Coding means writing instructions that a computer can follow. Many people wonder who decides what those instructions look like, and whether computers are stuck with 0s and 1s forever. This article answers both questions in plain English, using the latest facts we could verify.
Short answer
- Code is written in layers. People write readable text, and programs called compilers or interpreters translate it into machine code, the instructions a chip understands [1].
- At the very bottom are switches. Chips are made of transistors (tiny electronic switches) that are either on or off, and those two states are labelled 1 and 0 [2].
- The chip’s basic vocabulary is set by its designer or an open standard. This vocabulary is called the instruction set [3]. One open version, RISC-V, is free to use [4].
- Language rules are set by the creator first, then often by a group. Python is run by an elected council [5], JavaScript by a company-backed committee [6], and C by an international standards body [7].
- Anyone can create a new language. Rust began as one person’s side project in 2006 and reached version 1.0 in 2015 [8]. A new language’s first translator is written in an older language [9].
- 0 and 1 are a design choice, not a law of nature. A Soviet computer used three-state digits in 1958 [10], memory chips use 8 or 16 voltage levels per cell [11], quantum bits can be 0, 1, or a mix [12], and AI researchers use three-value numbers [13].
- Binary still dominates. Two clearly different states are the easiest to build and read reliably [14].
- Any set of signs can be used if people agree on the meaning. Unicode gives every character a number, such as 65 for the letter A [15], and reached 172,808 characters in September 2026 [16].
The bottom layer: switches that are on or off
Inside every chip are transistors. A transistor works like a light switch: if current flows the circuit is on, and if not it is off [2]. These two states correspond to the binary 1s and 0s used in digital computers [2].
A single chip can hold billions of these switches [2]. Everything a computer does, from showing a photo to sending a message, is built from patterns of switches turning on and off.
The chip’s own language: the instruction set
Each chip understands a fixed list of basic commands, such as “add these two numbers.” That list is called an instruction set architecture, or ISA [3]. It specifies the instructions the chip runs, the kinds of data it handles, and the small storage spots (registers) it uses while working [3]. Think of it as the vocabulary a chip speaks.
Usually the company that designs the chip decides this vocabulary. For example, most customers of the chip designer Arm receive ready-made design blocks and have little freedom to change the inside [3].
RISC-V takes a different path. It is an open standard instruction set looked after by a non-profit called RISC-V International, and there is no fee to use the ISA [4]. An open standard is open to anyone, like the shipping container standard ISO 668:2020, which lets countries build containers that fit each other [4].
From human words to machine code
Nobody types raw 0s and 1s. A compiler is software that translates code written in a high-level language such as C++ into machine-language instructions the processor can understand [1]. Think of a human translator who turns a letter from English into another language before it is posted.
Some compilers go through a middle step called assembly language, which gives machine instructions readable names, and an assembler then turns that into machine code. Other compilers produce machine code directly [17]. Compilers, assemblers, and interpreters are all kinds of translators that convert code from one language into another [18].
Who sets the rules of a programming language?
No single world authority decides. The rules (what words mean and how sentences must be built) belong to whoever looks after the language. Three common models are:
- A creator, then a community (Python). Its creator, Guido van Rossum, held the final say as “benevolent dictator for life” until he resigned from that role in 2018. The community then elected a five-person steering council under a rule document called PEP 13 [5].
- A committee backed by companies (JavaScript). The language standard is called ECMAScript, and it is developed by Technical Committee 39 (TC39) of Ecma International [6]. New ideas move through stages from 0 to 4, and the committee must agree before they advance [6]. Members meet about every two months, and the 116th meeting was held in Tokyo from 29 September to 1 October 2026 [6]. Delegates come from member companies such as Google, Apple, and Mozilla [6].
- An international standards body (C). The rules of C are published as ISO/IEC 9899. Its latest edition, C23, was adopted by ISO and IEC in 2024 and is maintained by a working group called WG14 [7].
What about the “functions” you use, such as print or sort? They come from three places: the language itself, libraries (bundles of ready-made code, some shipped with the language and some shared by other programmers), and code you write yourself. A function only works because the compiler or interpreter was built to follow its rule.
Can brand new languages be created from scratch?
Yes. A new language has a chicken-and-egg problem: to run it, you need a translator, and that translator must be written in a language that already exists [9]. This is called bootstrapping [9]. For example, Niklaus Wirth wrote the first Pascal compiler in the older language Fortran [9]. Once the new language works, its compiler can later be rewritten in the language itself [9].
Rust is a modern example. Graydon Hoare began it as a personal project at Mozilla in 2006, Mozilla started sponsoring it in 2009, and Rust 1.0 was published in May 2015 [8]. Inventing a language is the easy part, and taking one to a stable release can take years.
The people doing this work are increasingly global. GitHub reported more than 36 million new developer accounts in the year to August 2025, and India added over 5.2 million of them, which is 14% of all new accounts and the largest single source of sign-ups [19]. GitHub’s 2024 report also named Brazil, India, and Nigeria as especially fast-growing developer communities [20]. (An account is not always a professional programmer, so treat these as signs of growth, not exact head counts.)
Does everything have to be 0s and 1s?
No. Binary is an engineering choice. A digital circuit is designed to accept only voltages of certain values, and a circuit that uses only two states is called a binary circuit [21]. Two states are popular because they are easy to tell apart, which makes circuits more tolerant of small electrical noise [14]. A decimal circuit would have to tell ten voltage levels apart [14].
