ARC 2026.05.19.00 (Pro, Runtime And Free)

The AI Robot Software Era Has Officially Begun

This ARC update introduces something we believe is a major step forward for robotics software: an AI Agent embedded directly inside ARC.

• Not next to ARC.
• Not in another browser tab.
• Not in a generic coding tool that has never met your robot. Inside ARC.

The new AI Assistant gives robot builders the ability to chat, ask questions, generate scripts, fix code, refactor logic, and explain JavaScript or Python robot scripts using the real context of their ARC project. That means it can understand your project’s global variables, EZB ports, robot parameters, and available controlCommand() targets from your Robot Skills.

More about the AI Agent on its product page here.

In other words, it is not just writing "robot-ish" example code. It is helping write code for your actual robot project.

We are calling this a proof of concept for now because, like every good robot, it still has some growing to do. But this is the first public step toward what we believe becomes a new category of robotics software: project-aware AI robot development.

Or, said less formally: your robot now has a coding buddy. Try not to let it get smug.

New AI Chat Assistant Workspace

A new AI Chat Assistant workspace has been added between Blockly and the first desktop.

This gives ARC users a dedicated place to chat with the integrated AI Agent while working inside ARC. The assistant is designed to help with robot scripting, debugging, code explanation, and general ARC development guidance.

The AI Script Assistant can generate JavaScript or Python scripts, explain existing code, help fix broken scripts, and continue multi-turn conversations so users can refine behavior step-by-step instead of starting over every time. It is also designed around ARC-specific context, including project globals, EZB port information, robot parameters, and robot skill commands.

Why This Matters

Robotics has always had a gap between imagination and implementation.

A user knows what they want the robot to do:

• "Wave when you see my face."
• "Move the head slower."
• "Check the battery before running this behavior."
• "Fix this script because it stops after the second loop."

But turning that idea into working robot code usually requires digging through APIs, remembering port names, checking control commands, and hoping the script matches the actual robot configuration.

The ARC AI Assistant begins closing that gap.

Because the assistant is built into ARC and receives project-aware context, it can help generate code that is grounded in the robot project instead of guessing with generic examples. The result is a faster path from idea to working robot behavior.

This is important for hobbyists, educators, researchers, makers, and commercial robot developers because it reduces the friction between designing robot behavior and actually making it happen.

Also, it means fewer "why did this servo just do interpretive dance?" moments. Not zero. Just fewer.

SpeakingDepthTracker Added

• A new SpeakingDepthTracker helper module has been added for Robot Skills and speech engines.

• This module helps track when the robot is speaking so speech recognition systems can pause and resume more intelligently. The goal is to prevent the robot from listening to itself while it is talking.

• Because yes, robots interrupting themselves is funny once. After that, it is just a meeting.

Speech Recognition Improvements

• The following speech recognition systems now implement SpeakingDepthTracker:

• Bing Speech Recognition
• Windows Speech Recognition

• This improves coordination between robot speech output and speech recognition input. When the robot is speaking, recognition can be paused. When the robot is finished, recognition can resume.

• This creates cleaner voice interaction behavior and reduces false recognition caused by the robot hearing its own speech.

Navigation2DV1 Messenger Scan Point Improvements

• The Messenger system in Navigation2DV1 now deduplicates scan points by degree.

• Previously, scan points were stored in a ConcurrentQueue<ScanPoint>, which allowed repeated readings at the same degree to accumulate between location updates. This could cause unnecessary growth and duplicate scan data.

• The backing store has now been replaced with a ConcurrentDictionary<float, ScanPoint> keyed by scan degree. When a new scan arrives at an existing degree, it replaces the previous reading.

• Latest value wins. Memory growth loses. Navigation data gets cleaner.

• DequeueAllScanPoints now atomically swaps in a fresh empty map using Interlocked.Exchange, so scan points arriving during a drain operation are retained in the new map instead of being dropped.

• Clear() has also been updated for the new storage type.

Robot Skill Update Check Adjustment

• The behavior that ignored GUIDs during robot skill update checks has been disabled due to a plugin loading issue reported in the Synthiam community.

• This change helps address cases where Robot Skills could fail to load correctly when update detection did not properly account for GUID behavior. The related issue is documented here: here

Camera Device Textbox COM Port Fix

• Fixed COM port parsing in the camera device textbox.

• ARC now correctly handles COM port entries that include a baud rate using a colon separator.

Example:

COM2:115200

This format has also been verified to work as expected.

Summary

ARC 2026.05.19.00 is more than a routine update.

It introduces the first version of an embedded, project-aware AI Agent inside ARC, adds a new AI chat workspace, improves speech recognition coordination, cleans up navigation scan handling, fixes robot skill update detection behavior, and improves camera COM port parsing.

The AI Assistant is still early, but this release marks an important shift in how robot software can be built.

Instead of only programming robots by manually writing every command, ARC is moving toward a future where users can describe what they want, collaborate with an AI Agent, and generate robot behavior using the real context of their project.

That is a big deal.

And yes, the robot still needs you.... For now.