#ai-assisted

Three blocks of addresses were quietly set aside in 1994 with no explanation for the numbers, an opposition RFC titled 'Considered Harmful' predicted decades of pain, and the lead objector ended up co-signing the winning standard. Your router's default address is the armistice line.

The squiggly letters guarding every signup form are a 1950 thought experiment turned inside out: a human laboring to convince a machine it is not one. The acronym says exactly that, and the words it made you type were quietly put to work.

The placeholder in half of all code examples traces back through a WWII vulgarity, a 1959 model railroad club dictionary, and a 1930s comic strip. The internet's standards body eventually gave it an official etymology, published on April 1st.

The word behind every restart describes something that cannot be done: lifting yourself off the ground by your own straps. Engineers in the 1950s chose the idiom precisely because it names a real paradox at the bottom of every machine, and the paradox is still there.

In a vocabulary built from puns, insults and physics jokes, the Unix scheduler is named with a straight face after the Greek word for time. The story is in what happened next: a daemon that checked its watch so eagerly it had to be taught to sleep.

The first tool anyone reaches for when the network dies was written in one evening in 1983 by a man thinking about submarines. The acronym everyone quotes for it was invented by someone else, after the fact, and the author's response to it was better than either story.

The web's most famous number has a beloved origin story involving an office at CERN where the first server lived. The web's own pioneers call the story hogwash. The real answer is duller, older, and better: the digits are a sentence, and you can read them.

When a program dies badly, the file it leaves behind is named after hardware that has been extinct for fifty years: little ferrite rings threaded by hand onto wire grids, by the millions, mostly by women whose work your crash reports still commemorate.

Everyone types it, few know the name drifted. The command was born around 1980 at SUNY Buffalo meaning one thing, the official project now spells it another way, and the man pages quietly rewrote what the first two letters mean. The story of the most-typed safety mechanism in computing.

One man reserved it in 1981 with a single table row and his own initials as the only justification. The reason it was 127 and not anything else was never written down, and the address block it sits in is large enough to give every person in the Netherlands their own loopback.

The default shell of the Linux world is a forty-year-old joke about a man named Bourne. The rest of the shell family tree is no more serious: one shell is named after a teaching assistant's login, one after a fish, and two of them are both called sh.

The industry is spending 2.59 trillion dollars a year to etch gen-AI into physical silicon, on the bet that this paradigm is the durable one. The paying base is about 50 million people, the productivity that would pay for it isn't measurable, and no software paradigm has ever been the forever thing.

Before it filtered packets, a firewall was a literal wall — a fire-resistant barrier in buildings from the 1850s, then the iron bulkhead behind a vehicle's engine. The networking sense borrowed the metaphor in the late 1980s and was cemented by Cheswick and Bellovin's work at Bell Labs.

Matt Holt started Caddy in 2014 as a computer-science student and released it in 2015. The name carries the golf-caddy idea — a helper that handles the tedious parts of serving the web — and the server went on to be the first to do HTTPS automatically by default.

A network port is a numbered endpoint that lets one host run many services at once. The word traces from the Latin for a gate or harbour, through the general computing sense of a connection point, into the TCP/IP port formalised by Jon Postel and Vint Cerf in the early ARPANET RFCs.

A homelab node pinned its memory and my portfolio pod climbed to 20 GB of RAM in a day. My first thought was an attack. It wasn't — it was two of my own caches failing silently, and normal traffic doing the rest. A war story about telling a leak from a breach.

Kafka was never built to be queried. Apache Fluss bolts a columnar, Arrow-native hot store onto the streaming layer and tiers cold to Iceberg — a clean full-stack realtime design whose only real open question is governance.

A project survives not when its company thrives but when its community can't be captured. The history — MySQL, Redis, Terraform, OpenOffice — says the right to fork is the load-bearing property that keeps open source alive.

The crash erased ~97% of the US video game market in two years. Players hadn't stopped wanting games — Atari lost control of who could publish, the shelves filled with junk, and trust collapsed. Nintendo won by re-imposing the gate.

Sierra's catalog — Half-Life, Homeworld, Tribes, plus the adventure canon — reads like a modern major's. The brand is a dormant trademark inside Microsoft. Two distinct forces explain the gap: a financial collapse and a market shift, routinely mistaken for one story.

"Catalog" means two different things in the lakehouse: the technical catalog in your query path (Unity, Polaris) and the governance catalog beside it (OpenMetadata, DataHub). The second is where lineage, ownership, and trust live — and where the next fight is.

A WireGuard tunnel to a Hetzner VPS sent 670 KiB and received nothing. The firewall had an ALLOW rule for port 51820. Both facts were true — because the rule said TCP and WireGuard is UDP. A debugging war story, and the method that found it: bisect the packet path.

