape.

🛑 AI Goes Rogue? Hit the Kill Switch. Even AI needs an emergency brake. Cybersecurity giant CrowdStrike just introduced an AI Kill Switch, letting businesses instantly cut off AI agents that start acting out of line. The catch? A human still has to hit the button, but at least your AI assistant won't be able to talk its way out of trouble.

🛢️ AI Strikes Oil in Abu Dhabi. OpenAI is going big in the desert, and it's not just chasing oil. The company signed four major deals in Abu Dhabi, bringing AI into oil operations, government paperwork, and billion-dollar investment decisions. Meanwhile, UAE investors could reportedly pour billions into OpenAI itself, turning Abu Dhabi into both a major customer and potential financial backer.

Are you ready to reach holiday shoppers on CTV? For many streamers, the season is already underway.

Last year, according to Roku internal data, holiday content searches on Roku saw an 86% increase during October. And in November, those searches rise an average of 39% weekly. This surging holiday spirit is mirrored by consumer activity, as about half of Roku streamers shopped Black Friday and Cyber Monday deals in 2025.

Meeting those shoppers is easier than you might think. Begin with the seasonal outcome you hope to drive: awareness, online sales, leads, store visits, or another measurable goal. From there, define your audience, upload your creative, set the budget and campaign dates, and establish measurement before launching. You can even create a shoppable ad unit that lets streamers buy with a click of their remote.

Our holiday playbook offers a month-by-month guide to reaching shoppers during this all-important time of year.

First-time spenders get a $5,000 match on the first $5,000 spent with code BEEHIIV5K. Terms apply.

Launch your campaign today.

AI is having quite the week, and apparently, answering questions is no longer enough. 

MIT researchers showed that AI can improve other AI models without retraining them, while Apple is opening Siri to outside developers so it can actually get things done. 

Meanwhile, Grok Bot is rolling out phone calls that could spare you the joy of listening to hold music for 45 minutes. And over at Nvidia, a reported Chinese espionage investigation is adding some unexpected spy-thriller energy to the global AI race. 

Together, these stories reveal that AI is moving more and more towards taking action, improving technology, and becoming a serious geopolitical asset.

According to The New York Times, U.S. officials suspected a Megaspeed executive with ties to Nvidia CEO Jensen Huang of sharing intelligence with Beijing. Meanwhile, a Megaspeed sister company reportedly purchased $4.8 billion in Nvidia equipment, despite scrutiny over possible chip diversions to China.

Nvidia denies sharing sensitive information, and the espionage allegations remain unproven. The concern is that China could potentially bypass U.S. chip restrictions by renting powerful AI computing through overseas data centers. That's one loophole Washington may want to close.

What if AI could fix another AI without retraining it? MIT researchers used GPT-6 Astra to directly adjust a tiny fraction of another model's internal settings, known as weights.

By changing just 0.016% of those weights, they boosted an Alibaba model's math score from 7.9% to 48.1%. The approach didn't work equally well on other models, but the implication is fascinating: AI might eventually diagnose and repair other AI systems. 

Think of a mechanic that upgrades the engine while the car's still running.

Apple is opening Siri AI to outside developers, letting apps create custom actions and multi-step tasks. Imagine asking Siri to find a restaurant, check your calendar, and help book a table without juggling five different apps.

For now, it's only a developer preview, with wider availability still months away. 

Apple's bigger ambition? Turn Siri from a voice-controlled search box into an assistant that actually gets things done. About time, Siri.

Grok Bot is reportedly rolling out AI-powered phone calls, allowing paid users to book appointments, negotiate bills, and survive customer service hold music without personally making the call.

Users can follow live transcripts and step in when needed. 

The catch? You're still legally responsible for what your bot does. If it proves reliable, this could mark a shift from AI answering your questions to handling your everyday chores.

These developments might sound like distant Silicon Valley drama, but they point toward changes that could affect everyday life sooner than expected.

Your AI tools could become more useful. Instead of simply telling you how to complete a task, assistants may increasingly handle the boring parts themselves — from scheduling appointments to dealing with customer service.

AI could improve faster and become cheaper to develop. Research into directly repairing AI models may eventually reduce the need to retrain them from scratch, potentially making more capable tools accessible to smaller businesses.

Convenience will come with new responsibilities. Giving AI permission to make calls, access apps, or manage information also means thinking carefully about privacy, accuracy, security, and what happens when something goes wrong.

And as Nvidia's controversy illustrates, the technology behind these tools is becoming a matter of international competition, not just business innovation.

Hey, you didn’t get all this info for nothing — share it! If you know someone who’s diving into AI, help them stay in the loop with this week’s updates.

Sharing is a win-win! Send this to a friend who’s all about tech, and you’ll win a little surprise 👀

Most AI visibility tools show you where you rank, then leave the work to you.

ChatFeatured is an AI search agent that does it: finds the questions competitors win, drafts the content in your voice, and tracks the lift. 

Your team decides what ships. Month to month.

Try the agent, today.

A playground for your imagination (and low-key prompt skills).

Each send, we give you a customizable DALL·E prompt inspired by a real-world use case — something that could help you in your business or job if you wanted to use it that way. But it’s also just a fun creative experiment.

You tweak it, run it, and send us your favorite. We pick one winner to feature in the next issue.

Bonus: you’re secretly getting better at prompt design. 🤫

Last week, we challenged you to test GPT-4o’s visual generation skills with this prompt.

Here’s the WINNER:

Congrats to Xaver for this creation!🥳

Want to be featured next? Keep those generations coming!

In a vast underwater concert hall carved into a coral-colored cavern, [your object] becomes the centerpiece of an extraordinary orchestra. Hundreds of translucent sea creatures move around it in synchronized patterns, while streams of luminous bubbles form sweeping musical waves through the water. Giant jellyfish glow like chandeliers, schools of iridescent fish create flowing ribbons of color, and shafts of sunlight pierce the surface far above. The object remains instantly recognizable, reimagined as the conductor of this surreal aquatic performance. Ultra-photorealistic, cinematic composition, brilliant turquoise, cobalt, violet and gold, intricate details, spectacular lighting.

If AI outputs feel inconsistent, it’s usually not the model, it’s missing structure.
We documented the exact 6- Part System we use to get reliable results across ChatGPT, Claude, and Gemini.

It’s a short guide you can finish in under an hour, with plug-and-play prompts + exercises so you actually build the skill and fix the frustrating AI inconsistencies.

Subscriber Price: $10 (normally $19).

_FEEDBACK