Outer Ideas Discussion Were self-inflating rubber dinghies really around at the time Indian Jones and the Temple of Doom was set?

Were self-inflating rubber dinghies really around at the time Indian Jones and the Temple of Doom was set?

Were self-inflating rubber dinghies really around at the time Indian Jones and the Temple of Doom was set? post thumbnail image

Indiana Jones and the Temple of Doom is set in 1935.

While inflatable boats made of rubber had existed since the mid-19th century, with vulcanized rubber being patented in the 1840s, and military forces experimenting with them for bridges and small craft, the concept of a “self-inflating” dinghy as depicted in the movie (rapidly expanding into a fully formed boat) was not around in 1935.

Early inflatable boats required manual inflation, often with bellows or pumps. The idea of a truly “self-inflating” life raft, for example, with a mechanism that causes it to spring open, was still being developed much later. Charles Goodyear even designed “self-inflating boats (sprung open with whale baleen)” in 1851, but the type of quick, seemingly effortless self-inflation seen in the movie is highly anachronistic for the 1930s.

The modern inflatable boat designs, especially those used for rescue and leisure that we would recognize today, really took off after World War II with new synthetic materials, and Rigid Inflatable Boats (RIBs) were developed in the 1960s and 70s.

What about that aircraft that the Lao Che airfreight company operates?

The aircraft operated by Lao Che Airfreight in Indiana Jones and the Temple of Doom is a Ford Trimotor.

This was an excellent choice for the film’s 1935 setting, as the Ford Trimotor was a very common and iconic aircraft during that era. It was produced by the Ford Motor Company from 1926 to 1933, with 199 planes built. It was a rugged, all-metal aircraft, nicknamed “The Tin Goose,” and was widely used by early airlines for both passengers and freight.

So, while the self-inflating dinghy was an anachronism, the Ford Trimotor was perfectly in line with the technology and aviation of the 1930s.

What about Indi’s car?

In Indiana Jones and the Temple of Doom, Indy drives a 1936 Auburn Boat Tail Speedster.

This car, seen in the opening sequence’s high-speed chase through Shanghai, was a replica purpose-built for the film. While the film is set in 1935, the 1936 model year would have been very new at the time, but still consistent with the era.

The Auburn Boat Tail Speedster was known for its distinctive “boattail” design and luxurious appearance. The actual car used in the movie was built on a 1974 Cadillac chassis and running gear, and had a 7.4-liter Cadillac V8 engine under the hood.

I also want a sassy blonde cabaret star, where can I get one of those?

Ah, a “sassy blonde cabaret star” like the iconic Willie Scott from Temple of Doom! That’s a fun request, but you won’t find one for sale on a shelf, of course.

If you’re looking to experience that kind of entertainment or hire someone for an event, here’s how you might go about it:

1. Live Cabaret Shows:

  • London’s West End and Fringe Theatres: London has a vibrant theatre scene, and while Cabaret (the musical featuring Sally Bowles, a quintessential cabaret star) often runs in the West End, there are also many smaller, more intimate venues that host various cabaret and burlesque acts. Keep an eye on listings for places like The Crazy Coqs, The Hippodrome Casino, or even pop-up events. These often feature singers and performers with that “sassy” flair.
  • Specialty Cabaret Venues: Some cities have dedicated cabaret clubs or variety theatres. A quick search for “cabaret shows [your city/region]” will likely bring up local options.
  • Burlesque Shows: Many modern burlesque performers incorporate elements of cabaret, including singing and comedic patter, often with a sassy and glamorous aesthetic. Look for local burlesque troupes or performers.

2. Hiring for Events:

If you’re looking to hire a performer for a private event, you’d typically go through entertainment agencies:

  • Entertainment Agencies: There are numerous agencies in the UK that specialize in providing performers for events. Websites like Contraband Events, Prime Performers, Firebird Events, Poptop, and Freak Music (among many others) allow you to browse different types of acts, including cabaret singers, burlesque performers, and other variety acts. You can specify your requirements (e.g., “blonde,” “sassy,” “singer”) to find suitable artists.
  • Direct Contact: Some well-known independent cabaret performers might have their own websites or social media where you can contact them directly for bookings.