Other approaches exist:
- Three states (ternary). The Setun computer, built in 1958 at Moscow State University, used balanced ternary, with digits of −1, 0, and +1 [10]. About 50 were made, and production ended in 1965 [22]. A replacement binary computer performed equally well but cost 2.5 times as much [22].
- More voltage levels inside memory. In flash memory, a cell with 8 voltage levels stores 3 bits and one with 16 levels stores 4 bits [11]. Each level stands for a pattern of bits, so the outside world still sees 0s and 1s [11].
- Quantum bits. A qubit can be in state 0, state 1, or a mix of the two, and measuring it forces a definite 0 or 1 [12]. Two qubits can hold a superposition of four combinations at once [12].
- Three-value numbers in AI. Microsoft Research and the University of Chinese Academy of Sciences introduced BitNet b1.58, where every weight (a number the model learns) is one of three values: −1, 0, or 1 [13]. In 2025, Microsoft released a 2-billion-parameter model trained this way from scratch, and its benchmarks showed results comparable to similar-sized full-precision models [23]. It ships with software to run on ordinary CPUs [23].
Can a code use 2s, 3s, or different signs?
Yes, as long as people agree on what the signs mean. Computers already do this with text. Unicode gives every character a unique number called a code point [15]. The letter A is U+0041, which is 65 in everyday numbers [15].
In September 2026, Unicode 18.0 added 13,007 new characters, bringing the total to 172,808 [16]. Earlier, Unicode 17.0 (2025) added scripts used by Zaghawa communities in central Africa, Kurukh communities in northeast India, and Tai Yo communities in northern Vietnam [24].
Myths vs facts
| Claim | Verdict | What the evidence says |
|---|---|---|
| Computers can only ever use 0s and 1s | Myth | Binary is a design choice. A ternary computer was built in 1958, and flash memory cells use 8 or 16 levels [10, 11]. |
| Almost all computers today are built on two-state switches | True | Transistors act as on/off switches, and two-state circuits are called binary circuits [2, 21]. |
| Programmers write code in 0s and 1s | Myth | Compilers translate readable code into machine code [1, 17]. |
| One company or person decides every programming language | Myth | Python has an elected council, JavaScript a committee, and C an ISO group [5, 6, 7]. |
| Chip instruction sets are always owned by companies | Myth | RISC-V is an open standard with no fee to use [4]. |
| Anyone can invent a new programming language | True | Rust began as a personal project, and first compilers can be written in other languages [8, 9]. |
| Ternary computers have been proven better than binary ones | Unproven | Setun was cheaper in one comparison, but only about 50 were built and production ended in 1965 [22]. |
| AI models can work with just three values per weight | True | Microsoft’s BitNet b1.58 research shows this works at 2 billion parameters, based on the researchers’ own benchmarks [13, 23]. |
| A new set of signs needs a new kind of computer | Myth | Unicode assigns numbers to characters, and 17.0 added African and Asian scripts [15, 24]. |
Bottom line
Coding works as layers of translation, from readable text down to on/off switches in a chip. The rules at each layer are set by people: chip designers or open standards at the bottom, then creators, committees, and standards bodies for languages. New languages can be created by anyone, and 0s and 1s are a practical choice, not a law of nature. Three-state numbers, many-level memory, quantum bits, and any set of signs mapped to numbers all exist, but binary stays dominant because two states are the easiest to build and read reliably.
Disclaimer: This article is for general education only and does not replace professional advice from a qualified software engineer, electronics engineer, or computer science teacher. If you notice an error or would like to send feedback or corrections, please contact us.
References
- [1] compiler, Britannica
- [2] transistor, Britannica
- [3] RISC-V: An Open Instruction Set Architecture, DesignSpark (RS)
- [4] RISC-V International FAQ, RISC-V International
- [5] Report on the Python steering council election under PEP 13, LWN.net
- [6] TC39 meeting report: attending the committee that decides JavaScript’s language rules (Japanese-language article), Findy Media
- [7] ISO/IEC JTC1/SC22/WG14 (C) official home page, Open-std.org
- [8] Rust (programming language), Wikipedia
- [9] Bootstrapping (compilers), Wikipedia
- [10] Ternary computer, Wikipedia
- [11] Device, system, and method to verify data programming of a multi-level cell memory (patent text), USPTO
- [12] Quantum superposition and qubits, NIST
- [13] The Era of 1-bit LLMs: All Large Language Models are in 1.58 Bits, Microsoft Research
- [14] 2.3.2 Binary Data Format, TutorChase
- [15] What is a Unicode code point?, Unicode Fundamentals
- [16] Keyed in: UAE dirham symbol makes debut in Unicode 18.0, The National
- [17] compiler (print article), Britannica
- [18] Translator (computing), Wikipedia
- [19] GitHub’s 2025 Report Reveals Some Surprising Developer Trends, It’s FOSS
- [20] Octoverse: AI leads Python to top language as the number of global developers surges, GitHub Blog
- [21] integrated circuit, Britannica
- [22] Setun, Wikipedia
- [23] Microsoft Native 1-Bit LLM Could Bring Efficient genAI to Everyday CPUs, InfoQ
- [24] Unicode 17.0 Release Announcement, The Unicode Blog
