Did you know that we have LinkedIn, Instagram and X accounts
that you can follow?
Hi {{first_name|everyone}}, Reflection AI raised over $4 billion before releasing a single model, and now Beam is here, an open-weight model that's meant to take on the best Chinese options. After that, we've got a few handy tools, a couple of quick AI updates, and one of the more interesting AI-made films we've seen lately.
Our friends are also sharing a hands-on Claude Code bootcamp this week, so keep an eye out for that.
Alright, let's dive into it.
🤿 DEEP DIVE
Pentagon’s Partner Finally Has a Model
Reflection AI raised more than $4 billion before releasing its first AI model.
Now we know what all that money was building toward.
The US start-up has unveiled Beam, an open-weight model it says can compete with Z.ai’s GLM-5.2, one of the leading Chinese models in a part of the AI market where Chinese labs have become unusually strong.
And that open-weight part is really the story here.
Unlike the most advanced versions of ChatGPT or Claude, open-weight models can be customised and run on a company’s own hardware. That can make them much cheaper to operate.
Demand for them is also growing quickly. In August, open-weight models accounted for 56% of all tokens processed through Vercel’s AI Gateway. Back in December, it was 7%.
The problem for the US is that many of the strongest options are Chinese.
So Reflection is arriving at a useful moment
Reflection says Beam matches GLM-5.2 and beats other US open-weight models on coding and agentic tasks while using a fraction of the token cost.
There is an important caveat, though. The company hasn’t published the full technical details behind those performance claims yet.
Still, Reflection has some serious backing.
Nvidia has invested $800 million and given the company access to its chips. Reflection has also partnered with the Pentagon and the US Department of Energy, while working on AI models for US allies including South Korea.
That makes this about more than another model launch.
The Trump administration has been pushing American companies to build stronger open-weight alternatives, partly because US businesses are increasingly using Chinese models. Treasury secretary Scott Bessent has even raised the possibility of sanctions against Chinese models suspected of taking intellectual property from American competitors.
At the same time, Nvidia has argued against shutting those models out completely. Jensen Huang has said Chinese open-weight models offer good value for American companies, while Nvidia helped lead an industry letter arguing that a strong open-weight ecosystem is important for US innovation.
So there’s a slightly awkward tension here.
American companies want the cheap, customisable models Chinese labs are producing. Washington is worried about depending on them. And now Nvidia is helping fund an American alternative.
The quick take
Reflection AI has launched Beam, its first open-weight AI model
It says Beam performs on par with Z.ai’s GLM-5.2, although full technical performance details have not yet been published
Nvidia has invested $800 million in Reflection and provided access to its chips
Reflection has partnerships with the Pentagon, the Department of Energy and US allies including South Korea
The launch comes as US officials push for stronger American open-weight models to reduce reliance on Chinese alternatives
🤝 FROM OUR FRIENDS
A quick share we thought you might find useful
Build with Claude Code.
Prepare for Claude Certified Foundations.
Work through 10 hands-on projects and learn CLAUDE.md, reusable Skills, subagents, hooks, guardrails, testing, and more.
Get practical Claude Code experience plus downloadable Skills you can reuse after the workshop and a free full-length mock test to benchmark your knowledge and prepare for Anthropic certification.
Sunday, October 18 | Starting at $189.99
⚡ QUICK HITS
The latest in AI, tech, and productivity worth knowing
OpenAI says it will start watermarking ChatGPT and Codex text in the EU by subtly shaping word choices so a detector can recognize the pattern. The company says editing can weaken it quite a bit, with one test dropping detection from about 92% to 66% after replacing 10% of words with synonyms.
PewDiePie unveiled Ajax, a fine-tuned Qwen3.5-9B model for his self-hosted Odysseus workspace, and says OpenAI banned him twice while he was developing it. In his video, he says one ban specifically cited “distillation,” after he used another model to help create training data for Ajax.
⚒ TOOL SNAPSHOTS
Futuristic tools within AI, no-code, and productivity
Gets a social video surprisingly close to post-ready from a script.
Why it’s useful: Handy when you want presenter-style social content without piecing together B-roll, captions, music, and edits yourself.
Gives your app one place to choose and manage LLMs.
Why it’s useful: Useful for teams that don’t want model choice, provider outages, and cost controls scattered across a bunch of separate integrations.
Lets a coding agent do the video editing while you direct.
Why it’s useful: Kind of an interesting setup for people already using Codex or Claude Code who’d rather describe edits and refine the result than build every cut manually.
🤖 AI MADE THIS
Interesting and inspiring creations made with AI
BENCH - an AI film that just won Best AI-Generated Film
BENCH uses generative AI for something surprisingly simple - a story about an elderly man who promises to run a race with a young friend. It just beat a global field to win Best AI-Generated Film at the Sparknify Human vs. AI Film Festival, where judges praised it for pulling off something AI films often struggle with: genuine emotional storytelling.
AI used
Generative AI filmmaking but no models mentioned
ℹ️ ABOUT US
The Intelligent Worker helps you to be more productive at work with AI, automation, no-code, and other technologies.
We like real, practical, and tangible use-cases and hate hand-wavy, theoretical, and abstract concepts that don’t drive real-world outcomes.
Our mission is to empower individuals, boost their productivity, and future-proof their careers.
We read all your comments - please provide your feedback!
Did you like today's email?
What more do you want to see in this newsletter?
