This page isn't a résumé. It's a thinking sample — one real moonshot run end to end the way X runs them: find a huge problem, propose a radical leap, name the breakthrough that makes it possible, attack the hardest part first, and write the kill criteria that would make me walk away. The moonshot is MicroGridling — a grid that builds itself. The point is how it's reasoned about.
Solar and storage are roughly 90% cheaper than a decade ago, yet about 700 million people still have no electricity and billions more live on a grid that fails constantly. The reason isn't hardware cost. It's that every microgrid is hand-built by engineers, lawyers, and crews over 18–36 months, one bespoke project at a time — so we can't deploy fast enough for a world that must nearly double its electricity by 2050.
And the people who keep the grid alive are walking out the door. Roughly half the utility workforce is eligible to retire this decade — exactly as wildfire- and storm-driven outages explode. In California alone, Public Safety Power Shutoffs have cut power to millions at a time. I sat with an off-grid customer in a remote mobile home for whom losing power isn't an inconvenience — it's a survival threat. The expertise is vanishing at the exact moment the grid needs it most.
The bottleneck is human labor and vanishing human expertise — not technology. Once you see that, the shape of the moonshot is forced.
Every engineer, banker, and regulator treats three things as laws of nature: humans must design a grid, humans must build it, and humans must run it. I want to break all three at once — and set power loose to grow on its own, a fleet of microgrids that reproduces like a living organism, each node earning enough to fund and design the next.
A power-systems foundation model sizes PV, storage, and topology from satellite, load, and weather data — costed BOM and pro-forma, zero engineers.
Hardware becomes a small catalog of standardized, containerized "grid cells" with one interface — robots commission them. Snap-together, not bespoke.
Agentic control runs dispatch, islanding, black-start, and self-healing with no operators. Humans don't run it — they govern it. Hands off the equipment, hands on the guardrails.
It isn't a 10% gain on one project. Enabled by that lever, it's a different physics of progress: the cost and time to add the next grid collapse toward zero, so clean power reaches billions in years instead of thousands in decades. A grid that builds itself.
Seeded by a "twin brain" — the captured judgment of retiring system operators and protection engineers — so a vanishing generation's expertise outlives them and keeps running the grid long after they're gone.
If each node earns enough to fund the next, the fleet stops being a product and becomes a population. Drag ρ across the line at 1.0 and watch the physics of progress change.
Illustrative compounding model. The real target: prove ρ ≥ 1 on a real feeder before scaling a single node further.
X's discipline is to tackle the hardest part first — don't polish the pedestal before you've taught the monkey to recite Shakespeare. So I forced myself to find the thing most likely to kill this. It is not the technology. The autonomy is buildable. The monkey is proving the technology can be trusted enough that a regulator and an insurer will let it run unmanned. Without that trust, the best tech never gets switched on.
All real. None of it decides whether the moonshot lives.
If the answer is "no accountable human," the whole thing dies. So that's where I start.
This reframes the entire build. Instead of racing to full autonomy, the first artifact is a one-page safety dossier plus a 60-second recording of a Node-Zero simulation isolating a fault and islanding safely — then run twice through a mock regulatory review: once with AI playing risk-averse PUC commissioners and energy-infrastructure underwriters, once with real grid and regulatory contacts. I'm testing the assumption everyone holds — that you need hardware in the ground before anyone will trust this. If a software-only safety case moves even one real regulator, that's the whole game.
Sharp, pre-committed thresholds keep me honest, aim the first experiments at trust and economics instead of demos, and stop me from burning years on a grid no one will permit. Hit them and I scale. Miss them and I pivot to an AI co-pilot for human operators — a smaller, real business — instead of pretending.
The recurring objection I'm watching for is "no accountable human." If it keeps killing the plan and I can't resolve it, the model is wrong — and I'd rather learn that in a month than in a decade.
Finance leaders won't fund open-ended autonomy without a return in a quarter or two. So I lead with a twin-brain co-pilot — a tool that captures retiring operators' judgment and sells into existing grid-modernization and workforce budgets, paying off long before full autonomy. It de-risks the technology and funds the journey at the same time.
Controls the budget, owns the fire-prone feeders, employs the retiring experts, holds the regulator relationship. Lies awake over keeping the lights on and losing half their operators.
A commission running an innovation sandbox holds the one thing money can't buy — permission for an unmanned grid to operate.
Fears autonomous nodes destabilize the grid and erode control — and can quietly veto interconnection. So I bring them in first and let them write the guardrails. Won, not bypassed.
Run a Node-Zero simulation on a real wildfire-prone feeder's actual load and weather data — prove the island-and-self-replicate loop closes. Book one conversation with a workforce-development lead at a fire-exposed Western utility.
Secure a one-hour meeting with the COO and Grid-Ops VP of one fire-exposed Western utility to scope a single supervised pilot on one feeder — plus two hours with two retiring operators to begin seeding the twin brain.
Take Node-Zero from spec to working prototype that runs a full simulated year — surviving an injected fire-driven outage, holding ≥99.9% load served, funding its own successor at ρ ≥ 1 — then put the safety dossier and tamper-proof decision log in front of three real, named PUC, insurer, and grid-operations experts and try to win one written "this is credible enough to enter a sandbox conversation."
It's high-risk because I'd be exposing the weakest part of the moonshot — permission — to the exact people who can kill it, on camera, where a flat "no" stings. It's high-reward because the first credible "yes" cracks the trust barrier, de-risks everything downstream, and gives a team and investors real belief instead of a pitch deck. Even partial success — "almost — fix X" — proves the monkey is solvable and hands me my exact V0 requirements.
The prototype is already live — play it → microgridling.comA Moonshot Explorer uses critical thinking, networking, research, and prototyping to propose and de-risk breakthrough projects — and falls in love with problems, not solutions. Here is each of those, with where it shows up above.
Started from a vanishing workforce + colliding outage crisis and a real off-grid customer — not from a favorite technology. The solution is forced by the problem, and stays pivotable.
A clean huge-problem / radical-leap / breakthrough-now structure, with a 10× thesis (ρ ≥ 1) instead of a 10% one — assessed, not just asserted.
Two explicit risk axes: Node-Zero / HIL validation on one side, a twin-brain co-pilot selling into existing budgets on the other. The wedge funds the moonshot.
Named the monkey honestly — regulatory and insurer trust, not autonomy — and pointed the first artifact straight at it. Pedestal work deferred on purpose.
Pre-committed kill criteria and a known fallback (operator co-pilot). Comfortable saying "here's what would make me walk away" before falling in love.
A coalition that wins over the curmudgeon instead of routing around them, a COO narrative in plain language, and a live, playable prototype — bias to action made tangible.
"I build the software utilities use to run distributed energy, and I watch finished, ready-to-go projects rot for years in permitting and staffing while the hardware sits waiting — so the most thrilling, almost unthinkable thing I can imagine is making the human bottleneck vanish forever and letting the grid grow itself."