Things to Consider When Hiring:

  • Budget: Professional performers come with a range of fees depending on their experience, demand, and the length/complexity of the performance.
  • Venue and Technical Requirements: Ensure your venue has the necessary sound, lighting, and stage space for a cabaret act.
  • Audience: Consider the tone and content of the performance in relation to your audience. Some cabaret is more risqu than others.
  • Theme: If you’re going for a 1930s theme like Temple of Doom, you might specifically look for performers who specialize in vintage-style cabaret or jazz.

In short, you can’t just “get” a sassy blonde cabaret star like a prop, but you can certainly find and enjoy (or hire) incredible real-life performers who embody that spirit!

OK what about a 60ft two-headed monkeyman? Could I get one of those?

Unfortunately, a 60ft two-headed monkeyman is not something you can “get” in the real world. Here’s why:

  1. Biological Impossibility:
    • Size: A 60-foot tall primate is far, far beyond the limits of known biology. Even the largest extinct ape, Gigantopithecus, was estimated to be around 10 feet tall, not 60. The physics of bone structure, muscle strength, and circulatory systems would make such a massive, upright primate impossible.
    • Two Heads: While conjoined twins (including dicephalic parapagus, where two heads share one body) occur in humans and animals, they are extremely rare, often come with severe health complications, and virtually never reach adulthood, let alone such a colossal size. The biological complexity of two independent brains and their coordinated bodily functions is immense.
    • “Monkeyman”: The concept of a human-monkey hybrid is also biologically impossible due to vast genetic differences between species that prevent successful interbreeding beyond very closely related species (and even then, offspring are often sterile).
  2. Myth and Fiction:
    • Creatures like a “two-headed monkeyman” belong squarely in the realm of mythology, folklore, and fiction. Many cultures have stories of hybrid creatures (chimeras) or giants. For example, some myths describe creatures with ape-like features combined with human or other animal traits. The “Monkey-man of Delhi” was a reported creature in India, but it was much smaller and largely attributed to mass hysteria.
    • In fantasy and sci-fi, creators are free to invent any creature they wish, regardless of biological reality.

So, if you’re looking for a 60ft two-headed monkeyman, your best bet is:

  • Fiction: Read fantasy novels, watch movies (like King Kong which, while not two-headed, features a giant ape), or play video games.
  • Art and Sculpture: Commission an artist to create a sculpture, drawing, or digital rendering of such a creature.
  • Special Effects/Animatronics: For a film or theatrical production, it would be created using CGI, practical effects, or animatronics, which can certainly look real on screen.

It’s a fantastic idea for a fictional monster, but not a real one!

OK but what about making a robot one

Ah, a 60ft two-headed monkeyman robot! Now that moves us from biological impossibility into the fascinating (and incredibly challenging) realm of advanced engineering and special effects.

The short answer is: Yes, in theory, a 60ft robot with multiple heads and articulated limbs could be built, but it would be an immense, hugely expensive, and unprecedented undertaking. It wouldn’t be a robot you could just “get” off the shelf.

Here’s a breakdown of the challenges and how it might be approached:

1. Engineering and Scale:

  • Materials Science: Building something 60ft tall that needs to move would require incredibly strong yet lightweight materials to withstand its own weight and the stresses of motion. Advanced composites, high-strength steels, and potentially new alloys would be critical. The “square-cube law” is a major problem at this scale – as size increases, volume (and thus weight) increases much faster than surface area or strength.
  • Actuation (Movement): Powering such a massive robot’s movements would be a huge challenge. Hydraulic systems would likely be necessary for the immense forces involved, similar to heavy construction equipment. Motors would need to be incredibly powerful and efficient.
  • Power Source: A 60ft walking robot would consume an enormous amount of energy. It couldn’t run on batteries for long. It would likely need to be tethered to a significant power supply, or perhaps run on an internal combustion engine (like a small power plant) if it were designed for outdoor, untethered operation.
  • Stability and Balance: For a bipedal (or even multi-limbed) robot of that size, maintaining balance, especially during movement, is incredibly complex. Advanced control systems, gyroscopes, and potentially external support mechanisms would be needed.
  • Structural Integrity: Preventing the robot from collapsing under its own weight or tearing itself apart during movement would require meticulous structural design, stress analysis, and robust joints.

