In partnership with
Hey {{first_name | there}},
A man with no left hand put on a headset and moved a robotic arm using only his thoughts.
This wasn't a lab demo buried in an academic paper. This happened at WAIC 2026 in Shanghai, in public, on stage.
China's BrainCo just unveiled what they're calling the world's first integrated Brain-to-Robot AI platform. You wear an EEG headset. AI reads your brain signals, decodes your intent, and sends commands to a robot, all in under 200 milliseconds.
For context: 200 milliseconds is roughly how long it takes you to blink.
Now I don't want to be the guy who says 'this changes everything', but I do want to point at what this represents when you put it next to everything else happening this week.
Alibaba Dropped Qwen 3.8
2.4 trillion parameters
To put that in scale: GPT-4 is estimated to have around 1.8 trillion. Qwen 3.8 claims to be second only to Anthropic's Fable 5. It's multimodal: text, images, video, documents. It has a 1 million token context window.
This dropped days after Moonshot AI released Kimi K3 as open-weight.
The Chinese AI race isn't just matching the West anymore, it's escalating on its own terms, on its own timeline.
But here's what I actually want to talk about.
Not the spec sheets or benchmark wars. But the implication of “thought as input”.
What Happens When Intent Becomes the Interface?
Right now, the bottleneck between you and AI is language.
You have to translate what you want into words. Words into prompts. Prompts into outputs. It's a lossy process, something always gets lost in translation.
BrainCo's demo hints at a world where that bottleneck compresses toward zero. You think it. The machine does it. No language overhead.
This is not commercially available yet. And the current tech requires wearing an EEG headset, which has obvious limitations. But the trajectory matters more than the current state.
Neuralink is already doing invasive BCIs in humans. BrainCo is showing non-invasive options that work well enough to move robotic arms at the speed of a blink. The gap between "interesting demo" and "consumer product" is closing faster than most people expect.
And Qwen 3.8 going open-weight matters for a different reason. When models this powerful become free to run, access to frontier AI stops being gated by which API you can afford.
A developer in a city with no access to OpenAI enterprise contracts will be able to run a 2.4-trillion-parameter model locally. That's not a technical detail, it's a redistribution of who gets to build with the most capable tools.
What This Means If You're Building Right Now
Two things are converging:
1. The interface between human and machine is collapsing, from GUI to voice to text to, eventually, direct neural input
2. The raw intelligence available to builders is becoming democratized through open weights
If you're building AI products, the combination of these two trends suggests something important: the competitive advantage is no longer "access to a powerful model." It's what you build with it, and for whom.
Because when everyone has access to a 2.4-trillion-parameter model and robots you can control with your thoughts, the only differentiator left is whether you understand the human problem deeply enough to solve it in a way no one else has thought of.
Same as always. Just faster. And with better tools.
What are you building that couldn't have existed 12 months ago?
I'm genuinely curious. Hit reply.
– Aashish
P.S. The AI Enthusiasts community is where I share stuff like this before it hits the newsletter, come join: https://chat.whatsapp.com/IXt9FJIblNs8tu36JIwWbd
You've seen the AI demos. Viktor does it without you watching.
The AI tool you tried last quarter waited for a prompt, hallucinated a number, then asked if you'd like a summary.
Viktor opened a PR at 2am, rebased it against main, ran your test suite, and posted a note in #eng: "Two flaky tests in payments service, both pre-existing. Recommended merging after fixing them." Then drafted the customer reply for the support ticket the bug created.
That's 619K autonomous actions per day across 20,000+ teams. Not chat replies. Real work shipped to GitHub, Stripe, Linear, Notion, and 3,000+ other tools, from inside Slack and Microsoft Teams.
You don't supervise him any more than you supervise a senior engineer.
SOC 2 certified. Your data never trains models.
"It's what you probably originally thought AI was going to be when you first heard of it in sci-fi movies." Tyler, CEO.