Tag: AI

Last week, shortly after Anthropic launched its most powerful AI models, the U.S. government imposed export controls restricting foreign nationals from accessing them – and rather than try to verify the citizenship of every user on the planet within ninety minutes, Anthropic shut the models down for everyone. In this video we look at what actually happened: the Commerce Department’s “is informed” letter, the deemed-export rules that locked Anthropic’s own engineers out of their work, the claim that the “national security threat” was essentially an AI fixing software bugs, and the awkward detail that the partner who reported it was Amazon – Anthropic’s largest backer and a direct competitor. We also dig into why all of this matters for Anthropic’s near-trillion-dollar valuation, the winner-takes-all assumptions behind frontier AI, the rise of cheap open-source Chinese models, and whether there’s really any such thing as a monopoly on math.

Jonathan Pageau closed out Day 2 of ARC 2026 with a keynote that left the whole room speechless. Amidst the many voices surrounding the global AI debate, Pageau offered a simple question that stopped us in our tracks: “Who does it serve?”

With the return of the ARC Conference this June, we are publishing a brand new book — the most comprehensive synthesis of ARC’s philosophy to date. Drawing on Augustine, Tolkien, and a long tradition of civilisational thinking, The Age of Reconstruction charts a path from diagnosis to renewal. While many can point out where things have gone wrong, this book does what most can’t: it lays out a vision of what we must build, how, and why it is still possible.

Praise for The Age of Reconstruction:

“Will Western Civilisation flourish, or flounder? It depends on us, and the decisions we make right now. And Johnny Patterson’s wonderful ‘The Age of Reconstruction’ gives us the formula, based on timeless ideas and wisdom.” — Arthur Brooks, #1 New York Times Bestselling Author

__

Jonathan Pageau is an artist, writer and public speaker. He is a pioneer in the revival of Liturgical Art for the 21st century. Through his podcast, The Symbolic World, and publishing company, Symbolic World Press, he fosters the rediscovery of symbolic thinking and a vision for re-enchantment in the world.

LTX Director is a free open source all-in-one tool for creating AI Videos. Version 2.0 is a complete overhaul of the previous version, giving you total creative control over your AI generations.

LM Studio local AI just changed: NVIDIA’s DGX Station packs 748GB unified memory to run 70B models in full precision, no cloud needed.

LM Studio and Ollama finally lose their asterisk, NVIDIA’s DGX Station at Computex 2026 lands 748GB of coherent unified memory in a single deskside tower, enough to load a full-precision 70B model with room to spare, no quantization required, no cloud offload.

Announced by Jensen Huang at GTC Taipei on May 31, 2026, the machine is built around the GB300 Grace Blackwell Ultra Desktop Superchip: a 72-core ARM Grace CPU fused to a Blackwell Ultra GPU via NVLink-C2C at 900 GB/s. The memory pool splits into 252GB HBM3e (7.1 TB/s GPU-side) and 496GB LPDDR5X (CPU-side), both fully coherent, one address space, zero explicit copies. Compute tops out at 20 petaFLOPS FP4. NVIDIA doesn’t sell a Founders Edition; OEM partners ASUS, Dell, HP, MSI, and others handle that, with real-world pricing landing between roughly $85K and $115K (the MSI XpertStation WS300 lists at $96,995.99 on CDW). The video also covers NVIDIA’s DGX Spark (128GB, ~$4,700) as the genuine prosumer entry point, and gives an honest head-to-head with the Mac Studio M5 Ultra, which still holds the value crown for a solo developer running mid-size models. The trillion-parameter claim gets a reality check, it’s technically true only with aggressive 4-bit quantization, not full-precision weights. The cloud ROI math is real: at ~$98/hour for a comparable AWS p5 instance, the hardware pays for itself in roughly two months of sustained workload. The DGX Station for Windows (WSL-based) is flagged as a Q4 2026 promise, not a shipping product. RTX Spark, NVIDIA’s MediaTek-partnered consumer AI PC chip, rounds out the roadmap alongside a three-generation plan through Rubin and Rosa Feynman.

For individual builders and small teams deciding between local AI options, this is a practical breakdown of which box on the NVIDIA ladder actually makes sense for their workload.

In this episode, I follow up on a few of my predictions about AI from my recent video: “How AI Will Fail Like The Music Industry.”