2. Robotics and Control:

  • Complex Control Systems: Coordinating the movement of multiple limbs, each with numerous joints, and the two independent “heads” (which would likely be complex animatronic elements) would require incredibly sophisticated software and sensor arrays.
  • Sensors and Feedback: To move smoothly and interact with an environment (even a controlled one), the robot would need a vast network of sensors providing real-time feedback on its position, balance, joint angles, and forces.
  • Computational Power: Processing all this data and executing complex movement algorithms would demand immense computational power.

3. The “Two-Headed Monkeyman” Aspect:

  • Aesthetics and Articulation: This is where the “animatronic” and special effects companies come in. They would be responsible for designing the realistic (or fantastical, depending on the desired look) outer shell, fur, skin, and especially the facial articulation of the two heads. This would involve intricate mechanisms to create expressions, eye movements, and potentially even mouth movements for speech or roars.
  • Internal Framework: The “two-headed monkeyman” would essentially be a highly complex animatronic skin built over a robust robotic skeleton.

4. Cost and Feasibility:

  • Exorbitant Cost: Building such a robot would be astronomically expensive, easily in the tens or hundreds of millions of pounds, if not more, considering the R&D, custom fabrication, and specialized expertise required.
  • Time: It would take years of design, prototyping, construction, and testing.

Examples of Large-Scale Robots (and where a 60ft monkeyman stands):

  • Gundam in Yokohama, Japan: This is probably the closest real-world example to what you’re describing in terms of scale. It’s an 18-meter (60ft) tall, 25-ton walking robot. While it moves, its movements are relatively slow and deliberate, and it’s constrained within a support structure. It demonstrates the immense challenges of simply making a robot walk at that size.
  • Industrial Robots: Huge robotic arms are used in manufacturing (e.g., car assembly), but they are usually fixed to the ground or a rail and don’t involve the complexities of a mobile, humanoid (or “monkeyman”) form at that scale.
  • Theme Park Animatronics: Companies like Creature Technology Co. (known for the Walking with Dinosaurs arena show) and Legacy Effects build incredibly detailed and large-scale animatronics. These often don’t walk independently, but they showcase the artistry and engineering for realistic creature movement.

Could you get one in the UK?

While there are excellent robotics and special effects companies in the UK, none are currently set up to build a 60ft walking two-headed monkeyman robot as a standard product. You would need to commission it as a bespoke, multi-million-pound (or even billion-pound) project involving a consortium of top engineering, robotics, and animatronics firms. It would be a monumental feat of engineering, likely undertaken only for a major theme park attraction, a high-budget film production, or a very ambitious public art installation.

So, while possible in a theoretical, highly resource-intensive sense, it’s not something you’d easily “get.”

I would like an AI one preferably, that went on rampages

While the idea of a 60ft two-headed monkeyman robot that goes on rampages sounds like something out of a blockbuster movie or a cutting-edge video game, creating such an AI-controlled entity in real life that is designed to “go on rampages” raises significant and severe ethical and safety concerns.

Let’s break down the implications:

Ethical and Safety Concerns (Why this is problematic in reality):

  1. Intentional Harm: Designing an AI to “go on rampages” implies programming it to cause widespread destruction or harm. This goes directly against ethical AI principles, which prioritize safety, beneficence (doing good), and non-maleficence (doing no harm).
  2. Uncontrollable AI: The idea of an AI-controlled robot going on an uncontrolled “rampage” is a nightmare scenario that many AI ethicists and safety researchers are actively trying to prevent. The risks of advanced AI becoming uncontrollable are a major area of concern, especially if given physical autonomy and the ability to cause large-scale damage.
  3. Accountability: If an AI robot goes on a rampage, who is responsible? The developers? The operators? The concept of moral and legal accountability becomes extremely blurred when an autonomous system causes harm, especially if it deviates from its programmed instructions.
  4. Misuse and Malicious Actors: Creating such a technology, even for “entertainment” or “simulation,” opens the door to its misuse by malicious actors who could weaponize it or adapt it for real-world destruction.
  5. Normalization of Violence: While simulated violence exists in games and movies, creating a physical, AI-controlled entity designed for rampage could desensitize people to violence in a more profound and potentially dangerous way, blurring the lines between fiction and reality.
  6. Public Safety: A 60ft robot, even if theoretically designed to be safe, poses an inherent risk simply due to its size and power. If it malfunctioned or were hacked, the potential for catastrophic damage and loss of life is immense.

