Benji Barash

🔍 Today we're launching root cause analysis in @roboto_ai Roboto Agents now debug across your robot data and your codebase, tracing a failure from your logs to the exact code that caused it - and even drafting a fix. A triage label tells you what failed. Root cause analysis figures out why. That's always been the hard part. Knowing a run hit a localization failure is just the starting point. The explanation is buried somewhere across the telemetry, source code, configuration, and engineering context. That's why debugging often takes days and pulls in multiple people. An agent that only reads your robot logs can tell you what happened. To explain why it happened, it needs the same context your engineers do. Now it has both. Roboto Agents follow the evidence from your robot data into your codebase, test hypotheses against real runs, and come back with the most likely cause, the evidence behind it, and even a proposed fix to review. Instead of spending days bouncing between logs, plots, source code, and teammates, you can get straight to understanding what changed.

Ask a question about a robot run, and watch the agent build the answer and pull up the data for you. That's live in @roboto_ai today. The agent can drive the visualizer now, too. Debugging a robot log in a visualizer used to mean driving it yourself. Hunt for the right topics, plot them one at a time, scrub the video, then start over on the next analysis. Now you can just ask: "Look at the IMU and the front camera and show me any unusual movements." The agent analyzes both the time-series and video, then takes you straight to the moment that matters. The best part is that it's still the same visualizer underneath. You're not locked into a chat response or a static report. Ask a follow-up. Add another topic. Keep exploring. When interacting with robot data becomes this easy, engineers naturally ask more questions. You check the hunch you would've ignored. You compare runs you wouldn't have had time to compare before. And it's not limited to a single run. Point the same conversation at your robot's entire history and ask whether you're looking at one-off behavior or a fleet-wide pattern. This morning we launched Agentic Triage - the first pass over every robot run. This is what happens when you want to go deeper. Part 2 of Day 2 at Agent Week is live now. @OpenRoboticsOrg @PX4Autopilot #Robotics #PhysicalAI #Agents #Analysis

🚀 Today we're launching agentic triage in @roboto_ai. Roboto Agents can now review every robot log the moment it arrives and open labeled issues in #Jira, #Linear or #GitLab with supporting evidence attached. It’s the first line of defense that every #robotics team needs. Failure modes and anomalies are detected automatically. The runs that warrant escalation get surfaced. You can spend more time improving your system instead of debugging the same issue for the fifth time. Robotics teams often underestimate how high the support burden will be once they start putting robots into the field. A robot that's 99% reliable sounds great until you deploy 100 of them. Suddenly you're dealing with failures every day. The statistics catch up quickly. Tomorrow’s issues are already sitting in today’s logs - but nobody has time to review them. So you wait for a customer to complain instead. That’s a huge business risk. Agentic triage gives you the tools to find issues before your customers do. We’ve been testing this with select engineering and field support teams collecting #ROS bags, #MCAP and #PX4 ULG files over the past few months. They think it rocks, and they’ve discovered and fixed some surprising issues before anyone noticed. Today is day 2 of #AgentWeek, and it's live now. If you want to try agentic triage on your own data, reach out or check out our blog post below. #PhysicalAI #Triage #Agents @OpenRoboticsOrg @PX4Autopilot @NVIDIARobotics

Today we're launching Roboto Agents - AI that turns every robot run into a faster learning loop. Since late last year, our team has been working on a new way for robotics teams to analyze the massive amounts of data their robots generate. Every robotics team hits the same wall: you generate far more data than you can analyze. Most of it goes ignored, and the lessons buried in those runs never make it back into the robot. That's the problem we built Roboto Agents to solve. Instead of analyzing runs one at a time in a visualizer, you can just describe what you want to accomplish: "Review these logs for any anomalous behavior." "Investigate why this vehicle stopped during takeoff." "Which software and hardware versions have this issue across the fleet?" "Highlight all grasping failures and put them in a collection." Agents do the work and show you the evidence behind their findings. The faster you learn from every run, the faster your robots get reliable. That's the learning loop that separates the teams that successfully scale Physical AI. The best part: this isn't just a vision for the future. Robotics teams have been using Roboto Agents in production for months to investigate issues, find patterns, and build better datasets. A huge thank you to the teams who trusted us early, and to everyone at Roboto who made this launch possible. This week we'll share how Agents accelerate triage, root cause analysis, and data curation. #Robotics #PhysicalAI #Data #Agents

An important lesson from my time at Amazon Robotics: a single log tells you what happened. But the real value comes from the corpus of historical logs you build over time. We had years of robotics data. But when a robot failed, we still struggled to answer the question every engineer asks: have we seen this before? This is why I'm so excited about AI agents in robotics. Not because they replace infrastructure, but because they finally unlock it. An agent can take the anomaly you're investigating today and get you to answers: Have we seen this before? What changed beforehand? Which robots are affected? But an agent is only as good as the tools and infrastructure available to it. The better your robotics data is organized, indexed, and made searchable, the more effective the agent becomes. Agents don't replace the foundation - they leverage it. Agent Week at @roboto_ai is almost here 👀 #Robotics #Data #Agents #PhysicalAI