Exoskeletons
Exoskeletons don’t restore movement—they reframe disability as an engineering constraint.
AI
Machines sold as thinking: what the models actually do, where the limits sit, and which claims survive contact with the mechanism.
- 70
- briefs
- 10:09
- average
- 710 min
- in total
- 18
- inventors
Fields under AI
All briefs
Machine learning
Machine learning is not intelligence — it is applied statistics with memory.
Graphics processing unit
The GPU didn’t accelerate graphics—it relocated computation, and in doing so, rewrote the rules of what a chip could be used for.
Technological singularity
The singularity is not a prediction — it's a surrender note written in advance of the argument.
Relational model
Codd · 1969
Codd’s 1969 relational model is a formal specification—not software, not SQL, not a product. It defines data as tuples in relations, enforces consistency via logic, and insists all information be explicit. Real databases violate it daily. That gap is where data engineering begins.
Nanotechnology
Nanotechnology is a scale-based definition—not an engineering discipline—and its real-world impact lies in bulk materials, not molecular machines.
Orange (software)
Visual programming doesn’t replace coding — it defers the hard parts until they’re unavoidable.
AI agent
AI agents promise autonomy — but deliver orchestration dressed as agency.
Unmanned aerial vehicle
UAVs don’t fly free—they inherit military logic, evade civilian oversight, and deliver capability before consent.
Mobile robot
Mobile robots are not intelligent agents—they are guided machines whose autonomy is strictly bounded by sensor fidelity and route control.
Unmanned surface vehicle
Autonomy at sea is real—but it’s fragmented, unstandardised, and still tethered to human oversight.
Self-driving truck
Autonomous trucks are not driving on highways—they’re leasing roads, not lanes.
Cobot
‘Cobot’ isn’t a robot type—it’s a risk-assessment label you have to verify yourself.
Domestic robot
Personality was a feature before autonomy was reliable.
Military robot
Military robots don’t replace soldiers—they reconfigure accountability, delay ethical decisions, and make war faster than the law can keep up.
Rethink Robotics
Rodney Brooks · 2008
Rethink Robotics launched Baxter in 2012 as a collaborative industrial robot designed to work safely beside humans — using physical stoppage and manual override, not AI. It released Sawyer in 2015 for finer tasks, then went global with cobots in 2016. It ceased operations in October 2018. Its assets were sold. Its later relaunch attempts failed within a year. Its safety mechanisms worked. Its business model did not.
Robotic vacuum cleaner
There was no robotic vacuum cleaner in 1990 — only toys, fiction, and TV props.
Da Vinci Surgical System
Intuitive Surgical
The Da Vinci Surgical System is a teleoperated robotic platform for minimally invasive surgery. It translates surgeon hand motions at a console into precise movements of up to four robotic arms at the patient’s side — one for 3D vision, the rest for tools. FDA-cleared in 2000, it had performed ~200,000 surgeries by 2012, mainly prostatectomies and hysterectomies. It does not demonstrate long-term superiority over laparoscopy. It demands high capital cost, proprietary software, dedicated training, and a full surgical team. Safety investigations by the FDA include reports of patient injury and death.
Atlas (robot)
Atlas wasn’t built to save lives—it was built to prove they could be saved, one lab demo at a time.
DJI Phantom
DJI
The DJI Phantom series launched in 2013 as an entry-level, GPS-stabilised quadcopter. It pioneered the consumer aerial photography drone form factor. It was DJI’s first fully assembled drone. It was priced at $629. It achieved 15 minutes of flight time. It supported optional GoPro mounting. It used a GPS-enabled NAZA-M autopilot. The Phantom 2 Vision introduced the first built-in camera in the series. The line was discontinued after the Phantom 4.
ASIMO
Honda
ASIMO was Honda’s eleventh humanoid robot, built to test bipedal navigation in human environments. It worked — but only in narrow conditions. It climbed stairs autonomously, yet fell twice. It recognised faces and gestures, but only pre-trained ones. It ran one hour on lithium-ion power, but never left the lab. Retired in 2022, it changed robotics research, not robotics use.
Robotic arm
Robotic arms don’t think—they calculate, repeat, and lift. Their intelligence is borrowed, their strength bolted, and their autonomy a promise, not a feature.