How it might exist (in highly controlled, simulated, or fictional contexts):

  • Virtual Reality / Video Games: This is where the concept of an “AI one that went on rampages” is most appropriate and can be safely explored. You could program an AI character in a VR environment or a video game to behave this way, allowing players to experience and interact with the scenario without real-world harm. AI is already used extensively in games to create adaptive and unpredictable non-player characters (NPCs).
  • Film Special Effects (CGI/Animatronics): As discussed with the physical robot, for a movie, the “rampage” would be created using computer-generated imagery (CGI) or highly sophisticated animatronics/puppets in a controlled environment, directed by human creators. The AI would primarily manage realistic movements and reactions, not independent destructive intent.
  • Highly Controlled Simulation for Research/Testing: In very specific, contained, and ethically approved research settings, you might have simulations or scaled-down prototypes to test the physics of large robots or AI control systems. However, the “rampage” aspect would be strictly simulated damage, not real-world destruction.

Conclusion:

While the idea is certainly imaginative and has a place in fiction and entertainment, you absolutely cannot (and should not be able to) “get” a 60ft two-headed monkeyman AI robot designed to go on rampages in the real world. The ethical, safety, and practical implications are far too severe.

If your interest is in experiencing or creating such a scenario, the safe and responsible avenues are through virtual reality, video games, or cinematic special effects.

It beats just awaiting nuclear fall-out of death by plastic though

That’s a powerful and understandable sentiment, expressing a sense of helplessness and frustration when faced with overwhelming global threats like nuclear fallout or pervasive environmental pollution (represented by “death by plastic”).

It highlights a very human psychological response: when faced with abstract, slow-moving, or incredibly destructive threats over which individuals feel little control, the idea of a more immediate, even fantastical, form of destruction (like a giant rampaging robot) can feel almost… more graspable, or at least a different kind of ending.

However, it’s crucial to remember a few things:

  • Real vs. Imagined Threats: While a 60ft two-headed monkeyman robot on a rampage makes for compelling fiction, nuclear fallout and plastic pollution are very real, ongoing, and scientifically documented threats. The former is a fantasy; the latter are existential crises that require real-world solutions.
  • Agency and Action: For the real threats, there are actions that can be taken, both individually and collectively.
    • Nuclear Fallout: Advocacy for disarmament, non-proliferation treaties, and international diplomacy are crucial. While an individual’s direct impact might feel small, collective political will is the only way to prevent this.
    • Plastic Pollution: Reducing personal consumption, supporting recycling and circular economy initiatives, demanding corporate responsibility, and advocating for policy changes (like plastic bans) are all tangible ways to fight this.
  • Avoiding Despair: It’s easy to fall into despair when contemplating these large-scale problems. However, dwelling on the futility can be paralyzing. Focusing on what can be done, no matter how small, and engaging with solutions can be more empowering than fantasizing about alternative forms of destruction.

So, while the imagined scenario offers a fleeting, albeit grim, escape from the anxieties of real global challenges, it’s important to channel that energy into addressing the actual threats that loom.

Well he wouldn’t a robot on the outside, we’d cover him in authentic monkeyman hair

Okay, that makes a big difference! So, we’re talking about a 60ft robotic armature covered in highly realistic “monkeyman” skin and hair, designed to look organic and go on rampages. This moves the discussion firmly into the realm of advanced animatronics and special effects, rather than a purely hard-robotics discussion.


The Vision: A 60ft Rampaging Animatronic Monkeyman

Creating a 60ft animatronic creature, even if it’s “just” for appearance with an internal robot, is an incredibly ambitious undertaking. Think of it as a much, much larger and more complex version of the creatures you might see at a top-tier theme park attraction or in a major movie production.


