Some Languages Die. Others Simply Refuse.
In 2020, Adobe ended Flash Player support on December 31st. In the weeks that followed, 98% of ActionScript job postings vanished from the internet. The language did not decline gradually. It ceased. There was no long tail, no graceful migration, no retirement party. A runtime decision by a California software company deleted an entire programming ecosystem — the only language in history where we know the precise calendar date of its death warrant.
But ActionScript's extinction is the exception, not the rule. Most programming language deaths are geological: long, slow, almost imperceptible until they are undeniable. COBOL has been "dying" since 1992. Perl's obituary has been written at least a dozen times in this century. Fortran outlived the programmers who first wrote it. These languages persist because software is infrastructure, and infrastructure does not retire — it accumulates.
The Programming Language Mass Extinction Index scores 30 languages across six dimensions of survival risk. What emerges is a picture that confounds most intuitions about which languages are dying and which are merely dormant. The most surprising finding: the language developers call "dying" most often is not even close to the top of the list.
The index measures six things: TIOBE search presence (a proxy for how many developers are looking for help with a language), Stack Overflow activity (what developers are actually doing), GitHub repository creation (what people are building), job market demand (what employers will pay for), ecosystem health (whether the package registry is alive or calcifying), and community vitality (Reddit, conferences, organized activity).
Higher score means higher extinction risk. A score of 100 means the language has already effectively ceased to exist as a living development environment. A score of 2 means the language is so embedded in the global infrastructure that discussing its extinction is almost absurd.
The chart below shows the fifteen most endangered languages in the dataset. Look at what's near the top. Then look at rank ten.
The Language That Won't Die: COBOL at #10
Every year, technology writers declare COBOL dead. Every year, banks quietly post another 1,400 open positions — paying an average of $92,000 annually, with 30% of COBOL developers earning over $100,000. COBOL runs an estimated $3 trillion in daily financial transactions. IBM has active apprenticeship programs training a new generation of COBOL engineers. The language has almost no GitHub activity and no modern community — but it has institutional oxygen tanks with infinite refill. COBOL is not dying. It is becoming rarer and therefore more valuable, like a trade skill nobody trains for anymore.
The data resolves into four groups that tell four different stories about language death. They are not simply ordered by score — they represent fundamentally different modes of decline and survival.
The Fossils — Already Gone (Score 75–100)
The Fossils share one defining trait: no engineering team is starting a greenfield project in any of these languages in 2026. The work that exists is maintenance, not creation. Note that this group contains both ActionScript (genuinely extinct) and Assembly (genuinely immortal in hardware) — which illustrates how the index captures different failure modes rather than a single axis of "deadness."
The Haskell entry deserves its own paragraph. By every market metric, Haskell should be extinct. It ranks #42 on TIOBE, appears in only 2% of Stack Overflow developer surveys, and produces perhaps 7,000 new GitHub repositories annually. But its Reddit community has 73,900 subscribers. ZuriHac attracts hundreds of contributors every summer. The Haskell Foundation is funded and active. Haskell is the only language in this dataset that is simultaneously near-extinct in the market and culturally thriving. It is a language kept alive by love, not economics.
The Fading Stars — Contracting to Their Niches (Score 55–74)
These languages are not dying. They are contracting toward their permanent niches. Swift will always exist while iPhones exist. Go will always exist while Kubernetes exists. Rust is growing, not shrinking — its score reflects that it is still tiny in absolute developer numbers despite its reputation. The question for this group is not survival — it is whether the niche is large enough to sustain an active ecosystem for another generation of developers.
The Undead — Declared Dead, Still Answering Emails (Score 35–54)
PHP has been declared dead since 2010. It runs WordPress, which runs 43% of the internet. C is sixty years old and still in the Linux kernel. These languages were written for machines, not developers, and machines do not care about developer experience.
The Immortals — Beyond Extinction Risk (Score 0–34)
JavaScript scores 2.0 — essentially zero extinction risk. It runs in every browser on the planet, every Node.js server, every React Native app. TypeScript's score of 23.2 understates its momentum: its TIOBE rank (#35) is dramatically low because most TypeScript-related searches are captured under "JavaScript" in TIOBE's methodology. On GitHub repo creation and Stack Overflow activity, TypeScript is a top-3 language. The index captures this through the non-TIOBE metrics, but the bump chart below tells the story most clearly.
Where Does Your Language Land?
The most important chart in this dataset is not the bar chart. It is the chart that shows how languages change rank depending on which metric you use to evaluate them.
"The decline was not sudden. It was geological. Perl ranked #12 on TIOBE in January, as it has for years. Nobody noticed that GitHub had quietly stopped generating new Perl repositories at any meaningful rate."
The bump chart below shows each language's rank across five independent metrics: TIOBE presence, Stack Overflow activity, GitHub repository creation, job market demand, and community size. The X-axis is not time — it is measurement perspective. A line that crashes from left to right means a language looks healthy on one metric and moribund on another. The crossings are the story.
Watch Perl's line. It enters from the left at a respectable rank (TIOBE has it in the top 12) and falls sharply as you move right toward jobs and community. Watch Visual Basic do the same in reverse — it has an inflated TIOBE rank driven by legacy documentation searches, but its GitHub and job presence are near-zero. Watch COBOL: it enters modestly on TIOBE but rises sharply in job demand, the only language in the Fossils tier that does this.
What the extinction index teaches is that language death is almost never about age. Python is older than Lua. Java is older than Ruby. C predates most of the people who write in it. Age is not the variable. The variable is whether the next generation of developers decides to learn you.
COBOL failed that test — no new developer chooses COBOL for a side project or a startup. But COBOL passed a different test: the institutions that run on it cannot afford to replace it. That is a different kind of immortality than Python's or JavaScript's. It is reluctant immortality, the immortality of a system too expensive to decommission.
ActionScript passed neither test. When Adobe decided that Flash was over, it made the decision for everyone. There was no community resilient enough to fork it, no institution dependent enough to maintain it. The language that powered a generation of internet creativity disappeared in the span of a quarterly report.
The lesson of the extinction index is not that some languages are better. It is that some languages are load-bearing walls, and others are decorative arches. The walls survive. The arches may be more beautiful, but when the building is renovated, they are the first to go.
What the numbers don't show is the hundreds of thousands of developers who learned Flash, built careers in ActionScript, made things that mattered — and who then had to start over. The index measures survival. It cannot measure the cost of the extinction events it catalogs.