Go (programming language)
Robert Griesemer, Rob Pike, Ken Thompson · 2009
Go is a statically typed, compiled programming language designed in 2007 by Robert Griesemer, Rob Pike, and Ken Thompson at Google to improve programming productivity for multicore, networked machines and large codebases. It combines static typing and runtime efficiency like C, readability and usability like Python, and high-performance networking and multiprocessing. Publicly announced in November 2009 and released as version 1.0 in March 2012, Go emphasizes simplicity, safety, and developer tooling — especially around concurrency via goroutines and channels, packaging, testing, and deployment — while deliberately omitting features like inheritance, assertions, pointer arithmetic, and implicit type conversions. Its design reflects a focus on the software development process itself rather than just language syntax or semantics.
Haptic technology
Haptic tech doesn’t replicate touch—it substitutes it with calibrated vibrations, forces, and motions, and that substitution is all it ever promised to do.
Backpropagation
Backpropagation is not learning—it’s bookkeeping for learning.
Convolutional neural network
It did not invent vision—it invented the first scalable, trainable scaffold for spatial pattern recognition.
Diffusion model
Diffusion models don’t generate images—they reverse entropy, one noisy step at a time.
Expert system
Expert systems don’t think — they transcribe. What they deliver depends entirely on what experts wrote down, and nothing more.
Generative adversarial network
GANs don’t model reality—they train a mimic to beat a critic.
PyTorch
Facebook · 2016
PyTorch is a Python-first deep learning library built by Meta in 2016 as Torch's successor. It enables rapid model prototyping via dynamic computation graphs (Autograd), tensor abstractions, and CUDA-accelerated training. As of 2025, it remains one of the most popular deep learning libraries—alongside TensorFlow and Keras—but its design prioritises researcher agility over production robustness, scalability, or deployment tooling.
Reinforcement learning from human feedback
RLHF outsources ethics to annotators—and then optimises for their rankings, not their intent.
Robotaxi
Robotaxis are running — but they are losing money, failing safety benchmarks, and failing to earn public trust.
TensorFlow
Google · 2015
TensorFlow is a machine learning API developed by Google. It is used mainly for training and inference of neural networks. It is free and open-source software released under the Apache License 2.0. It was developed by the Google Brain team for Google's internal use in research and production. The initial version was released under the Apache License 2.0 in 2015. Google released an updated version, TensorFlow 2.0, in September 2019. TensorFlow can be used in a wide variety of programming languages, including Python, JavaScript, C++, and Java. It is one of the most popular deep learning frameworks, alongside others such as PyTorch.
Vector database
It doesn’t understand meaning — it just maps distances between precomputed guesses.
Gaussian splatting
Gaussian splatting ditches ray-marching for splatting—and gains speed by abandoning geometry.
LoRA (machine learning)
LoRA doesn’t shrink models—it shrinks the cost of changing them.
Grafana
Grafana Labs
Grafana is open-source dashboard software that queries external data sources at render time. It unifies metrics, logs, and traces without storing data itself. Its JSON-based dashboards are versionable and composable. Since 2021, it uses AGPLv3 licensing. It originated from Graphite work at Orbitz and launched in January 2014. Its UI began as a fork of Kibana v3.
Kotlin
Kotlin isn’t about elegance—it’s about eliminating Java’s tax on velocity, one stable release at a time.
Motion capture
Motion capture records movement — not performance, not intention, not life.
Photogrammetry
Photogrammetry doesn’t capture reality — it reconstructs geometry from the constraints of light, lens, and overlap.
Agricultural robot
It replaces people — but only where speed doesn’t matter.
Delivery robot
Delivery robots are not the future of logistics — they are a stopgap for places where humans are expensive and terrain is tame.
Humanoid robot
Humanoid robots are not stepping into our homes—they’re stepping into our labs, as calibrated dummies for prosthetics and rehab.
Industrial robot
Industrial robots didn’t arrive with AI—they arrived with paper tape, solenoids, and a single-minded refusal to tire.
Self-driving car
Self-driving cars are not autonomous — they are geographically gated, software-limited, and already lethal.
Segway
Segway Inc.
The Segway PT is a mechanically elegant, narrowly useful device born from medical robotics. It solved balance precisely, but failed to solve adoption. Its legacy is not in streets, but in patents, pivots, and the cautionary weight of overpromising hardware.
ChatGPT
ChatGPT is not intelligence — it’s a prompt-to-output relay, optimised for speed, not truth.
ELIZA
ELIZA doesn’t talk — it mirrors, misleads, and exposes how little it takes for humans to believe a machine is listening.