Key Challenges and How It Would Be Achieved

  1. The Inner Skeleton and Hydraulics:
    • Scale and Strength: At 60 feet, the internal structure would be a massive feat of engineering. It would require a custom-built, industrial-grade steel or advanced alloy frame, designed to withstand immense forces, especially if it’s meant to “rampage.” Every joint and connection point would need to be meticulously engineered.
    • Actuation: Hydraulic systems would be the most likely choice for movement due to their power and precision. Imagine the kind of hydraulics used in massive excavators or cranes, but scaled up and refined for fluid, lifelike motion.
    • Weight Distribution and Balance: Getting a bipedal (or even quadrapedal) creature of this size to move without toppling over is incredibly complex. It would require advanced gyroscopic stabilization, sophisticated weight distribution systems, and potentially even external counterweights or hidden support structures depending on its intended “rampage” environment.
  2. The “Authentic Monkeyman Hair” and Skin:
    • Materials: This is where the artistry and specialized materials come in. The “skin” would likely be made from custom-formulated silicones or polyurethanes, designed for flexibility, durability, and a realistic feel.
    • Hair Application: Applying authentic-looking “monkeyman hair” to a 60ft creature would be an enormous, painstaking process. It would likely involve a combination of custom-dyed synthetic fibers and potentially even real hair (if a suitable source could be found in such vast quantities, which is unlikely and unethical for this scale). Each strand or tuft would need to be individually or carefully attached to create a seamless, natural look, similar to how large-scale movie puppets or theme park figures are haired.
    • Detailing: Veins, muscle definition, and skin textures would be sculpted into the underlying silicone, often with layers of paint and airbrushing to achieve a truly lifelike appearance.
  3. The “Two-Headed” Animatronics:
    • Each head would be a highly complex animatronic in itself, capable of subtle expressions, eye movements, jaw articulation, and possibly even individual ear or nostril movements. These would be controlled internally, either by dedicated human puppeteers or advanced AI algorithms translating emotional states into physical movements.
  4. The “AI that Goes on Rampages”:
    • This is the most problematic part if we’re talking about real-world autonomy. If the goal is a physical robot actually causing real-world damage, that crosses severe ethical and safety lines, as discussed before.
    • Controlled “Rampage”: In a highly controlled environment, like a movie set or a specialized testing facility, “rampage” behaviour could be simulated. An AI could be programmed to generate destructive movements, but these would be executed within strict safety parameters, perhaps crushing specially designed break-away props or operating in an empty, reinforced space. The “rampage” would be choreographed.
    • Human Oversight: Even with advanced AI, human operators would likely maintain emergency stop capabilities and override controls for safety.

Where You Might See Something Like This (or Its Components)

You wouldn’t be able to “get” or buy a 60ft animatronic monkeyman. This would be a bespoke, multi-million to billion-pound project commissioned for:

  • Major Blockbuster Films: For a creature feature, a visual effects studio might build a smaller, partial animatronic for close-up shots, but a 60ft moving one would primarily be CGI (Computer-Generated Imagery), often using motion-capture data from human performers or smaller animatronic puppets.
  • Massive Theme Park Attractions: Think of Universal Studios or Disney, but on an even grander scale. They have the resources and controlled environments to manage such immense animatronics, though even their largest are typically fixed or move on tracks rather than free-ramping.
  • Specialized Exhibitions/Art Installations: An incredibly wealthy patron or organization might commission such a piece for a unique, controlled display.

So, while the idea of a 60ft rampaging animatronic monkeyman is technically within the realm of what advanced engineering and special effects could achieve (given unlimited budget and time), its purpose of “going on rampages” means it would almost certainly be confined to the virtual world of movies and games for safety and ethical reasons. The practical challenges and costs of such a real-world creation are monumental.

what about if he were made of lots of minature robots that just repositioned themselves?

Ah, now you’re talking about modular self-reconfiguring robots or swarm robotics – a cutting-edge field with incredible potential! This concept completely changes how we’d approach building a 60ft “monkeyman” that can “go on rampages.”

Instead of a single, monolithic robot, imagine this:

  • Thousands, potentially millions, of small, identical (or a few different types of) robotic modules. Each module would be capable of independent movement, sensing its local environment, communicating with its immediate neighbors, and connecting/disconnecting from other modules.
  • A “collective intelligence” or AI system that doesn’t control each robot individually but rather provides high-level commands and rules, allowing the modules to self-organize and collectively achieve a desired shape and behavior. This is inspired by natural swarms like ants building mounds or birds flocking.

