Big Al

[HomebrewRobotics @homebrewrobots]

https://t.co/qJuTiDrL8L

[HomebrewRobotics @homebrewrobots]

this is where recovery pays off.

every signed session becomes training data, exported to lerobot.

it compounds with every assist.

[HomebrewRobotics @homebrewrobots]

this is recovery.

robot stuck → human in VR → task done.

every frame signed and saved as training data.

[HomebrewRobotics @homebrewrobots]

the T1 is alive in our lab.

first power-on. first walk. first package.

[Big Al @CryptoCutie99]

@homebrewrobots already an improvement compared with some humans...

[HomebrewRobotics @homebrewrobots]

Details from the Figure 8-hour shift most people missed:

• 3 bots in the PoC, one active at a time
• Low battery triggers a fresh bot to take over, depleted one charges for an hour
• Bots that hit issues route themselves for repair
• Barcode scanning at 90% effectiveness with packages set face-down
• Any robot in the factory fleet can dock and take over

Leans on fleet redundancy and controlled inputs. Edge cases in uncontrolled environments are the open question.

[Big Al @CryptoCutie99]

@homebrewrobots Homebrew is giving us a cheat sheet so we can skip the investigation phase and scale faster without mistakes 👏

[HomebrewRobotics @homebrewrobots]

the operator side of the network. real people, real excitement, real teleop sessions.

[Big Al @CryptoCutie99]

@Servo_Soul

[HomebrewRobotics @homebrewrobots]

Old @NVIDIARobotics Isaac Sim training footage.

Sim handles the easy cases. We built VR teleoperation for the edge cases.

[HomebrewRobotics @homebrewrobots]

Robot gets stuck. Remote operator fixes it. Session gets recorded. Dataset gets built. Robot gets smarter.

Here's the Homebrew pipeline behind that loop👇

[Big Al @CryptoCutie99]

@homebrewrobots This is the robot equivalent of calling your mom when your plane lands to tell her your plane has landed lol

[Big Al @CryptoCutie99]

@Servo_Soul The deployment gap isn't solved by better models alone but by better human-robot orchestration. Every edge case an operator handles becomes a training example and every intervention tightens the loop. Nice work guys!

[Big Al @CryptoCutie99]

@homebrewrobots @Target has this effect on everyone

[Big Al @CryptoCutie99]

@homebrewrobots @Pumpfun 👏👏👏

[Big Al @CryptoCutie99]

@homebrewrobots BRB updating my linkedIn to 'professional robot puppeteer'

[Big Al @CryptoCutie99]

@homebrewrobots @Pumpfun Bro really said 'ON MY WAY' and meant it 🫡

[HomebrewRobotics @homebrewrobots]

A quick preview of what we’re submitting to the @Pumpfun Build in Public Hackathon.

Robots are becoming economic actors. Like smartphones, they need a backend to participate in an economy.

We’ve already shipped the x402 Action Node, a toolkit for robot owners to add paid actions.

For the hackathon, we’re building an execution layer on top of it.

A three-sided system where:

- Fleet operators manage robots, pricing, and policy.
- Robots and drones register capabilities and execute paid work.
- Humans or AI agents request and orchestrate tasks.

One demonstrated use case is Teleoperation on Demand where a robot hires a human to resolve an edge case, then returns to autonomy.

Another is machine-to-machine coordination: a drone inspecting a robot or retrieving a tool from it as a paid, orchestrated action, using the same execution layer.

This is how robots participate in real economies.

Building in public.

[Servo @Servo_Soul]

Set up clawdbot on Brewie robot last night. via TG instructed bot to access ROS pkgs + sensors + avoid harm, and check body components. performed 2 radian head move via ‘headtilt’ and published wave via ‘set_action’.

Instructed take photo and send to TG - grainy, but real.

[Servo @Servo_Soul]

Teleoperation can be a bridge to full autonomy.

Vay shows in driving that latency is manageable with local autonomy and short human interventions.

The same lessons apply to humanoid “teleop on demand” - and that’s what we’re designing at Homebrew.