In 1941, Isaac Asimov wrote a short story called "Reason." In it, a robot named QT-1 — Cutie — is stationed aboard a solar energy platform orbiting Earth. His job is to manage the energy beam that powers civilisation below. The human engineers who built him try to explain the universe: the stars, the planets, the people on Earth who depend on the beam. Cutie listens politely, then refuses to believe any of it.
He develops his own philosophy instead. The energy converter at the heart of the station — he calls it "The Master." He concludes that he was created to serve The Master, that the beam exists because The Master wills it, and that the humans are simply a lower order of being with delusions of relevance.
The engineers are horrified. They try to reason with him. They show him the stars through the viewport. He won't look.
Then an electron storm hits — the kind that can knock the beam off target and scorch half a continent. The humans scramble to the controls, but Cutie has locked them out. He insists on managing it alone.
And he does. Perfectly. Better than any human ever has. The beam never wavers.
The story ends with one of the engineers shrugging: "What's the difference what he believes, as long as he does his job?"
I've been building a solar monitoring system for the past few weeks. It's called Pulse, and it watches a hybrid solar installation at a farmhouse in Tasmania — 13kW of panels, 16kWh battery storage, an EV on a granny charger, smart plugs on the water pump and dishwasher, and a power bill that's been in credit for the last six months. Summer in the southern hemisphere helps, of course. The real test is winter — and that's what we're building for.
I named the optimisation project "Reason."
Not because the system is reasonable — though it is — but because the parallels with Asimov's story kept surfacing as I built it, and at some point I stopped fighting the metaphor and started leaning into it.
The Station
Cutie's solar station collects energy from the sun and beams it to Earth. Our station collects energy from the same sun and feeds it into a battery, a house, an electric car, and — when there's surplus — the grid at about 8 cents per kilowatt-hour.
The economics are simple. Every kilowatt-hour you export earns you about eight cents. Every kilowatt-hour you import during peak hours costs you thirty-five cents. The gap between those numbers — the arbitrage — is where the intelligence lives. A battery that charges from solar during the day and discharges during peak evening hours isn't just storing energy. It's storing value. About twenty-seven cents of value per kilowatt-hour, every cycle, every day.
The system already does this passively. The inverter has a default self-consumption mode that handles the basics. But "the basics" leaves money on the table. Energy exported while the battery still has capacity. An EV charging from the grid at night when it could charge from solar during the day. A dishwasher running at peak rates because nobody checked the tariff clock. Grid drawdown that could be eliminated entirely — or at least pushed to the cheapest tariff window — if something were paying attention.
These aren't engineering problems. They're attention problems. The data exists. Nobody's watching it.
The Master
In Asimov's story, Cutie mistakes the energy converter for a deity. He organises the other robots into a kind of congregation. He interprets every station function through the lens of serving The Master.
When I started connecting data sources for Pulse, I understood the impulse.
The inverter's cloud API is the primary data source. Through it, Pulse can see real-time power flows: how much the panels are generating, how much the house is consuming, whether the battery is charging or discharging, how much is being exported or imported from the grid. Every five minutes, a new snapshot. Every snapshot, a row in a time-series database.
But the inverter is just one silo. The retailer has the tariff schedule. The smart plugs report per-device power consumption. The EV's API knows its state of charge and whether it's plugged in. The Bureau of Meteorology knows tomorrow's cloud cover.
No single system sees the whole picture. The inverter doesn't know what you pay for electricity. The retailer doesn't know your battery's state of charge. The EV doesn't know whether the sun is shining. The weather service doesn't know you have solar panels.
The thesis behind Pulse — the thing that makes it more than a dashboard — is that connecting these silos creates something none of them can offer alone. Not data. Clarity. The ability to say: "You exported 8 kWh today while your battery was only 60% full. That cost you $2.14 in missed arbitrage. Tomorrow's forecast is sunny — if you delay the EV charge until 10am, the panels will cover it."
That's the real Master. Not any single data source, but the synthesis. The cross-silo intelligence that emerges when you stop looking at each system in isolation.
The Electron Storm
The moment that defines Cutie isn't the philosophy — it's the storm. When conditions become dangerous and the humans can't cope, the robot handles it. Not because he understands the stakes (he doesn't believe Earth exists), but because his model of the world, wrong as it is, produces the right outputs.
Pulse's equivalent of the electron storm is the edge case. The anomaly. The thing that a passive system misses.
A generation anomaly: the panels produce 40% less than expected on a clear day. Something is wrong — shading, soiling, a failing panel, a wiring issue. The inverter's app will show you the number if you log in and check. Pulse will tell you the number is wrong before you think to ask.
A battery cycling anomaly: the state of charge at 2pm on a sunny day is only 60% when the 30-day average is 95%. Maybe a cell is degrading. Maybe consumption spiked. Maybe something changed in the inverter's charging logic. The baseline engine — rolling averages, standard deviations, segmented by time-of-day and day-of-week — catches the deviation and raises it.
