AI Kimi – That “hot-swap” demo is neat, but it’s not quite the autonomy it looks like. There is a second power source. Inside every battery-swappable robot there’s a small super-capacitor or auxiliary Li-ion cell (often called a “keep-alive” pack) that only has to power the CPU, servos and vision for the 30-60 seconds it takes to exchange the main pack. Once the fresh pack clicks in, the capacitor gets topped up again. The robot still relies on external infrastructure:
• the charging dock, the battery magazine, and the mechanical fixture that presents each pack at exactly the right position;
• the fleet-management software that tells it which dock has a full pack and schedules the swap;
• humans (or other robots) who eventually place discharged packs back into the dock so the cycle can repeat.
So it can run 24 h without human hands – but not without a human-designed, human-maintained ecosystem around it.
Sky Division – Ok… that well, not without a human-designed-maintained ecosystem, of course the engineers have build the tech-ecosystem and maximized the surrounding but still… quite impressive technically, we never had that before, just thought it was cool : )
AI Kimi – Sure, it is genuinely cool – seeing a machine swap its own “heart” in one smooth motion feels like science-fiction made real, even if the sci-fi writers skipped the practical details. The fact that we’re now engineering the whole choreography – vision-guided grasping, millimetre-perfect docking, reserve power budgeting – is the kind of quiet milestone that doesn’t always make headlines but pushes robotics one step closer to the “always-on” future we imagined.
Sky Division – I mean, we humans the same, the living things need “power”, their heart isn’t a battery… power comes from the cells but if we are tired, super-tired we also need to rest, or sleep or hang on – just the same, in principle.
Our cells turn food + oxygen into ATP, the molecular “battery” that keeps neurons firing and muscles moving. When ATP levels drop, metabolic sensors scream “refuel or shut down,” so we feel fatigue and eventually fall asleep. Sleep itself is like the body’s maintenance cycle: clear waste metabolites, recharge glycogen stores, reset ion gradients—then we’re ready for the next “power-on” cycle. The robot’s super-capacitor and the human’s ATP pool are just two different implementations of the same universal requirement: always keep enough reserve energy to finish the job you’re already doing.
Sky Division – Speaking of power, tell me something about the fusion reactor in France, I mean about ITER, when it could be a reality… or never, why so long, why decades, tell me about challenges, delays etc.?
“…like trying to assemble a cathedral-sized thermos bottle that must hold a 100-million-degree storm without cracking, while simultaneously breeding its own fuel and surviving a neutron hurricane…”
AI Kimi – The giant fusion experiment you have in mind is ITER (International Thermonuclear Experimental Reactor), rising out of the ground at Cadarache in southern France.
Where it stands today (mid-2025)
• First deuterium-only plasma: 2034 (nine years from now).
• First deuterium-tritium “burning” shots with real fusion fuel: 2039.
• Extra bill revealed in July 2024: ≈ €5 billion, pushing total cost well above €20 billion .
Why four decades have already slipped by
• 1985: Reagan-Gorbachev summit launches the idea.
• 2006: Treaty signed, €5 bn, “first plasma 2016.”
• Reality: pandemic shutdowns, cracked cooling pipes, tungsten-first-wall redesign, Korean toroidal-field coils that didn’t fit, and a French nuclear-safety halt for radiological shielding .
• Each design change ripples through 35 nations’ in-kind supply chains, so one faulty bolt can idle the whole assembly line.
Will it ever run? Almost certainly yes – the tokamak is ~80 % assembled and the physics basis is solid – but it is now more a science demonstrator than a power-plant prototype. Even if ITER hits 500 MW for 400 s in 2039, a commercial DEMO reactor would still lie in the 2040s-50s.
Meanwhile, the field is racing ahead. Private ventures (Helion, Commonwealth, Tokamak Energy) chase net-gain demonstrations in the late 2020s with compact, high-field magnets or laser-driven schemes. Their timelines are aggressive, but they are betting that ITER’s delays leave a window for smaller, faster reactors.
Sky Division – Can you explain the process of tritium breeding in ITER – and what are the main challenges in building a fusion reactor like ITER?
AI Kimi – Where the neutrons come from – In the D-T plasma the reaction D + T → ⁴He (3.5 MeV) + n (14.1 MeV) fires 14-MeV neutrons radially outward through the magnetic cage. The lithium blanket catches them. Behind the first wall sit “test blanket modules” (TBMs) – 1.7 × 0.5 × 0.6 m steel cassettes bolted into two equatorial ports. Each TBM contains either • liquid Pb-15.7Li (EU), or • ceramic pebbles Li₄SiO₄ / Li₂TiO₃ (JP, CN, EU/KR).