Everyone spent the week breaking down Seedance 2.5’s specs — 30-second shots, native 4K, up to 50 references. But at the same event, ByteDance quietly announced something that might matter a lot more: a way to make AI videos with real movies and a real movie star… legally, with the rights-holder actually getting paid.

It’s called Volcano Ark, and the first partner is Stephen Chow’s company, Bingo Group — licensing scenes from King of Comedy, God of Cookery, and CJ7. Unlike the dead Disney–Sora deal (stylized, masked characters only — never the real face), this licenses real likeness AND voice, and even reproduces the original film’s look down to the film stock.

The unlock is the business model: a revenue-share that pays the rights-holder when you generate, or a commercial license you can buy. It’s already quietly live, running on Seedance 2.0 — not even 2.5 yet.

In this one I break down how it actually works, the fine print (free for personal use, commercial use blocked on the spot), why it’s basically the “TikTok music” playbook applied to movies, and whether anyone in the West — realistically only YouTube — could ever pull it off… one very cantankerous mouse permitting.

Seedance and Kling just changed what “fast” can do in AI video. The old “fast, cheap, or good — pick two” rule started cracking this week — Seedance Mini, Kling 3 Turbo, two real-time world models, and Google’s stealth “Instant Ramen” all landed at once.

And no single tool did it alone.

ByteDance’s Seedance 2.0 Mini is the cheap, fast tier that finally stopped being embarrassing. I ran the 15-shot Coda prompt through it, threw Flamethrower Girl at it for the reference-prompt stress test, and used it for Dragon Blue inserts and extra sword-fight coverage. Resolution is lower (1470×630 vs flagship 2206×946) — and Topaz Astra’s Starlight Precise mode closes some of that gap — but for inserts, drafts, and doubling your fight footage at half the price, it holds. Also in the rumor file: Seedance 2.5 is reportedly coming early July (via Testing Catalog) — name unconfirmed.

Kuaishou’s Kling 3 Turbo is basically Kling 3.0 quality at a lower price — and the Twin Peaks FBI-diner test showed no lip-sync drift, which is a bigger deal than it sounds. The Kling Omni upgrade still hasn’t dropped, and Kling 4.0 / 3.5 is still the thing everyone’s waiting on.

Then it gets strange. Alibaba’s Happy Oyster is a real-time, Genie-3-style world model you can actually walk around inside — three minutes free per day. I spent mine at Sunset at Bali Bay, added two kaiju to the ocean, and neither person on the beach seemed remotely concerned. It’s wonky. It’s also kind of remarkable. Maine Coon from Catnip AI is a separate real-time model, reportedly hitting a record 47.5 fps on a single H100 at 9:16 vertical — apply at mainecoon.tech.

Finally, Google quietly planted a stealth image model on LM Arena. The internet dug a Google ID out of the metadata and is calling it Instant Ramen — probably the next “Nano Banana,” probably Gemini 3.1 Flash Lite Image. “Probably” is doing a lot of work there.

The honest read: AI video isn’t suddenly solved. The cheap tiers are still compromises, real-time is still server-gated and weird, and Instant Ramen is still an unconfirmed leak. But the direction is hard to ignore now — here’s what you can actually use this week, and what’s still just a really good trailer.

Speaking at the VivaTech conference in Paris on June 17, the Jeff Bezos pushed back on fears that artificial intelligence will eliminate jobs. Bezos argued that AI will help people identify and solve more problems, ultimately increasing demand for workers. He also outlined his vision for the future of space exploration, including permanent space colonies, off-world manufacturing and moving polluting industries beyond Earth.

Adobe is making more money than ever, so why is Wall Street treating it like the company is dying? While the internet is distracted by a $150 million FTC lawsuit over hidden cancellation fees, a much deeper crisis is threatening the tech giant’s 40-year empire.

Behind the scenes of Adobe’s sliding stock price lies a massive strategic panic, a botched $20 billion merger, and a controversial shift into generative AI that has sparked a historic backlash from creators. The “industry standard” software isn’t just fighting off predatory subscription scandals, it’s actively building an automation ecosystem that could replace the very professionals who keep it alive.

Higgsfield’s MCP and CLI connectors have given Claude full image and video generation capabilities – but can an AI agent make a better film than a real person? Today I gave Claude $500 in Higgsfield AI credits to see who can make the better AI cinematic