Short answer: yes. Long answer — below.

---

Introduction

The Moon has no atmosphere to slow down a landing vehicle. No airport to receive it. Every kilogram landing on its surface has to brake using engines alone — burning fuel that itself had to be flown there.

It’s a closed loop. Expensive, single-use, energy-intensive.

And yet the Moon has something we rarely think about in the context of landing: it has craters. Curved, kilometre-wide walls carved over billions of years by meteorite impacts. It has regolith — material that can be sintered into plates. It has gravity weak enough that a vehicle entering at a shallow angle onto a curved surface won’t be crushed, but gently guided onto a horizontal trajectory.

This article isn’t a finished engineering project. It describes something else: how, in a conversation with a language model, step by step, from an “what if…” idea — a coherent technical concept emerges, one with numbers, geometry and economic sense. A concept that without AI would have been an interesting thought for a moment, and then would have disappeared.

That’s exactly what human–AI collaboration is about.

---

Where the idea came from

I didn’t start with a landing pad. I started with a simple question about fuel — how much does a vehicle descending to the Moon from 500 metres burn. Then from a kilometre. Then for a 1-tonne payload, then for 2.5 tonnes.

This is just a regular conversation with the LLM Claude. Prompt, response, and so on in circles.

Somewhere in the middle of those calculations another question appeared: what if part of that energy were taken not by the engine, but by the surface?

Not as an abstraction. As a concrete thought: a ski jump. An aircraft carrier. Something that changes the force vector and transfers it into material, not into exhaust.

The model didn’t propose this idea. I proposed it — clumsily, in a single sentence. The model expanded on it, ran the numbers, showed where it works and where it doesn’t. And came back with questions: what entry velocity? What arc radius? How many cables?

That’s exactly the division that matters.

---

What the human does, what the AI doesn’t

An LLM won’t pick the idea. Won’t judge whether it’s worth pursuing. Won’t decide which way to go — toward realism or toward a bold concept. Won’t propose that the project should be forward-looking, but rooted in what the Moon actually offers.

These were all my decisions:

• start from physics, not from vision
• adopt 2.5 tonnes as the horizon, because it covers today’s landers and tomorrow’s plans
• root the geometry in a crater — not in an abstract structure, because the crater is already there
• choose skids over legs, because it’s a logical consequence of the system, not a default solution
• decide when enough is enough — that the concept doesn’t have to be a detailed engineering project

The model contributed something else: calculations that wouldn’t fit in a head in reasonable time. The Tsiolkovsky equation for every scenario. Comparison of kinetic energies. Cross-checks against what has actually landed on the Moon and what is being planned. A web search to see whether anyone had a similar idea (someone did — Lunar Xistera from 2014, but at a completely different scale and without the key element: skids).

None of those tools replaced the choice of direction. But each of them turned a direction into something concrete.

---

Concept summary

Problem

Classical lunar landing is purely retropropulsion — engines brake the vehicle from orbit to the surface. Every kilogram of fuel had to be flown in from Earth, which at a cost of ~50,000 USD/kg on the lunar surface makes every saved tonne a financial issue, not just an engineering one.

On top of that: the vehicle needs legs. Shock absorbers. Structure capable of safely touching down on uncertain ground. That’s another 250–450 kg of flying structure, at 50,000 USD per kilogram.

Idea

If a permanent base infrastructure exists on the Moon — then the vehicle doesn’t have to land vertically on engines. It can approach at a shallow angle, slide down the natural crater slope, enter a curved arc (built from sintered regolith plates at the bottom of the slope), be redirected onto a horizontal trajectory and brake on a runway with arresting cables — exactly like an aircraft on a carrier.

Instead of legs: skids. Instead of shock absorbers: hook and cable. Instead of hover: geometry.

Geometry

• Descent trajectory: starts ~160 m above the runway level, at the crater rim. Of which:
  • ~130 m of drop is the natural crater slope at ~18° (about 420 m along the surface)
  • ~30 m of drop is the built curved transition (the arc) that straightens the velocity vector to horizontal
• Braking arc: radius of curvature R = 600 m, transition from 18° to 0° (horizontal)
• At this geometry and a speed of 80 m/s on the runway, the load factor on cargo is ~1.25 g Earth-equivalent (centripetal force + lunar gravity at the bottom of the arc) — safe for most cargo
• The radius is independent of vehicle mass (it’s determined by speed, not mass)
• The whole profile fits naturally into the morphology of a large crater — most of the work is done by the slope itself, the infrastructure is just ~30 m of curved transition made of sintered regolith
• Runway: 900 m long, surfaced with sintered regolith plates (ISRU)
• Arresting cables: 4 sets every 150 m, with hydraulic cylinders
• Energy to absorb at 2,500 kg and 80 m/s: 8.0 MJ distributed across 4 points (2.8 / 2.2 / 1.8 / 1.2 MJ)

The vehicle — skids

Instead of a classical lander with legs:

Element	Classical lander	Skids
Legs + shock absorbers	280–350 kg	—
Skid plate	—	55 kg
Hook + mechanism	—	18 kg
Remaining structure	120 kg	80 kg + rails
Total	~440 kg	~200 kg

Structural saving: ~240 kg.

Balance for a 2,500 kg vehicle

Source of saving	Mass
Structure (skids vs. legs)	~240 kg
Fuel (Δv ~80 m/s absorbed by the arc)	~95 kg
Total	~335 kg / mission

At 50,000 USD/kg → **16.75 million USD saved per mission**.

At 10 missions per year → ~167 million USD/year. The infrastructure (CAPEX ~400–700 million USD) pays for itself after 2.5–4 years of regular operation.