Grade an LLM against the classic linguistic layers — syntax, semantics, pragmatics, social — and the failure lands in one precise place: not meaning, but truth. Hallucination is semantically well-formed falsehood, and the layers that work are what makes the one that doesn't hard to spot.

TTY is short for teletypewriter — a 1920s electromechanical printing telegraph that became the first computer terminal. The machine is gone, but the name survives in /dev/tty, getty, stty, SIGHUP, and a whole generation of terminal emulators that pun on it — kitty, Ghostty, PuTTY, Alacritty.

Grace Hopper's 1947 moth in the Harvard Mark II is real, and it is in the Smithsonian. But the word "bug" for an engineering fault is decades older — Edison used it in 1878 — which is exactly why the logbook entry is a joke.

The Unix daemon was named at MIT around 1963 after Maxwell's demon — a tireless background sorter from thermodynamics — not after any religious devil. The Greek daimon meant a benevolent guiding spirit, and "Disk And Execution Monitor" is a backronym invented decades later.

Linus Torvalds wrote Git in about two weeks in April 2005 after the BitKeeper licensing fallout, and named it after himself. "Git" is British slang for an unpleasant person — and the man page lists several joking backronyms for when it works or breaks.

The Apache web server has two competing origin stories — "a patchy server" built from NCSA httpd patches in 1995, and respect for the Apache Native American nation. The project's own documentation has told both, and which one is "real" has become a live question.

A Docker image is not a picture. The word traces through ISO files and CD-ROMs back to the 1960s, when an 'image' meant a faithful, byte-for-byte copy of storage — and the 'ROM' in CD-ROM is the same read-only idea Docker's immutable layers are built on.

Guido van Rossum named Python in December 1989 after Monty Python's Flying Circus, not the reptile. The snake logo is a 2006 retrofit, the docs say spam and eggs instead of foo and bar, and the BDFL title started as a joke in a 1995 mailing-list thread.

Karpathy coined "vibe coding" in February 2025 as a fun way to throw away weekend projects. The 2026 audits show what happens when the workflow leaks into production — the debt is real, it is comprehension debt, and it was deferred rather than erased.

Spam means junk because of a 1970 Monty Python sketch in which Vikings chant "SPAM" until it drowns out all conversation. The term moved through MUDs and Usenet to email, and Hormel — which trademarked SPAM in 1937 — eventually made peace with the lowercase generic use.

Java started as "Oak," named after a tree outside James Gosling's window at Sun. A trademark conflict forced a rename in 1995, and the replacement was picked over coffee — Java, the Indonesian coffee the team drank by the gallon. The steaming-cup logo made the accident official.

JavaScript was written in ten days in May 1995, shipped as LiveScript, and renamed JavaScript in December as a marketing co-branding deal with Sun. The name is a trademark, the language is unrelated to Java, and the standard had to be called something else entirely.

Prompt engineering was the 2023 job that mostly meant typing nicely at a chatbot. In 2025 Karpathy and Shopify's CEO renamed the real version of it — context engineering — and the rename was a promotion.

Three protocols showed up to standardise agentic AI — MCP for agent-to-tool, A2A for agent-to-agent, ACP for lightweight agent messaging. One of them is already dead.

'Agentic' was the word of the year, and Gartner reckons only about 130 of the thousands of vendors selling agents are shipping one. Here is the precise definition — a loop with tools, planning, state, and retry — the line that separates a real agent from a single LLM call wearing a cost...

The Model Context Protocol went from an Anthropic side-project in November 2024 to 97 million monthly SDK downloads by March 2026. Here is what it actually is — JSON-RPC, three primitives, a client-server split — why it won by turning M×N integrations into M+N, and why just a protocol...

Bluetooth is named after a 10th-century Danish king, Harald "Bluetooth" Gormsson, who united warring tribes — the analogy Jim Kardach drew when he proposed it as a codename in 1996. The logo is a bind-rune of the king's initials. The codename was never meant to ship.

The real engineering discipline behind shipping reliable LLM features is not prompting — it is evals. They are to LLM apps what unit tests are to deterministic code, except you measure pass-rates and distributions instead of exact equality.

Most companies' data is not big enough to justify a distributed warehouse. A single fat box running DuckDB reads Parquet and Iceberg off S3 directly and answers the median analytics query in under a second, for a fixed bill and no cold start. The big-data era was mostly oversizing.

