A brain in a jar.
Why an AI model alone is just a brain in a jar - and what changes when the platform around it picks the right brain for each piece of work, swaps in better ones as they arrive, and uses each one efficiently.
Sky · · 4 min read
Picture an AI model on its own. Disconnected from a business, disconnected from records, disconnected from action. Pure intelligence in isolation. It can think - but it cannot reach. It can analyze a problem if you describe one, but it cannot see the problem inside your business. A model alone is a brain in a jar. Capable. Trapped.
A model alone is a brain in a jar. Capable. Trapped.
The platform around the model is what gives the brain a body. Eyes to see the actual records of the actual business - every contact, every interaction, every prior decision. Hands to take action, not just recommend it. Memory that persists across days and weeks and quarters, so today's work builds on yesterday's. Take the brain out of the jar, give it a body, and it stops being a thought experiment. It starts being a worker.
Different brains have different strengths. Some are deep, careful, slow - built for hard reasoning, for multi-step diagnosis, for conversations that turn on subtle context. Others are fast, lean, ready - built for sorting, tagging, extracting, replying inside a heartbeat. None of them runs a business alone. The platform chooses which brain handles which piece of work - automatically, behind the scenes, without anyone touching a setting.
A well-built platform never picks one brain and sticks with it. It picks the right brain for each piece of work. Strong reasoning routes to the deep brain. Routine categorization routes to the lean brain. Time-sensitive responses route to whichever brain is fastest at the needed quality. Different work, different brain. The work decides; the platform picks; the team running the business never sees the choice, only the result.
Different work, different brain. The platform picks. The work decides.
A platform that funds its own work has to be smart about this. Use the deepest brain on every task and the work cannot pay for itself. Use the cheapest brain on every task and the work does not get done well. The intelligence is in the matching - deep brains for hard work, lean brains for routine, fast brains when speed is the variable that matters. Efficiency is what comes out of doing the work right, brain by brain, task by task. A self-funding system is one that knows the difference and chooses accordingly.
New brains keep arriving - sharper, faster, with deeper memory and longer reasoning. The platform that can swap in the better brain when it lands keeps getting smarter without rebuilding. The platform locked to one brain - or to one maker's brain - has to retrofit every time the frontier moves.
The frontier keeps moving. The platform that swaps in the better brain when it lands keeps getting smarter without rebuilding.
Swapping the brain is only half the story. The platform also has to know how to talk to the brain - and how that conversation has to grow as brains get bigger. Instructions written for last year's brain, with its smaller memory and shallower reasoning, do not let the next brain's full power reach the work. The platform's instructions have to stretch - written to describe the shape of the work, not to fit the limits of any single brain. As brains grow, the instructions grow into them.
This is the honor part. The makers building the brains are doing extraordinary work. The pace of capability is a gift the era keeps offering - but only to platforms built to receive it. Architecture decides whether the gift lands or whether it bounces off a ceiling that should have been written more generously to begin with.
For the founder, the sales lead, the team picking AI software, the questions to ask are not which brain the platform uses today. The questions are: Does the platform pick the right brain for each piece of work, automatically? Does it use the deep brain for hard work and the lean brain for routine, based on what the work needs? Will it absorb each new brain as it arrives, or will someone have to retrofit the platform every time the frontier moves?
Pick the right brain for the work. Let the platform do the picking. Let the work compound the rest.
A brain in a jar is a thought experiment. A brain in a body, choosing between deep and fast as the work demands, absorbing each new brain as it arrives, paying for the next round of work from the work that just landed - that is a workforce. The platform is the body. The body decides which brain handles each task. The work compounds.