Questions, answered simply.

What is RoboCast?

RoboCast is a non-neural style of digital intelligence. It casts humans and machines into defined roles, like the cast and crew of a production: performers follow the script, and every decision is recorded.

What is Related Intelligence (RI)?

Related Intelligence is non-neural generative intelligence. It is related to each source by its address, to each teachable moment (each recorded decision), and to each performer’s role, and it builds results by clear-box composition. It is the kind of intelligence RoboCast performers carry.

How is RI different from AI?

On this site, AI means neural-network systems, which generate by black-box approximation. RI is non-neural and generates by clear-box composition. RoboCast can see into any RI performer and record what it did and why; it cannot see inside an opaque neural model. That is why RI takes the roles that must be audited and controlled.

So is RoboCast against AI?

No. AI is a major part of the production. Its biggest credits are co-writer and co-producer of the RI workflows that RI performers then run reliably. The formula is AI + RI + Human Workers = Safer SuperIntelligences.

What is a Dirigent?

A RoboCast-style agent. The word comes from the conductor paradigm, chosen to replace the servitude at the root of “robot.” Dirigentic design is the practice of building systems around them.

What is a RoboShow®?

A show built from addressed nodes (web pages, files, apps or other tools’ output), scheduled in sequence, branching or side by side, and played with pause, rewind and fast-forward while every step is recorded. An approved RoboShow is an RI workflow that anyone can adopt. See RoboShows →

Why the name “RoboCast”?

“Robo: Cast” is a command to cast the production team, with humans and machines in interchangeable roles. “Cast” carries many meanings beyond that: to broadcast, multicast or narrowcast, to cast a phone to a TV, to cast from a mould, even a healing cast.

What does “clear-box” mean?

Clear-box behavioral truth is a record of what each performer actually did, at every step. Every RI output carries an exact trace of the parts, rules and decisions that produced it, so the path can be replayed. It is different from black-box, white-box and glass-box approaches. RoboCast makes no claim to mechanistic interpretability of neural networks; it records the behavior of the performers it can see into.

Is RI compositional, combinatorial or deterministic?

All three, at different levels. RI is compositional by design: like a composer’s score, every part has a role, any qualified performer can play it, and you always know who played what. It is combinatorial in range: a finite cast and crew yields an almost endless repertoire. And once a workflow is approved and frozen, it runs with deterministic execution, the same sequence every time. Because sources are live, the promise is a reproducible record, not identical content.

How do AI and RI work together?

RoboCastOS, part of the governance layer, decides whether each task goes to AI, RI or both, like two hemispheres of one brain. AI drafts the scripts and handles high-touch moments with confirmation; RI performers compose from addressed parts and execute approved workflows step for step.

Can AI still hallucinate inside RoboCast?

Roles are assigned by liability. High-stakes roles go to RI performers that follow the script, or to AI interactions that confirm before acting. For simpler, low-stakes moments, an AI performer may improvise, and may occasionally get something wrong.

What is the human gate?

The paid human role at every judgment and approval point in a RoboCast structure. People run, test, refine and approve workflows before production: Human Moderated Recursive Improvement (HMRI), in place of recursive self-improvement.

What about my privacy?

RoboCast lets AI write the instructions for a task without ever seeing your personal information, while its Related Intelligence performers carry out each step, touching only the data that step names and recording exactly what they did. Because it learns from a record you can see and erase, rather than absorbing your data into an AI model, it can show you what it did with your information and forget it when you ask.

Much of today’s web runs on a surveillance economy: services watch what you do, build a profile you never see, and use it to predict and sell your attention, often feeding it into AI models where it can’t be taken back out. RoboCast keeps a record too, but it’s meant for you: it shows each step taken with your data, it never needs to hand that data to the AI writing the instructions, and it can be erased when you ask. Read more →

Does RI learn?

Yes, without a neural network. Each decision is written to a decision database, and that record revises future outputs and personalization. The Patent Hub animates the patent’s own FIG. 3, where each decision is recorded.

How does RoboCast protect my privacy?

By where it runs. RI performers run on ordinary CPUs, so an approved workflow can run on your own laptop, phone or server instead of someone else’s cloud. AI writes the script once without seeing your personal data, each step touches only the source it names, and the record of what happened stays with you to review or erase. How RoboCast protects privacy →

Is RoboCast new?

No. The architecture has been built four times since 1996 and deployed with ZDNet, Hachette Filipacchi and Microsoft, among others. It debuted as a working system at Internet World in 1998. See the Archives for the full timeline.

What patents describe it?

US 7,155,451 and four continuations, all invented by Damon C. Torres. The Patent Research Hub reproduces the issued claims verbatim, alongside the patent’s own figures, animated.

What does “public domain” mean for me?

The architecture described in the RoboCast patent portfolio, US 7,155,451 and its four continuations, is in the public domain: anyone may study it, build on it and use it. Some things are not in the public domain. RoboCast’s trademarks belong solely to RoboCast. RoboCast coined the terms Related Intelligence (RI) and Dirigents. You are free to use them, with attribution, to describe the concepts defined in this FAQ. The Generative RI abilities don’t come from the patent text alone: they come from running the RoboCast system with the right workflows, apps, and building-block files, and an active, full learning loop of digitally signed administrators and users. Those are available through the Testing Group, the Training Cohort and licensing.

Is there a white paper?

Yes. Conservation, Control, Collaboration: A Blueprint for Generative Related Intelligence sets out clear-box, compositional and combinatorial design as a blueprint for developers of all sizes building generative RI apps and Dirigents. It is available on request: request a copy.

What’s the difference between the Testing Group and the Training Cohort?

The Testing Group offers private, shared-screen sessions for decision-makers, by invitation. The Training Cohort is open to everyone: pre-register to hear when training on the standards candidate architecture becomes available. Companies can also license the curriculum.

Does it cost anything?

Testing Group sessions are by invitation at no cost. Training Cohort pre-registration is free. Curriculum and other licensing terms are being announced soon.

My organization wants to use RoboCast. Where do we start?

Start with a RoboShow: request a briefing for a pilot, partnership, investment or research collaboration.

I’m a journalist or event organizer.

The Media Kit has company descriptions, the founder bio, logos and an interview request form.