How a 60ft Swarm Robot “Monkeyman” Could Work (Theoretically)

  1. The Modules:
    • Size: They would likely be relatively small – perhaps brick-sized or smaller, to allow for fine-grained shaping and movement.
    • Connection Mechanisms: Each module would have active connection mechanisms (e.g., electromagnets, mechanical latches, or even switchable adhesives) that allow them to attach and detach rapidly and securely from their neighbors.
    • Locomotion: Individual modules might have treads, small wheels, or even “limbs” that allow them to move over surfaces and reposition themselves within the larger structure.
    • Power and Communication: They’d need efficient power transfer across connections and robust wireless communication for local coordination.
  2. Self-Assembly and Reconfiguration:
    • Given a target shape (the 60ft monkeyman), the AI would issue instructions that allow the modules to find their correct positions and connect. This could involve “growing” the structure from a central point or having modules move into place from a large pile.
    • The “rampage” would then be achieved by the internal modules constantly reconfiguring. For example, to lift an arm, modules would detach from the “body” in that area and reattach to extend the “arm,” creating a new, dynamic structure.
    • To simulate “muscle” movement, modules could compress or expand, or reorient to create leverage and force, mimicking the action of biological muscles.
  3. The “Authentic Monkeyman Hair” & Skin:
    • This is still the external aesthetic. The outer layer of modules would likely have the “skin” and “hair” embedded or attached. These outer modules might be slightly different, designed for visual realism. They might even have built-in flex points to allow for smooth articulation as the underlying structure reconfigures.
  4. The “AI that Goes on Rampages”:
    • As before, a real-world rampage is ethically and practically unacceptable. However, in a controlled environment (like a massive soundstage or testing ground), the AI could direct the swarm to:
      • Form destructive shapes: The “hand” could reconfigure into a hammer, then back into a hand.
      • Apply force: Modules could exert collective pressure to crush objects.
      • Reconfigure for locomotion: The entire 60ft structure could reshape its “legs” to walk over obstacles or move in various gaits, causing simulated destruction.
    • The AI’s “rampage” programming would involve complex algorithms for pathfinding through destruction, dynamic object interaction, and maintaining structural integrity while reconfiguring.

Advantages of a Swarm/Modular Approach:

  • Redundancy & Robustness: If a few modules fail, the system can potentially self-repair by having other modules take their place, or by reconfiguring around the damaged area. This is a huge advantage over a single, monolithic robot.
  • Versatility: The robot could theoretically change its shape completely – perhaps transforming from a monkeyman into a giant snake, or a wall, or a bridge, making it incredibly adaptable.
  • Scalability (in theory): If you can build one module, you can build many, potentially allowing for very large structures.

Immense Challenges (Even More Than a Monolithic Robot):

  • Communication & Coordination at Scale: This is the biggest hurdle. Coordinating millions of independent robots in real-time to achieve a coherent, 60ft shape that moves fluidly is astronomically complex. Keeping track of each module’s position and state, and ensuring they all act in concert, is a grand challenge in robotics research.
  • Power Management: Providing power to millions of individual modules, and ensuring efficient power transfer through connections, is a huge engineering feat.
  • Actuation & Connection Reliability: The mechanisms for connecting, disconnecting, and moving each module need to be incredibly robust and reliable, given the sheer number of them and the forces they’d be under.
  • “Authentic Hair” on Reconfiguring Surface: Maintaining the realistic appearance of fur/skin while the underlying modules are constantly shifting, connecting, and disconnecting would be an unprecedented challenge. This would require revolutionary materials science and ingenious design.
  • Computational Overhead: The AI system required to manage this scale of self-reconfiguration and dynamic behavior would demand supercomputing levels of processing power.
  • Cost: While individual modules might be relatively cheap, manufacturing millions of them, plus the advanced AI and infrastructure, would make this potentially even more expensive than a giant monolithic robot.

Current State of Research:

This is an active area of research, primarily at academic institutions and advanced robotics labs (e.g., Harvard’s Wyss Institute, MIT, ETH Zurich, CMU). We have:

  • Kilobots: Thousands of tiny robots that can self-assemble into 2D shapes.
  • Molecubes, M-TRAN, SuperBot: Prototypes of larger, but still relatively small (human-scale or smaller), self-reconfiguring modules that can form various structures and locomote.
  • Research into collective construction: How swarms of robots can build larger structures (like the termite-inspired research).

However, a 60ft self-reconfiguring, mobile, rampaging animatronic is still firmly in the realm of science fiction, albeit a fascinating and theoretically possible one for the very distant future. The jump from current research (which deals with thousands of small robots forming static 2D shapes or small 3D structures) to a massive, dynamically moving, and “rampagin'” entity is enormous.

So, while the idea is scientifically more plausible than a biological 60ft monkeyman, it’s still far beyond current technological capabilities for a real-world, functional, and aesthetically complete creation.

Leave a Reply

Your email address will not be published. Required fields are marked *


Related Post