AI agent
AI agents promise autonomy — but deliver orchestration dressed as agency.
Algorithms
Procedures that decide things, increasingly without being asked.
- 29
- in technology
- 9:54
- average
- 287 min
- in total
- 114
- across the network
All briefs
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.
Machine learning
Machine learning is not intelligence — it is applied statistics with memory.
Search engine
Search engines don’t find truth — they map attention, then sell the map.
AltaVista
AltaVista didn’t make search useful — it made search possible, then left relevance unsolved.
Antivirus software
Antivirus software is necessary — and mathematically guaranteed to fail at its core promise.
Apache Spark
Spark replaces MapReduce’s disk-bound rigidity with memory-resident flexibility—so long as your cluster has enough RAM.
Arista Networks
Arista didn’t reinvent networking—it rebuilt the OS so failures stop spreading.
Backpropagation
Backpropagation is not learning—it’s bookkeeping for learning.
Computational photography
Computation replaces glass—but only where the maths is exact and the optics are weak.
Content delivery network
CDNs don’t speed up the internet — they relocate its bottlenecks.
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.
Gaussian splatting
Gaussian splatting ditches ray-marching for splatting—and gains speed by abandoning geometry.
Generative adversarial network
GANs don’t model reality—they train a mimic to beat a critic.
LoRA (machine learning)
LoRA doesn’t shrink models—it shrinks the cost of changing them.
Neo4j
Neo4j Inc. · 2010
Neo4j is a graph database built for relationship-heavy workloads. It delivers expressive querying and transactional safety—but horizontal scaling, vector search, and unified standards arrived over a decade after launch. Its value is real, but bounded by what it ships—not what it promises.
PageRank
PageRank didn’t rank pages — it ranked links, and in doing so, rewired how machines read trust.
Redis
Salvatore Sanfilippo · 2009
Redis is an in-memory key-value database introduced in 2009 by Salvatore Sanfilippo to solve scalability issues at his startup LLOOGG. It delivers low-latency operations by holding all data in memory and using the fork system call to enable concurrent client service and disk persistence. It supports two persistence methods—RDB snapshotting and AOF journaling—with journaling introduced in version 1.1 and considered safer. Default configuration limits data loss to a few seconds upon full system failure. Redis supports master-replica replication with hierarchical trees and introduced clustering in April 2015 with version 3.0. It grew into one of the world’s most popular databases for caching, message brokering, and primary storage. In December 2024, Sanfilippo returned to Redis and implemented the Vector Set data structure for vector similarity search.
Reinforcement learning from human feedback
RLHF outsources ethics to annotators—and then optimises for their rankings, not their intent.
Texture mapping
It didn’t make 3D look real—it made 3D look cheap, fast, and editable.
Vector database
It doesn’t understand meaning — it just maps distances between precomputed guesses.
ALGOL
ALGOL didn’t run programs—it rewrote how we think about them.
Google Search
Google Search didn’t organise the web—it redefined authority as a function of linkage, then language, then generation.
High Efficiency Video Coding
HEVC cuts bandwidth—but doubles your compute bill and adds patent risk.
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.
JPEG
Joint Photographic Experts Group
JPEG is a 1992 DCT-based lossy compression standard. It delivers predictable 10:1 size reduction with acceptable visual trade-offs. It defines only the codec—not the file format. Its lossless mode is unused. It remains dominant not by superiority but by inertia.
MP3
MP3 is not a codec—it’s a decoder contract with a thousand incompatible encoders.
RSA cryptosystem
Ron Rivest, Adi Shamir and Leonard Adleman · 1977
RSA is the first publicly described, usable public-key cryptosystem. It works by exploiting the asymmetry between modular exponentiation and root extraction modulo a composite number whose factorisation is secret. Its security depends on factoring difficulty—but that dependence is not proven equivalent. No published attack breaks it with large keys. It changed cryptography by enabling open protocols, trustless key exchange, and standardised role names like Alice and Bob.