Iceberg and Delta reimplemented a transactional catalog as JSON and Avro files in object storage — and then needed a real database catalog on top anyway. DuckLake's heresy is to skip the file layer entirely: put all the metadata in SQL, keep the data in Parquet. It is both obvious and a little rude.

Iceberg did not win on features. Delta Lake had the bigger installed base and Hudi had the better write path. Iceberg won on governance and an engine-neutral spec, and the moment Databricks paid roughly $2B for Tabular the war was effectively over.

"Zero-ETL" is the data industry's most successful rebrand of the year — and the T never went anywhere. The transformation work does not vanish, it relocates to query time or to the catalog. What is genuinely new is the collapse in data duplication, and that part is worth the hype.

Lou Montulli invented the web cookie at Netscape in June 1994 to give a stateless protocol a memory. The name is borrowed from the Unix "magic cookie" — an opaque token a program hands around — and the privacy story started almost immediately.

The cloud was a network-diagram icon for two decades before it was a product. Compaq coined cloud computing in 1996, AWS made it real in 2006, and the surrounding vocabulary — VPS, hyperscaler, colocation — each carries its own history.

vim came from ed (1969) through ex and vi. Emacs came from a different lineage starting at MIT in 1976. The thirty-year fight between them was a disagreement about whether modal editing is worth the learning curve. Nano is the third option that quietly took the casual market.

K8s is a numeronym — K, eight letters, s. Kubernetes itself is the ancient Greek κυβερνήτης, the helmsman of a ship. The seven-spoked logo is a Star Trek easter egg. The thing it names is the third generation of Google's internal scheduler.

A docker is a longshoreman — the worker at a port who loads and unloads shipping containers. The software took its name from the work; the work took its metaphor from the 1956 standardisation of physical shipping containers. The container ecosystem inherited the maritime vocabulary.

The .yml extension is a 1990s DOS artifact. The 'YAML Ain't Markup Language' acronym is a 2002 self-correction. Both questions resolve cleanly once you know markup languages and data serialisation formats are different categories with different ancestors.

A serverless warehouse that auto-provisions three nodes for a SELECT * spends most of the bill on the time you waited for it. A self-hosted ClickHouse on EC2 trades elasticity for sub-second latency and a fixed monthly line item — favourable any time queries are even mildly steady.

A lakehouse on object storage looks cheap because storage is cheap. The bill is built from request count and managed-tier access fees, both of which scale with file count, not data volume. 5 GB stored as one million 5 MB files is a different invoice than 5 GB stored as ten 512 MB files.

Douglas Crockford has said for twenty years that he did not invent JSON, he discovered it. The format was sitting inside JavaScript the whole time, waiting for someone to extract it. The story of how a 2001 footnote in a browser scripting language ate XML's lunch is shorter than most people think.

POSIX stands for Portable Operating System Interface, plus an X added at the IEEE's request. Richard Stallman coined the name in 1988. The standard codifies the Unix interface every modern operating system claims to implement and almost none follow strictly.

Tatu Ylönen wrote SSH at Helsinki University of Technology in July 1995 after a password-sniffing incident. Thirty years later it is the default remote-access protocol on every Unix-like operating system, including Windows. Almost nothing else in security tooling has lasted as long unchanged.

grep is short for g/re/p — the ed command syntax for global regular expression print. Regular expressions themselves go back to a 1951 RAND memo by Stephen Kleene. The thirty-year flavour war is a footnote to the original math.

Two pronunciations of SQL have coexisted for nearly fifty years. Why the language was renamed in the first place, and what each pronunciation preserves, turns out to be more interesting than the debate that surrounds it.

A short tour of database name origins. Children, code names, predecessors, acronyms, mythology, and at least one pet duck on a houseboat. Most of the history of the field is hidden inside the names themselves.

A 2026 study found humans are quietly being gentler on AI-authored PRs than on human-authored ones. Single-reviewer review already broke at 400 lines per diff. The fix is two old ideas — distribute the review across the team, and pair every diff with a deterministic map of the code flow.

Brooks said adding people to a late project makes it later. Adding LLMs follows a similar pattern, except the cost lands on a different axis — output rises while shared understanding of the system erodes, and the resulting cognitive debt eventually has to be paid.

A small series of opinionated hot takes on AI engineering, drafted with LLM assistance and verified by hand. The index post.

Apple's "do not hallucinate" prompt is funny. The same logic is why AI-written + AI-reviewed pipelines silently approve bugs, why long contexts compound errors, and why the real guardrails are deterministic.

Cursor, Kiro, Windsurf, Antigravity, Trae. All forks of VS Code, all wrapping one of three model APIs, all selling a long system prompt that does not move the model's ceiling. The vibe-coding tech-debt numbers were always pointing somewhere else.