|
|
|
TikTok's parent company just dropped a full AI model lineup at once. ByteDance announced Seedance 2.5 at its Volcano Engine conference this week, a video model that can generate a full 30-second native video clip in a single pass. Most AI video tools today top out at 10-15 seconds before you have to start stitching clips together like a TikTok editor in 2018. |
Oh, and they also dropped a new flagship language model, an image model, and an audio model in the same keynote. One conference. Four models. Meanwhile, some labs take six months to ship a blog post. |
The model isn't live for most people yet (early July is the target), but benchmarks already claim it rivals Google's Veo 3 on video quality, at a fraction of the price. The gap between Chinese AI labs and Western ones is closing fast. Which is either exciting or terrifying, depending on how you feel about the company that also owns TikTok. |
Here’s what happened in AI today: |
😸 Meta launched new $299 AI glasses, no Ray-Ban branding, Kylie Jenner edition included. 📰 IBM joined OpenAI's Daybreak cybersecurity program to find enterprise software vulnerabilities faster. 📰 Groq raised $650M to expand its AI inference cloud after Nvidia licensed its chip tech. 📰 Companies are cutting jobs over AI productivity gains that haven't actually arrived yet. 🎓 AI Skill of the Day: Why every SKILL.md needs a PITFALLS.md sitting next to it.
|
…and a whole lot more that you can read about here. |
Hey: Want to reach 700,000+ AI-hungry readers? Advertise with us! |
P.S: Love robots? We’re starting a new robotics newsletter! Sign up early here. |
|
😸 Meta Wants To Be on Your Face, For $299 |
Mark Zuckerberg just made his cheapest pitch yet for the AI-glasses era. |
Meta launched Meta Glasses on Tuesday, a new line of AI smart glasses starting at $299. That's at least $80 cheaper than the Ray-Ban Meta models, and nearly $500 less than Snap's AR Specs that dropped last week. These are available now in 10+ countries including the US, UK, Canada, and Australia. |
Here's what happened: |
Three frame styles: the Adventurer (rectangular), the Fury (boxy), and one designed with Kylie Jenner (slim oval) Built-in camera, open-ear speakers, and Meta AI (voice-activated and hands-free, with live translation coming in 14 new languages including Hindi, Mandarin, and Japanese) 8+ hours of battery, 40 hours with the charging case Powered by Muse Spark (Meta's first model out of its new Superintelligence Labs)
|
Unlike Snap's $2,195 glasses, these don't have a display in the lenses. But they let you speak to Meta AI about what you're seeing in real time, getting directions, translations, or answers to questions hands-free. |
Why this matters: Smart glasses shipments surged 167% in Q1 2026, and Meta already owns 69% of that market. Zuckerberg has called glasses "the main way people will access personal superintelligence," and at $299, he's clearly trying to make that reality before Google (partnering with Warby Parker), Apple, and OpenAI can ship theirs. |
The strategy is straightforward: own the hardware platform at scale first. At $299, these are closer to "impulse buy for someone who already wears glasses" than "niche tech gadget." There are ongoing privacy questions (including how footage from the built-in camera is handled), but demand has clearly outpaced those concerns. The daily number of people using Meta's glasses has tripled year-over-year. |
Our take: The glasses race is officially on. Meta is moving fast and cheap while everyone else is moving slow and expensive. The real test isn't whether people buy them; it's whether they reach for them daily over their phone. And whether the Kylie edition comes with a custom chime sound when you put them on. (It does. We're not making this up.) |
|
|
|
|
Plenty of companies can launch an AI pilot. Far fewer know how to make it stick. Explore this resource hub, sponsored by Dell AI Factory with NVIDIA, for strategies, decisions, and real-world lessons on turning AI into something scalable, useful, and worth the investment. |
Learn More: |
|
|
You've probably run into this: you correct an AI mistake in one session, and then the exact same mistake comes back next session like nothing happened. |
The fix, according to Maximiliano Contieri, is simple: pair every SKILL.md (a reusable instruction file you give AI coding tools) with a PITFALLS.md in the same folder. One file tells the AI what to do. The other tells it what not to do. |
Here's how to structure each entry: |
Trigger: what situation caused the bad behavior Wrong behavior: what the AI did Correct behavior: what it should have done
|
