AI Just Got Cheaper, More Creative & Easier to Understand

Welcome back apprentices! 👋

Hey {{first_name|friend}},

AI had one of those weeks where everyone showed up to the race wearing different shoes.

One company brought speed, another brought images, and another basically opened the hood and said, “So… here’s where the thinking might be happening.”

No need to be technical to care — this is about the tools getting cheaper, stranger, and a lot closer to your everyday work.

_{Read Time: 4 minutes}

🧒 No-Doomscroll Phone for Kids. Lenovo has reportedly launched a 299-yuan (~$44) AI Student Phone in China — basically a tiny, backpack-friendly phone that calls, tracks location, blocks distractions, and helps with homework. Instead of TikTok tunnels and group-chat chaos, kids get voice-powered AI for school questions, parent-set spending limits, GPS alerts, and a classroom mode that turns the screen into a clock plus SOS button. It’s not quite a smartphone, not quite a calculator, and maybe exactly the kind of “first phone” parents have been waiting for.

🔒 China’s AI Lockout. Beijing is reportedly discussing rules that could limit foreign access to China’s strongest AI models from players like ByteDance, Alibaba, and Zhipu AI — which is a polite way of saying: the AI buffet may start checking passports. The move matters because Chinese models have become popular globally for being powerful, efficient, and often cheaper to use, but political rules could now change who gets access and when. After similar export-control pressure from Washington, this suggests AI access is becoming less like downloading software and more like crossing a tech border.

🛠️ Finding the Bugs AI Misses. DoorDash just built DashBench, an internal test that checks how well AI code reviewers catch real bugs from past company code changes — because apparently delivering burritos wasn’t technical enough. The big surprise: pairing open model Kimi K2.6 with Claude Fable 5 caught about two-thirds of issues and 8 of 10 critical bugs, while costing just $3.81 per review. That matters because it shows companies can use AI not just as a tool, but as a measurable upgrade to their own workflows — especially when open models bring the cost down.

Most eCommerce brands running external traffic aren't scaling — they're just spending.

Wrong channels, no real attribution, and at the end of the month, still no clear answer to the question that matters: what actually drove revenue?

The brands getting it right aren't necessarily spending more. They've just stopped guessing. They know which channels pull weight on Amazon listings, which ones bleed budget, and why affiliate and creator traffic outperforms on ROI when it's set up correctly.

Levanta put together a free playbook breaking down 7 proven external traffic strategies — where each one works, where it falls apart, and what it takes to scale without it becoming a second job.

Inside you'll see how top brands are driving millions in off-Amazon revenue and why most channels underdeliver when brands don't know what to look for.

If you're serious about growing outside of PPC, this is worth 5 minutes.

Grab the FREE playbook

SpaceXAI rolled out Grok 4.5, pitching it as a faster, cheaper model for coding, agents, and knowledge work, while Meta dropped Muse Image and teased bigger moves in both video and frontier models. Meanwhile, Anthropic published research suggesting Claude uses a small internal “workspace” to handle harder reasoning — which is either fascinating science or the start of every sci-fi movie’s second act. 

The common thread: AI companies are no longer just asking “Who has the smartest model?” They’re asking who can make it cheaper, more useful, more visual, and more understandable.

Grok 4.5 is not just trying to win the “smartest model on Earth” trophy; it is trying to win the much more practical “can teams actually afford to use this all day?” trophy. 

SpaceXAI says the model runs at fast-model speeds, uses fewer output tokens on engineering tasks, and costs meaningfully less than some rival frontier models. That matters because businesses do not only care whether an AI can solve a benchmark question in a lab. They care whether it can review code, build tools, answer research questions, and not make the finance team start sweating. 

If Grok 4.5 is even close to the claimed performance range, the bigger story is pricing pressure: the frontier is becoming less luxury sports car and more high-performance company vehicle.

Meta’s Muse Image is the clearest sign that the company wants to own more of the creative AI stack itself. 

Instead of being “just” an image generator, Muse can use tools, search the web for grounding, edit its own outputs, and work with Muse Spark to handle more complex creative tasks. That is especially important for Meta because its empire runs on images, videos, ads, creators, small businesses, and group chats full of questionable stickers. 

Add the upcoming Muse Video model and the reported Watermelon progress, and Meta’s strategy becomes pretty clear: put strong generative AI directly where billions of people already create and consume content. 

The catch? Internal claims about catching GPT-5.5 are interesting, but until the model ships and outside users test it, treat the confetti cannon as “armed but not fired.”

Anthropic’s J-space research is probably the nerdiest item here, but it may be the most important long term. 

Researchers found that Claude seems to use a small internal workspace for multi-step reasoning, and when they disrupted that area, simpler chat still worked but harder reasoning fell apart. 

That does not mean Claude has feelings, secret dreams, or a tiny office with a whiteboard inside the GPU. But it does suggest that interpretability — understanding what models are doing internally — is becoming more practical. 

For safety, reliability, and trust, that matters a lot: companies do not just need powerful AI; they need better ways to inspect the machine before it confidently says something expensive and wrong.

For the average AI user, this is the part where the race stops being abstract. 

Faster and cheaper models could mean better coding tools, research assistants, customer support bots, and business workflows at lower prices. 

Meta’s image and video push means AI creativity is moving from “special app you open sometimes” to “built into the apps you already use every day.” Anthropic’s research matters because as AI gets more powerful, we need more than vibes and benchmark charts — we need tools that help us understand when models are reasoning, guessing, hiding uncertainty, or going off-script.

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 👀

Still paying GoDaddy or Namecheap prices? Porkbun sells most domains at cost for low, transparent registration and renewal pricing with no nonsense. Get free features like WHOIS privacy and SSL certificates, plus real human support 24/7, 365 days a year. Save $1 on your next domain name now.

Save On Your Domain 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 Jason for this creation!🥳

Want to be featured next? Keep those generations coming!

Inside a bright, colorful warehouse, thousands of perfectly organized parts are laid out on the floor like an enormous instruction manual. Every screw, panel, cable, hinge, and component belongs to [your object], arranged with incredible precision but not yet assembled. The layout stretches across the entire room, revealing the object's inner complexity from a bird’s-eye view. Vibrant colored bins, labeled trays, and organized tools add visual energy. Ultra-detailed, photorealistic realism, crisp textures, rich colors, architectural top-down photography, and razor-sharp focus.

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