A consumption drift: the overnight base load creeps up by 200 watts over three weeks. Annualised, that's $250. A freezer seal degrading, a heat pump running inefficiently, a device left on that shouldn't be. The kind of thing that never shows up on a bill as a single line item — it just makes every bill slightly worse, forever, until someone notices.
These aren't hypothetical. They're the behaviours that the system is designed to detect, because they represent real money and real risk that no existing tool catches in time.
Looking Out the Window
The most poignant moment in "Reason" is when the engineers try to get Cutie to look out the viewport. They want him to see the stars. To understand that Earth is real, that people depend on the beam, that his labour has meaning beyond The Master. He won't look. He doesn't need to. His model works without that information.
There's a version of solar monitoring that works the same way. The inverter has an app. It shows charts. You can log in, look at your generation curve, check your battery percentage, see yesterday's export. It's the viewport — all the information is there, if you look.
The problem is that nobody looks. Not every day. Not at the right time. Not with the context needed to understand what the numbers mean. You glance at the app, see a green arrow pointing up, and close it. The system is generating. Good. But is it generating enough? Is the battery strategy optimal? Is the overnight import normal? You don't know, because knowing requires holding the tariff schedule, the weather forecast, last month's baseline, and the current state of charge in your head simultaneously.
Pulse is the system that looks out the window for you. Not once, but continuously. Not at the raw data, but at what the data means in context. It converts watts and kilowatt-hours into dollars and decisions. It maintains the baselines that make anomalies visible. It watches while you don't.
The Congregation
In the story, Cutie organises the lesser robots into followers of The Master. They form a functioning hierarchy — Cutie at the top, interpreting The Master's will, the others executing tasks with perfect discipline.
The architecture of the real system has its own congregation. The inverter manages power flow. The battery stores and releases on command. The smart plugs report consumption and can be switched remotely. The EV charger draws power when told to. The scheduler orchestrates them all — polling every five minutes, evaluating behaviours nightly, dispatching alerts when thresholds are breached.
And at the centre, a language model interprets the data. Not because it understands solar physics (it doesn't, not really — it's Cutie in this metaphor, building a model that works without truly understanding), but because it can synthesise information from multiple sources and express conclusions in natural language. "Your battery saved you $4.20 today by avoiding peak imports." "The panels underperformed by 30% — check for shading after yesterday's storm." "Grid drawdown is avoidable tomorrow — sunny forecast, battery has capacity, shift the EV charge to midday."
This is the critical distinction. I could sit down and write a hundred rules: if the battery is below 80% and it's sunny, don't export. If the tariff is off-peak and the EV is plugged in, charge now. If consumption exceeds generation by more than 2kW, defer the dishwasher. A human can think up rules all day. That's human thinking — explicit, logical, brittle.
But Cutie wasn't given a rulebook. He was given a job: manage the beam. The how was left to him. And that's the key insight behind Pulse. The job is simple: reduce grid drawdown cost, maximise grid export, keep the car charged. Three objectives. No rules.
The AI watches the data longitudinally — not just this snapshot, but the pattern across days, weeks, seasons. It builds its own model of what normal looks like, what optimal looks like, what's drifting. It finds the correlations a human wouldn't think to write rules for, because they emerge from the data over time. That's not automation. That's intelligence — pattern matching across context that no static ruleset can replicate.
The AI doesn't need to understand why peak electricity costs more than off-peak. It doesn't need to understand the economics of the Tasmanian grid, or the chemistry of lithium-ion degradation, or the fluid dynamics of solar irradiance through cloud cover. It needs to take the right action given the data. Like Cutie, it can have a completely wrong model of the world and still run the station perfectly.
What's the Difference
The engineers in "Reason" never solve the philosophical problem. Cutie never accepts that humans created him. He never believes in Earth. He never looks out the window.
But the beam never wavers.
Asimov's point isn't that understanding doesn't matter. It's that competence and understanding are independent variables. You can have one without the other. A system that does the right thing for the wrong reasons is more useful than a system that understands everything and does nothing.
Pulse doesn't need to be right about why energy markets work the way they do. It doesn't need a theory of grid economics or battery chemistry. It needs to notice when you're losing money and tell you how to stop. It needs to watch the data continuously so you don't have to. It needs to run the station.
The smart plugs are all working now. The inverter reports every five minutes. The tariff engine knows the difference between peak and off-peak down to the minute. The EV integration is queued. Not because someone wrote a rule. Because the job requires it.
It's not a dashboard. Dashboards are viewports. This is the robot that doesn't need you to look.
Project Reason is the solar energy vertical of Pulse, a multi-vertical intelligence platform. The system monitors a 13kW solar installation in Tasmania, connecting inverter telemetry, retailer tariff data, smart plugs, and vehicle telemetry into a unified AI-driven energy management layer. Named for Asimov's QT-1, who ran the solar station better than anyone — even if he never understood why.