Then reference PITFALLS.md from inside your SKILL.md so the AI loads it automatically every session. Treat the file as append-only: never delete entries, even after a problem seems fixed. Solved pitfalls can come back after a skill update. |
## Don't use regex to count H2 sections
Trigger: counting sections by heading level
Wrong: regex-based heading detection (/^##/m)
Correct: match section names explicitly by string
Reason: code blocks with # fool regex heading counters
|
Contieri calls it "the scar tissue that lives next to the blueprint." Which is a great way to think about it. |
Total AI beginner? Start here (goes with this video). |
Have a specific skill you want to learn? Request it here. |
|
|
|
|
|
|
|
Did you know we have a podcast (The Neuron: AI Explained) where we talk to fascinating people in the industry who teach us how it actually works? Check it out: |
 | Click to view these episodes on YouTube! |
|
New episodes air every week on: Spotify | Apple Podcasts | YouTube |
|
📰 Around the Horn |
 | Anthropic's Claude Tag turns Claude into a Slack teammate you can @mention in any channel to delegate real work, with full context from the conversation thread already baked in. It's available on Team and Enterprise plans now, with existing Claude-in-Slack integrations switching over on August 3. |
|
Trump signed two quantum executive orders requiring federal agencies to migrate to quantum-resistant encryption by 2030-2031, and separately pushing to build a large-scale quantum computer at a Department of Energy facility by 2028. SpaceX signed a $6.3B compute deal with open-source AI startup Reflection AI, which will pay $150M per month for Nvidia GB300 chip access at SpaceX's Colossus 2 data center in Memphis through 2029. NVIDIA launched Halos for Robotics, the industry's first full-stack safety system for humanoid robots working alongside humans in factories and warehouses, with Agility Robotics as the first adopter. IBM joined OpenAI's Daybreak Cyber Partner Program, launching a security service that uses OpenAI's models to find and validate software vulnerabilities faster than traditional scanning tools. Groq raised $650M to expand its AI inference cloud (the infrastructure that runs AI models at scale), six months after Nvidia licensed its chip technology and hired away several top executives. Meta paused its Model Capability Initiative (an internal program that records employee keystrokes, mouse clicks, and screen activity to train AI) after a permissions error made all that sensitive data readable by everyone at the company.
|
|
|
|
|
For business owners and AI enthusiasts who want more than browser-based tinkering, Dell Pro Max with GB10 brings NVIDIA GB10 Grace Blackwell power to a compact AI development desktop. |
Check out the Dell Pro Max with GB10. |
|
Midweek Wisdom: |
Screentime correlates more with kids' brains than IQ (Erik Hoel, The Intrinsic Perspective) — a roundup that also covers why literary fiction awards keep falling for AI-generated submissions and whether AI app quality is keeping pace with model quality; worth the read if you've been wondering why the apps feel dumber than the models powering them. The AI shenanigans that I'm not into (Stephanie, The Simple Freelancer) — a freelancer's honest take on which AI use cases feel genuinely useful vs. which ones just feel gross; good gut-check if you've been using AI in your work and wondering where the line is. Your curiosity is your superpower in an age of AI (Midlife Unfiltered) — argues that as AI gets better at answering questions, the skill that compounds is knowing which questions to ask in the first place. Alibaba's T-Head chip entity just tripled its registered capital (SemiAnalysis) — a quiet but telling filing: China's internal AI chip ambitions are accelerating even as U.S. export controls tighten on the outside. World Models in Pieces: Structural Certification for General Agents (arXiv, ICML 2026) — a new framework for formally verifying AI agent behavior in modular pieces rather than all at once; worth a skim if you're thinking seriously about when it's safe to actually trust an agent. OpenThoughts-Agent: Data Recipes for Agentic Models (arXiv) — open-source training recipes for agents that can reason and act, extending the OpenThoughts reasoning model work into the agentic world; the kind of paper that shows up in every serious agent project six months from now.
|
|
|
|
|
| That’s all for now. | | What'd you think of today's email? | |
|
|
P.S: Before you go… have you subscribed to our YouTube Channel? If not, can you? |
 | Click the image to subscribe! |
|
P.P.S: Love the newsletter, but only want to get it once per week? Don’t unsubscribe—update your preferences here. |