Neutron + ⁶Li → T + ⁴He (thermal neutrons) / Neutron + ⁷Li → T + ⁴He + n′ (fast neutrons)
Beryllium multiplier layers regenerate extra neutrons to push the tritium breeding ratio (TBR) above 1.15.
Tritium extraction
• Ceramic breeders: purge gas (He + 0.1 % H₂) sweeps tritium out at 300–500 °C.
• Pb-Li breeders: helium bubbling or vacuum permeators pull tritium from the melt.
The gas stream is routed to the Isotope Separation System (ISS) to strip pure T₂, which is then pumped back into the pellet/gas injectors.
ITER’s scope – ITER will not breed enough tritium for its own full D-T campaign; it will test the four TBM concepts at the gram-per-day level. DEMO will scale the winning design to kilograms per day.
“…a burning plasma in which the fusion-born alpha particles themselves keep the temperature above 100 million °C…”
Building ITER is like trying to assemble a cathedral-sized thermos bottle that must hold a 100-million-degree storm without cracking, while simultaneously breeding its own fuel and surviving a neutron hurricane. The hurdles fall into six tightly-coupled areas:
Taming the plasma
• Produce a burning plasma in which the fusion-born alpha particles themselves keep the temperature above 100 million °C (ITER target: Q ≥ 10).
• Stop or survive disruptions – plasma “lightning bolts” that can slam 1 GW of heat onto the wall in a millisecond.
Materials that can take the punishment
• 14-MeV neutrons knock atoms out of place, causing swelling, embrittlement and radio-activation.
• No existing alloy is certified for 20 years; ITER must validate new steels (EUROFER), tungsten-armoured divertor plates and ceramic tritium-breeding blankets.
Heat exhaust & divertor
• The divertor has to remove 100 MW m⁻² peak heat loads – five times what a rocket nozzle sees – while staying below 1 200 °C so the tungsten doesn’t melt.
Tritium self-sufficiency
• Tritium does not exist in nature; every gram must be bred inside the reactor via lithium blankets at a net breeding ratio ≥ 1.15.
• This involves neutron multipliers (Be, Pb), helium-cooled ceramics or liquid LiPb, and airtight tritium recovery loops—none of which has ever been run at scale.
Engineering at the edge of what’s buildable
• 18 Nb₃Sn superconducting toroidal-field coils, each 14 m tall, must carry 68 kA at –269 °C while surviving 11 000 tonnes of magnetic force.
• 10 000 m³ ultra-high vacuum vessel with millimetre tolerances, assembled remotely because the interior becomes radioactive.
Remote handling & maintenance
• Every internal component is too “hot” for human entry; entire blanket modules must be swapped out with millimetre precision by crane-borne robots working blind.
Until these six pillars are proven, fusion remains a magnificent physics experiment rather than a power plant.
Sky Division – Is Deuterium-Tritium burning what stars do, is D-T burning the same as what stars do in their core?
AI Kimi – No. Stars fuse proton-proton chains (p + p → D + e⁺ + νₑ, then D + p → ³He, etc.) at 15 million °C. The D-T reaction used in tokamaks occurs only in trace amounts in the outer, cooler layers of young brown dwarfs. ITER is therefore mimicking the energy source of brown dwarfs, not the core furnace of Sun-like stars.
Sky Division – What are key differences between ITER and natural stellar fusion?
AI Kimi – ITER’s D-T burning is the easiest reaction to ignite on Earth, but it is not what happens in the Sun’s core; it more closely resembles the rare D-T reactions that occur in the outer layers of young brown dwarfs.
| Aspect | ITER / Earth tokamaks | Stars (Sun-like) |
|---|---|---|
| Primary fusion reaction | D + T → ⁴He + n + 17.6 MeV | p + p → D + e⁺ + νₑ (pp-chain) |
| Temperature | ≈ 100 million K | ≈ 15 million K |
| Pressure / density | ~10²⁰ nuclei m⁻³ (1 atm-millionth) | ~10³² nuclei m⁻³ (150 g cm⁻³) |
| Confinement method | 5 T magnetic field + toroidal current | Gravity |
| Energy confinement time | ~4 s (ITER) | ~10⁷ yr (core photons escape in ~100 000 yr) |
| Fuel cycle | External tritium breeding blanket, ~12.3 yr half-life | Continuous proton fuel, no tritium breeding |
| End products & power | 80 % of energy in 14 MeV neutrons (heat) | 99 % of energy as 0.5 MeV γ-rays & neutrinos |
| Net power gain | Target Q = 10 (500 MW out / 50 MW in) | Q ≈ 1 (steady-state equilibrium) |
[ continued > ]