TL;DR
Figure raised over US$1B and launched “Project Go-Big”
Open research (AGILOped, DreamControl) shows progress
Unitree G1’s security audit reveals serious privacy & safety risks.
Robotics startup Figure just raised over US$1B in funding. Alongside the money, they announced Project Go-Big: a plan to collect tons of “first-person” human videos (think GoPro-style footage) to teach robots by watching how we move and act.
With this data and their “Helix” control system, Figure says robots could learn new tasks just by watching human videos—no extra training for each robot.
The funding is confirmed. The project outline is public. But there’s no independent proof yet that it works outside of controlled demos.
Why it matters: If successful, this could be a huge shortcut to making robots useful in daily life. The risk? Real homes and workplaces are messy, unpredictable, and dangerous if robots fail. Overpromising here could backfire.
This milestone is critical to unlocking the next stage of growth for humanoid robots, scaling out our AI platform Helix and BotQ manufacturing.
— Brett Adcock, Figure founder & CEO, via company release
“Uncanny Intel Weekly” — Your weekly briefing on AI and humanoid robotics. Each week, we cut through the hype to deliver verified news, research breakthroughs, product launches, security alerts, and industry insights. Stay informed, stay skeptical, and know what really matters in the fast-moving world of intelligent robots
DEEP DIVE 2
Unitree G1 Security Audit
Security researchers took apart Unitree’s new G1 humanoid and found weak protection for its data. The robot was sending video, audio, and movement logs without clear consent—and the encryption (the digital lock) was flimsy.
These aren’t just “maybe” problems. The team showed real ways hackers could move from spying on the robot to actually taking control of it.
The preprint includes code analysis, network logs, and even proof-of-concept hacks. It’s not speculation.
Read the full report here.
Why it matters: If a humanoid like this goes into homes, factories, or hospitals with these flaws, it’s a serious security and privacy risk. Buyers should push for fixes, independent audits, and full transparency before trusting these machines.
IN FOCUS
AGILOped and DreamControl
Two open-source research teams dropped eye-catching humanoid projects this week
AGILOped is all about making robots walk like us. The system uses reinforcement learning (trial-and-error training in simulation) to give humanoids more natural, agile steps—even on uneven ground. In plain terms: it’s a push toward robots that don’t wobble like toddlers.
DreamControl takes a different shot: it trains robots with “dreamed-up” scenarios—AI-generated simulations that look like the real world. Instead of relying only on expensive, messy real-world data, the robot learns from these fake-but-plausible situations.
Together, these projects show two sides of the same coin. AGILOped is making robots move better in real life. DreamControl is making robots practice smarter before real life. Both could speed up how quickly humanoids become useful beyond labs and demos.
TARGET VS. REALITY
Tesla Optimus
Tesla keeps repeating its ambitious goal of building 5,000 Optimus robots by the end of 2025, with talk of mass production to follow. But independent reporting points to only a few hundred units actually built so far. Until audited numbers or customer deliveries surface, treat these claims as targets, not facts. For investors, that means high execution risk despite the hype.
So, it’s just a regular Tuesday.
Editor’s Wrapup
This week felt like the industry turned a corner: the money, the papers, the ambitions are scaling. But it’s not all sunshine. Figure’s “Project Go-Big” might define who wins the future of humanoid control — if they don’t oversell what they’ve got. AGILOped and DreamControl show real engineering progress. But the Unitree audit is proof that ignoring security is a gamble you can’t win. If you’re in this game — start demanding safer, verifiable, transparent robots before money exchanges hands.
Start demanding safer, verifiable, transparent robots before money exchanges hands.
▶ Coming Soon: WIRED Podcast review: Move Aside, Chatbots: AI Humanoids Are Here
