One Job: Six Tools That Do a Single Thing Perfectly

The JSON That Fought Back

Here is a small, specific frustration that anyone who works with data has felt. You have a JSON file. It is full of the answer you need. And you cannot grep it. Grep, the old reliable search tool, works on lines. JSON does not really have lines in any meaningful sense. It has nesting, brackets inside brackets inside brackets, and the thing you are looking for is four levels deep with no flat path to it. So you open the file, you scroll, you count braces with your finger like a child learning to read.

A British developer who goes by the handle tomnomnom, real name Tom Hudson, looked at that exact annoyance and built a tool with one job. It is called gron, and what it does is almost stupidly simple. It takes JSON and turns it inside out into flat, boring, fully-qualified lines. Every single value becomes its own line, with the complete path spelled out, like json dot contact dot email equals a string. Suddenly the thing is greppable. You search it like any other text. And when you have found what you wanted, gron runs backward, ungron, and reassembles the survivors into real JSON again.

That is the whole program. Nobody stores data in gron format. Nobody builds a system on it. It is a verb you apply to a noun for about thirty seconds in the middle of solving something else. It is the perfect example of the category this episode is about. Six tools that would be absurd as a general choice, and unbeatable at the one tiny thing they were carved to do.

Numbers, Drawn in the Terminal

In twenty eleven, a developer named Zach Holman sat down on a Monday to mess around. By the end of the week the thing he had built had a thousand watchers on GitHub and five hundred upvotes on Hacker News. It was called spark, and it is one of the smallest useful programs you will ever meet. It is a shell script. You hand it a list of numbers. It draws them as a tiny bar chart made of Unicode block characters, the kind that fit on a single line, right there in your terminal. One, five, twenty-two, thirteen, fifty-three becomes a little skyline of stacked blocks.

That is everything spark does. It has no precision mode, no axes, no labels, no export. Holman himself said part of why it spread was that he wrote it in shell instead of in a real programming language, because shell is the common glue that everyone, in every community, already has. The restraint was the feature. You would never reach for spark to make a chart for a report or a client. But to glance at the shape of a thousand numbers streaming past, to see at a glance whether a trend is climbing or collapsing, nothing lighter exists. It is a graph that costs you nothing.

The Oracle of Metadata

Every file on your computer is quietly carrying a second, hidden file inside it. A photograph knows the camera that took it, the lens, the shutter speed, often the exact spot on Earth where you stood. A PDF knows the software that made it and sometimes the name of the person who clicked save. A song file knows things you never told it. This hidden layer is metadata, and there is one piece of software that understands it better than anything else humans have written. It is called exiftool, and it has been maintained, almost single-handedly, by a Canadian physicist named Phil Harvey since the early two-thousands.

exiftool is written in Perl, which already tells you it comes from another era of the internet, and it reads and writes metadata for hundreds upon hundreds of file formats, including obscure proprietary camera formats that the camera manufacturers themselves barely document. When forensic investigators, journalists, museum archivists, and photographers need to know what a file is really saying about itself, this is the tool. It is not a photo editor. It is not a file manager. It does one strange, deep thing. It is the closest thing the file system has to a confessional, and Phil Harvey is the only priest.

From Pixels to Curves

You have a logo. It is a black and white image, a grid of pixels, and you need it as a clean vector, the kind of smooth mathematical outline that scales to the side of a building without going blurry. The general answer involves opening expensive design software and tracing it by hand, or fighting with a menu full of options. The specific answer is a tool by a mathematician named Peter Selinger, called potrace.

potrace does exactly one transformation. It takes a bitmap, a field of black and white dots, and it finds the smooth curves hiding inside that staircase of pixels. It hands you back crisp vector paths. That is the entire surface area of the program. It does not do color. It does not do photographs. It does not edit anything. It traces. But the curve-fitting mathematics inside it are good enough that potrace quietly sits underneath other, larger programs, doing the tracing step while the bigger tool takes the credit. The narrow tool became a load-bearing part of the wide ones.

The Shell That Survives the Tunnel

If you have ever connected to a remote computer over SSH, you know the small daily deaths of it. You shut your laptop and the connection is gone. Your train goes into a tunnel and the session hangs forever, frozen, until you give up and kill it. Your wifi switches networks and SSH simply does not understand that you are still the same person. SSH was designed for a world of stable, wired connections, and that world is mostly gone.

In twenty twelve, a group at the Massachusetts Institute of Technology released mosh, short for mobile shell. It does one thing SSH does badly. It survives. You can close your laptop for three hours, open it on a different continent on a different network, and your session is simply still there, as if nothing happened. It does this with a clever trick, predicting and showing your keystrokes locally instead of waiting for the distant server to echo them back, so the connection feels instant even on a terrible link. mosh is not a replacement for everything SSH does. It is a replacement for the one part that hurts. For a laptop that moves, on networks that flicker, it is the tool.

Every Unit, Into Every Other Unit

The last one is a small piece of the GNU project, and it has been quietly correct since long before most software you use was written. It is called units, and you can probably guess its single job. It converts any unit of measurement into any other unit of measurement. Not just the common ones. You can ask it how many teaspoons are in a cubic mile. You can ask it how your speed in furlongs per fortnight compares to the speed of light. You can ask it genuinely useful things too, the cooking conversions and currency rates and engineering quantities, but the joy of it is that it does not judge the question.

units carries a database of thousands of definitions, including physical constants and historical and regional measures, all related back to fundamental quantities so it can chain conversions you would never think to combine. It is interactive. You type the thing you have, you type the thing you want, it gives you the number. It has no graphical interface, no app, no cloud account. It is a calculator that happens to know what everything means. You would never build a product around it. But the next time you need to know something measured in one strange unit expressed in another strange unit, there is exactly one right answer, and it has been sitting in the GNU project for decades, waiting.

The Pattern

Look at the six. gron, spark, exiftool, potrace, mosh, units. Not one of them is a platform. Not one of them wants to be your whole workflow. Each one picks a single, sharply defined task, often a task most people did not even consciously notice was a task, and then goes deeper on that one task than any general tool ever could afford to.

There is a real lesson hiding in here, underneath the fun. A tool that tries to do everything has to spend its design budget on breadth. It has to be acceptable at a hundred things. A tool that does one job spends the entire budget on depth. That is why potrace's curves are so good, why exiftool knows formats the manufacturers forgot, why units never gets a conversion wrong. The narrowness is not a limitation they are apologizing for. The narrowness is the whole reason they are excellent.

You do not build your house out of these. You keep them in a drawer. And on the one day a year you need exactly this, you open the drawer, and the perfect thing is in there, waiting, having thought about nothing else the entire time.