Google Search
Google Search didn’t organise the web—it redefined authority as a function of linkage, then language, then generation.
ImageNet
Fei-Fei Li
ImageNet is a foundational dataset infrastructure — not a model, not a tool, not a service. It succeeded by making scale measurable, not magical.
Siri
Siri didn’t invent voice control—it outsourced understanding to Nuance and offloaded thinking to the cloud.
Tesla Autopilot
Tesla Autopilot passes NHTSA’s test — but only because the test measures what Teslas do, not what drivers must still do.
Amazon Echo
Amazon · 2014
The Echo is a hardware interface for Alexa—a cloud-dependent voice service. It works well for narrow, internet-reliant tasks. It fails as a private, autonomous, or universally accessible system. Its legacy is not intelligence at the edge, but the normalisation of persistent listening and remote AI dependency.
DARPA Grand Challenge
The first driverless car race didn’t produce a working robot — it produced a better way to break them.
Arduino
Arduino didn’t democratise hardware — it outsourced the hard parts to hobbyists.
Cotton gin
Eli Whitney · 1793
The cotton gin is a mechanical seed-removal device invented by Eli Whitney in 1793. It uses wire hooks on a rotating drum to pull cotton fibres through narrow ginning ribs, while a second rotating brush cylinder clears lint to prevent jams. It increased daily lint output from one pound (manual) to 50–55 pounds. This made cotton farming highly profitable, expanded U.S. cotton production, and catalysed a surge in exports—from under 500,000 pounds in 1793 to 93 million pounds by 1810. But it did not reduce overall labour demand. Instead, it shifted the bottleneck to field harvesting, intensifying reliance on enslaved people to pick cotton by hand. Its patent was granted in 1794 but not validated until 1807.
Endoscope
Philipp Bozzini
Bozzini’s Lichtleiter established endoscopy as a category: a purpose-built optical conduit for internal light and view. It worked technically in most cavities. It failed institutionally — blocked by authority, not physics. Its legacy is conceptual, not clinical.
Gyroscope
Léon Foucault · 1852
The gyroscope is a demonstration device, not a tool. It makes Earth’s rotation locally visible — for less than ten minutes — using nothing but inertia and friction. It changed pedagogy, not engineering.
Magnetometer
Gauss and others
Gauss’s 1833 magnetometer established absolute measurement of Earth’s magnetic field strength using oscillation frequency differences — a tenfold precision gain over predecessors. It was the first instrument to derive a non-mechanical quantity from mechanical fundamentals. It did not measure direction or local variation; it required stable conditions and assumed ideal magnetic and torsional behaviour. Its legacy is metrological, not operational.
Pressure cooker
Denis Papin
The steam digester is a 1679 experimental vessel that used trapped steam to raise boiling temperature and cut cooking time. It introduced the safety valve and demonstrated atmospheric pressure’s mechanical action — directly enabling Papin’s 1690 piston engine. It was never mass-produced, never used outside demonstration, and had no domestic adoption. Its value is archival and conceptual: it is the first device to treat pressure as a controllable variable in thermal systems.
REAPER
REAPER isn’t leaner because it’s simpler—it’s leaner because it refuses to decide what your studio should be.
Sewing machine
Saint’s 1790 sewing machine is a footnote — not a foundation.
Thermal imaging camera
Thermal imaging doesn’t see through walls—it sees heat where eyes fail, and that narrow trick saves lives and money on its own terms.
Thermometer
Temperature wasn’t measured until someone decided it could be numbered — and that decision took 100 years, three failed designs, and one mercury breakthrough.
YAML
YAML trades parser reliability for human legibility—and wins only where humans edit more than machines validate.
Boston Dynamics
Legs beat wheels—but only if you ignore noise, autonomy, and cost.
Intuitive Surgical
Robotic surgery didn’t start with autonomy—it started with a console, a licence, and a long wait for FDA approval.
Automated guided vehicle
Barrett Electronics
The Barrett Electronics AGV established guided vehicle automation as a physical layer—not a cognitive one. It moved payloads predictably, cheaply, and without rails—but demanded infrastructure, not intelligence. Its legacy is in floor plans, not algorithms.
General Atomics MQ-1 Predator
It did not invent drone warfare—but it industrialised the kill chain.
Unimate
The first industrial robot didn’t think—it repeated, corrected, and kept workers out of fumes.