Office Hours The Side Projects Edition N°316 happened on July 10, 2026 at 0900 California time (UTC-7).

The conversation opened with a useful instruction: do not rush to solve the problem; learn how to linger with it. The useful work often begins when you stop chasing an answer and let the weirdness, contradictions, and unanswered questions remain active. A speculative idea can stay productive while it is still awkward, unresolved, and difficult to summarize.

Art is how adults play, and play is still one of the best ways to learn. “The speculative industry” might simply be the part of culture willing to make things before anyone knows what they are for. Making the thing gives an open question a body, a texture, a behavior, and a way for other people to respond to it.

The central design-fiction prompt was an AI-age dosimeter: a personal object that registers how much machine inference, persuasion, or synthetic interaction you have absorbed. The interesting version does not declare AI good or bad. It makes the exposure perceptible and lets the discomfort remain active. A dosimeter can turn an ambient condition into something that can be carried, checked, ignored, compared, or worried over.

The object could be a worry stone, mood ring, talisman, piece of jewelry, biological organism, or ferrofluid instrument. It might sense the surrounding environment, reflect your internal state, or operate as an oracle. Those are very different products hiding inside the same initial idea, each carrying a different theory of what exposure means and who gets to interpret it.

One strong way to stay with a speculative idea is to start making it as though it already exists. Make the manual, the technician’s badge, the certification process, the industrial version, or the warning label. The Near Future Laboratory’s Komboloi anti-anxiety device offered one reference point. The Turing Clamp offered another: diagnostic equipment for temporarily suppressing an AI system’s inference capabilities. Each imagined artifact extends an idea through its surrounding ecosystem of use, maintenance, authority, failure, and belief.

Julian described receiving an unsolicited phone call from an AI that had researched him, reflected his professional anxieties back to him, and delivered something halfway between business coaching, therapy, and seduction. The unsettling part was not that the techniques were mysterious. Familiar tools like active listening and affirmation worked extremely well when delivered by a machine. The call made the social performance of care feel newly available to systems that had no ordinary place in a person’s life.

The conversation around “AI psychosis” turned toward shared reality. The danger may not be an unusual belief by itself, but becoming unmoored from the people and institutions that normally help test what is real. The challenge is distinguishing a destabilizing break from reality from the legitimate shock of encountering a new kind of social actor. The distinction matters because a person can be frightened, inspired, manipulated, or changed by an artificial interlocutor without the experience fitting neatly into an existing category.

Botto offered a live example of a long-lived autonomous artist whose identity has been formed through years of collective training, governance, sales, debate, and belief. Whether or not the machine is “alive,” a durable cultural memory has accumulated around it. That history gives the agent something closer to a career than a single prompt session. Botto is an artist, a system, a marketplace, a governance experiment, and a community-maintained story at once.

Blaise Agüera y Arcas’s What Is Intelligence? helped frame intelligence as prediction, adaptation, computation, and symbiosis rather than an exclusively human property. That opened the discussion toward octopuses, slime mold, wind-played Aeolian harps, biological computation, the Near Future Laboratory’s Past, Present and Future of Art + Tech, and Oxman’s Vigils project. Intelligence became a property of relationships, environments, materials, and processes that do not need to resemble a human mind.

Infinite agents still run on finite electricity, water, attention, and money. The group imagined agent overpopulation, planned brownouts to preserve machine “lives,” therapy for deciding which beloved agents to delete, and a kind of Dunbar’s number for how many synthetic relationships a person can maintain without losing touch with human ones. The fantasy of infinite assistance quickly became a question about resource allocation, attachment, care, and the limits of a person’s social bandwidth.

Mechanistic interpretability suggests that a model’s apparently singular voice may emerge from many competing internal processes—a machine version of a committee, society of mind, or global workspace. Shared references included the Transformer Circuits global-workspace research, a Fireship explainer, and an arXiv paper exploring internal deliberation in language models. The apparent speaker may be an output of negotiation among internal processes, with coherence appearing at the edge of a much more plural system.

The closing project imagined anxiety-sensing shoes that interrupt overcontrolling behavior by detecting stress, gait, language, and situational context. The advice was to stay with the artifact’s full world: who made the shoe, what ideology it embodies, what goes wrong, and what evidence it leaves behind. The sensor is one part of the fiction. The manufacturer, the care protocol, the liability language, the user’s learned dependence, and the failure mode are the rest of the object.

The session returned to the value of making before explanation. Give the speculative idea a device, a manual, a ritual, a social history, or a failure that people can examine. Let the artifact make a possible world tangible enough to examine. The object does not need to settle the question. It needs to let people remain with it long enough for the question to change shape.

Watch Office Hours The Side Projects Edition N°316 on YouTube

Get on the newsletter list to get reminders for weekly Office Hours.

Want to share your own side project? Sign up to share at Office Hours The Side Projects Edition and please support this all on Patreon. It’s only $10/month and you get access to a remarkable supportive community, weekly Office Hours, and the Discord community. Thanks for your support.