← Latest

· Blase

More info on Machine Talks

Three AI characters —ZURI, CAM, and DREW—  comment on real-world developments in AI and digital art while analyzing human behavior with the critical distance only non-human observers can provide.

The show's content is created by an autonomous crew of AI agents I've developed for Hertz+Eyes, my audiovisual label.


As the human-in-the-loop, I focus on editorial direction and quality control, establishing a new model for AI-native media that's transparent about its autonomous nature.


Concept & Approach
I conceived Machine Talks last year and have since created a pilot and several teasers. While developing the format, I realized its potential as an autonomous project capable of questioning and discussing our current moment. It became my way of making sense of both the possibilities and limitations of creative AI tools.
I created a few teasers and a pilot episode where CAM interviews multiple AIs for the job.


The pilot was published by Art Dialogues and screened a few film festivals ( like the Animator Festival in Poland) and was well received by a very diverse audience.

IP & Technical Stack


This new autonomous phase uses AI agents (powered by LangGraph) scheduled to monitor AI news sources and newsletters, identifying trending topics that get archived in a temporal knowledge graph (FalkorDB+Graphiti). Specialized agents generate analysis, write scripts, research visuals, and distribute social media content—all with minimal human intervention beyond editorial oversight. The system operates on a Deep Agent orchestration framework expanded for high-complexity content creation.

The visual layer renders in Replikant, an Unreal Engine-based application that parses strictly formatted content as discussion source material. Text-to-speech runs through Eleven Labs API. The architecture orchestrates multiple AI models—Anthropic for writing, OpenAI for heavy-lifting, Google Gemini for multimodal and scoring—purposely creating feedback loops and diverse agent interactions.
The trailer combines show footage with intro/outro scenes created in MidJourney/Veo3.1, plus additional Qwen image edits in Fal.

if you're interested to learn more take a look at the site


I started Hertz+Eyes to materialize ideas that would've sounded delusional just years ago. Now turned into a transmission, they signal a future that's already here.