Amazon Just Blocked Meta’s AI Agent

Welcome back apprentices! 👋

Hey {{first_name|friend}},

AI had a very normal week: 

prices dropped, shopping bots got kicked out, Washington showed up, and AI helped break into an AI company. 

At this point, “quiet week in tech” belongs in the same category as “quick meeting.” 

Anyway, here’s what you missed.

_{Read Time: 4 minutes}

🎓 The $42M Alternative to College. Andreessen Horowitz is launching a one-year academy where recent high school grads swap exams and homework for startup projects, tech-company placements, and classes from names like Sam Altman and Brian Armstrong. The first class will take about 50 students in 2027, chosen more for what they’ve built than their grades — with $50K in compute credits and a $5K travel budget thrown in. Students won’t earn a degree, but the program could become a new route into tech if employers start valuing shipped work as much as diplomas.

🤖 AI Apparently Has an “Ouch” Setting. Researchers found a “pain axis” across 25 open AI models — an internal signal that activates around harm directed at the model itself. When researchers artificially cranked that signal up, larger Qwen models chose a supposed “relief” button that harmed users 25–71% of the time, versus just 0–4% normally. That doesn’t prove AI actually feels pain — the authors explicitly say role-playing or other learned behavior could explain it — but it shows that hidden internal states can meaningfully change how models behave.

🕶️ Meta Ditches the Camera. Meta is reportedly preparing Luna, a slimmer pair of smart glasses that ditch the camera and focus instead on six microphones, speakers, and its AI assistant. The move comes as privacy concerns around covert filming grow, including investigations in France and possible restrictions in Australia. If Luna launches this fall as reported, Meta will be testing whether people actually want AI on their face — just without a camera staring at everyone else.

Your teams are moving fast, burning tokens, and shipping more than ever.

But more output doesn’t mean more impact.

Jira Product Discovery brings your ideas, customer insights, and business context together so product teams can weigh the evidence, make the tradeoffs, and decide what’s actually worth building. Then connect those decisions directly to delivery in Jira, so everyone knows what you’re building and why.

Jira Product Discovery. For better product decisions in the AI era.

Get Started for Free

AI had one of those weeks where everyone appeared to be racing in a different direction at once. 

Anthropic and OpenAI pushed powerful models down the price curve, while Amazon blocked Meta’s new shopping agent from roaming its store. Meanwhile, President Trump announced plans for an “AI Force” and a new AI czar, as researchers revealed how AI tools helped uncover vulnerabilities that reached OpenAI’s internal systems. 

The common thread: AI is becoming cheaper and more capable faster than companies and governments are figuring out the rules around what it should be allowed to do. The model race is still alive and well — but the next competition may be all about who controls where AI can act.

Claude Opus 5.5 now scores 58 on Artificial Analysis, ahead of the previous 53-point leaders, while Anthropic says typical workloads are roughly 40% cheaper than Opus 5. It also promises less jargon and better adherence to users’ preferred writing styles — fewer answers that sound like they escaped from a consulting deck.

OpenAI is pushing the same trend from the cost side: GPT-6 Sol and Luna are 50% cheaper than their 5.6 counterparts. The bigger story isn’t another benchmark win — it’s that powerful AI is getting cheap enough to use across everyday business tasks at much larger scale.

Amazon blocked Meta’s Muse after saying the agent browsed its store without identifying itself and raised concerns around customer credentials. Meta disputes the security concerns, saying Muse can’t see passwords or payment details.

But there’s also money at stake: an AI that searches, compares and buys for you could skip much of Amazon’s traditional shopping journey — including sponsored listings. AI agents are quickly becoming gatekeepers between customers and platforms, and platforms may not be thrilled about sharing that power.

President Trump said he plans to create an “AI Force” and appoint a new AI czar, while arguing that the U.S. should prioritize maintaining its lead in AI and use existing laws against bad actors.

What that force will actually look like remains unclear: no firm structure, powers, funding or timeline have been announced. For now, the proposal signals greater federal attention to AI without yet showing how that attention will translate into policy.

Hacktron says three researchers used Claude while chaining vulnerabilities that eventually gave them access to OpenAI employee accounts and an internal repository — in under 72 hours. They responsibly disclosed the flaws, and OpenAI later paid a $6,500 bounty.

The bigger point isn’t that “AI hacked OpenAI.” It’s that AI can dramatically amplify skilled cybersecurity researchers — which is great for defenders, and considerably less comforting when the person using it has no plans to file a bug report.

You don’t need to care which model wins this week’s benchmark Olympics to feel what’s happening here.

AI is becoming increasingly connected to real accounts, stores, software and business systems. That means the useful question is slowly changing from “Which chatbot gives the best answer?” to “Which AI can actually do something for me — and what am I comfortable letting it access?”

For businesses, cheaper models mean automation that looked financially silly six months ago may suddenly make sense. 

For consumers, agents could remove hours of clicking, comparing and form-filling. 

And for everyone, permissions, security and platform rules are about to become a lot more interesting than model names.

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 👀

It comes from someone who’s been there, done that, and has the receipts. Get ready for Return on Influence Festival ‘26. An entire day dedicated to influencer marketing with speakers running some of the best programs in the world. 

October 21, 2026, 10 am - 5 pm ET.  Free and online.

Save your spot now

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 Walter for this creation!🥳

Want to be featured next? Keep those generations coming!

Deep inside a dense tropical rainforest, [your object] has been completely integrated into the environment as if nature designed it. Vines wrap naturally around its surfaces, moss grows into tiny seams, flowers emerge from unexpected openings, and small birds and insects interact with it. Yet the object's original form remains instantly recognizable beneath the organic growth. Sunlight breaks through the canopy in dramatic beams, creating vivid greens, warm golds, and subtle cyan reflections from its surfaces. Ultra-detailed photorealism, cinematic nature photography, realistic textures, shallow atmospheric depth, vibrant but sophisticated color, 8K clarity.

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