When it makes sense

The system is profitable when, simultaneously:

• mission frequency ≥ 10/year
• no local fuel production (fuel ISRU changes the equation)
• standardised vehicle (skids fit a single system)
• a permanent base capable of maintaining the infrastructure

Best window: early lunar base phase, before full ISRU — roughly 2035–2048.

---

Conclusion

The concept of a mechanical landing pad on the Moon isn’t new in absolute terms — similar thinking (Lunar Xistera, 2014) appeared earlier, at a completely different scale and without the skid element. What emerged from this conversation is different: smaller, more realistic, rooted in what the Moon actually offers — the crater as a natural arc, regolith as building material for the runway, cable-based braking as proven technology.

There are surely assumptions here worth refining — perhaps even some that need correcting. But for 12 iterations of conversation with an LLM, this is a solid conceptual foundation, one that could be built on further.

But more important than the concept itself is the question in the title.

Yes, you can design something like this with AI. Not because the model is creative. Because the model is able, in a few seconds, to compute things that would take me hours alone — and come back with a question that pushes the thinking forward. The decision about direction, scale, what’s worth pursuing — that was always my side of the conversation.

The model was a tool. A very good tool.

The difference between “an interesting idea that disappears” and “a concept with geometry, mass balance and a diagram” — that’s exactly this collaboration.

Example iterative LLM response

After version v3 was closed, the article went through one more verification loop — with the question of whether the 18° geometry actually fits real lunar craters. Below is the model’s verbatim response, left unchanged as a concrete example of what such an iteration looks like.

Measurement data

Slope angles of lunar crater walls (measured by the Lunar Reconnaissance Orbiter Camera):

• Average for craters <20 km: 25.5° ± 4.1°
• In fresh Copernican-aged craters, the upper 50% of the wall has slopes around ~36°, locally above 40°
• Lower wall section: 20–25° just before sharply meeting the floor
• Mass wasting threshold: ~25° (below this regolith is stable; above it, it starts to slide)
• Typical range for a “bowl”: 25–35°

18° lies below the mass wasting threshold — that’s a really very shallow angle for a crater wall. Your intuition of “27° or more” lands right on the observed median.

Three problems you rightly identified

1. The crater rim blocks an outside approach. Lunar crater rims are elevated above the surrounding terrain — a mix of ejecta and uplifted material, of which ~80% is uplifted substrate. On an outside approach at 18°, the vehicle would crash into the rim wall. Solutions: (a) approach from above, “falling” into the crater after de-orbit, (b) use a crater with a natural breach in its rim, (c) artificially carve a notch — but that’s expensive. The cleanest: the vehicle descends from orbit into the interior of the crater and only encounters the slope once it’s inside.

2. The arc is too short relative to the real shape. A real crater profile is more of a “ball impression” or a parabola — slope decreases from the rim toward the floor. With a realistic 27–30° angle, the straightening arc has these dimensions:

Angle	Drop on arc	Arc length	Drop on slope	Slope length
18°	29 m	188 m	131 m	424 m
27°	65 m	283 m	95 m	209 m
30°	80 m	314 m	80 m	160 m

At 27–30° the arc becomes a much more substantial structure — which better reflects the fact that the arc is the last, curved section of the slope rather than a short “clip” at the end of a long ramp.

3. Flat floor only in complex craters. Here the key distinction:

• Simple crater (<15 km diameter): bowl-shaped, d/D ≈ 1/5, no flat floor — a 900 m runway will not fit
• Complex crater (>15–20 km): terraced walls, flat floor, sometimes a central peak, d/D ≈ 1/10 to 1/20

For a 900 m runway you need a complex crater, e.g. ~20–30 km in diameter. Then the floor is a few km across and 900 m fits without issue. In a simple crater (up to 15 km) the runway simply won’t fit, because the floor is a bowl.

What this means for the concept

Good news: the physics of the arc doesn’t change. The ~1.25 g load factor depends only on v²/R — speed on the runway and the radius of the arc — not on the entry angle. The 8.0 MJ of energy stays the same too, because it depends on the runway speed. So does the Δv of ~80 m/s in fuel savings.

What needs to be updated in v3:

• Entry angle: 27° (not 18°) — consistent with the average wall slope of lunar craters
• Arc: 65 m drop, 283 m long (instead of 29 m / 188 m) — a larger structure, so the cost estimate likely shifts as well
• Natural slope: 95 m drop, 209 m along the surface (instead of 131 m / 424 m)
• Character of the approach: the vehicle descends from orbit into the interior of a complex crater (rather than approaching from outside over the rim), which requires a separate de-orbit and over-rim entry phase — non-trivial, because the velocity vector has to be managed so it meets the slope at 27°
• Crater requirement: minimum ~20 km diameter (complex crater with a flat floor), ideally somewhat larger for margin

My recommended choice: 27° — because it matches the observed mean and splits the descent roughly half-and-half between natural slope (95 m) and built arc (65 m). 30° would be even “nicer” mathematically (50/50 in metres), but 27° fits the observed distribution of slopes better.

Want me to prepare a v4 of the article with these corrections and an updated diagram? These are cosmetic changes for the economic logic (savings stay identical — 335 kg/mission, 167 million USD/year), but big ones for the credibility of the geometry and the description of the approach.

---

The calculations were correct — but for a specific slope on a surface, not for a crater. This doesn’t change the fact that it’s a good conceptual foundation; what it shows is that even well-computed geometry can be put in the wrong context, and that only a question from a different angle — “does this even fit the actual shape of a crater?” — pulls the problem to the surface. That’s exactly the collaboration from the title: the model didn’t notice that 18° doesn’t fit craters. I didn’t notice it until something twitched in my intuition. Only together did we arrive at the realisation that this is a problem.