.NET Day 2026 The Past, Present and Future of Programming Languages

Transcript

1. The Past, Present and Future of Programming Languages Kevlin Henney
2. @kevlin.bsky.social @kevlin@mastodon.social kevlinhenney.medium.com sessionize.com/kevlin linkedin.com/in/kevlin instagram.com/kevlin.henney threads.net/@kevlin.henney x.com/KevlinHenney about.me/kevlin
3. The past, present and future walked into a bar. It was a tense situation.
4. © Lucasfilm
5. Impossible to see the future is. © Lucasfilm
8. Idaho State Correctional Institution 2021–2030
9. Seymour
10. He spends every hour he can inside the computer lab. He has learned Excel, AutoCAD, Java, C++, and Python , taking comfort in the clear logic of code, input and output, instruction and command.
11. has learned Excel
12. Java, C++, and Python
14. Programming languages appear to be in trouble.
15. Java C++ Python
16. Python C C++ Java C# JavaScript Visual Basic SQL R Rust
17. tiobe.com/tiobe-index
18. tiobe.com/tiobe-index
19. tiobe.com/tiobe-index
20. Python C C++ Java C# JavaScript Visual Basic SQL R Rust
21. C Java Python C++ C# TIOBE Index for September 2020
22. C Python Java C++ C# TIOBE Index for September 2021
23. Python C Java C++ C# TIOBE Index for September 2022
24. Python C C++ Java C# TIOBE Index for September 2023
25. Python C++ Java C C# TIOBE Index for September 2024
26. Python C++ C Java C# TIOBE Index for September 2025
27. Python C C++ Java C# TIOBE Index for August 2026
28. top_5 = {" Python", top_5 = {" C ", top_5 = {" C++ ", top_5 = {" Java ", top_5 = {" C# "} for ranking in permutations(top_5): print(ranking)
29. 1 2 3 4 5 6 7 8 9 10 Python C C++ Java C# JavaScript Visual Basic SQL R Rust 18.53% 11.10% 8.62% 8.25% 4.09% 2.63% 2.18% 1.88% 1.56% 1.45%
30. 1 2 3 4 5 6 7 8 9 10 Python C C++ Java C# JavaScript Visual Basic SQL R Rust ~60%
31. 1 2 3 4 5 Python C C++ Java C# ~51%
32. Python C C++ Java C# JavaScript Visual Basic SQL R Rust
33. Python C C++ Java C# JavaScript Visual Basic SQL R Rust
34. Python C C++ Java C# JavaScript Visual Basic SQL R Rust
35. Python C C++ Java C# JavaScript Visual Basic SQL R Rust Delphi/Object Pascal Scratch PHP Go Fortran Ruby Swift Perl COBOL Assembly language
36. Python C C++ Java C# JavaScript Visual Basic SQL R Rust Delphi/Object Pascal Scratch PHP Go Fortran Ruby Swift Perl COBOL
37. Python C C++ Java C# JavaScript Visual Basic SQL R Rust Delphi Scratch PHP Go Fortran Ruby Swift Perl COBOL
38. Python C C++ Java C# JavaScript Visual Basic SQL R Rust Delphi Scratch PHP Go Fortran Ruby Swift Perl COBOL
39. Python C C++ Java C# JavaScript Visual Basic SQL R Rust Delphi Scratch PHP Go Fortran Ruby Swift Perl COBOL
41. Change always comes later than we think it should. Jean-Luc Picard blendermarket.com/products/uss-enterprise-ncc-1701-d
42. redmonk.com/sogrady/2026/04/14/language-rankings-1-26
43. Movement in our top 20 is relatively rare. Shifts within our top 5 are even more so. redmonk.com/sogrady/2026/04/14/language-rankings-1-26
44. JavaScript Python Java PHP C# TypeScript CSS C++ Ruby C Swift Go R Shell Kotlin Scala PowerShell Dart Objective-C Rust
45. JavaScript Python Java PHP C# TypeScript CSS C++ Ruby C Swift Go R Bash Kotlin Scala PowerShell Dart Objective-C Rust
46. JavaScript Python Java PHP C# TypeScript CSS C++ Ruby C Swift Go R Bash Kotlin Scala PowerShell Dart Objective-C Rust
47. JavaScript Python Java PHP C# TypeScript CSS C++ Ruby C Swift Go R Bash Kotlin Scala PowerShell Dart Objective-C Rust
48. Anomalously low worth of pull requests from GitHub. It could be that while code creation is accelerating, the percentage of code committed to open repositories is declining. redmonk.com/sogrady/2026/04/14/language-rankings-1-26
49. Stack Overflow’s position and prominence in our rankings has also come into question. redmonk.com/sogrady/2026/04/14/language-rankings-1-26
50. redmonk.com/rstephens/2025/06/18/stackoverflow
51. From this you can see that questions peaked in 2016 and have been in decline since. It is clear that the usage dramatically decreased post 2023 when ChatGPT became widely available. The number of questions asked are now about 10% what they were at Stack Overflow’s peak. redmonk.com/rstephens/2025/06/18/stackoverflow
52. spectrum.ieee.org/top-programming-languages-2025
54. Python Java C++ SQL C# JavaScript TypeScript C Shell Go R PHP Kotlin Rust Dart Swift Ruby HTML Objective-C MATLAB
55. Python Java C++ SQL C# JavaScript TypeScript C Bash Go R PHP Kotlin Rust Dart Swift Ruby HTML Objective-C MATLAB
56. Python Java C++ SQL C# JavaScript TypeScript C Bash Go R PHP Kotlin Rust Dart Swift Ruby HTML Objective-C MATLAB
57. Python Java C++ SQL C# JavaScript TypeScript C Bash Go R PHP Kotlin Rust Dart Swift Ruby HTML Objective-C MATLAB
58. There are only two kinds of languages: The ones people complain about and the ones nobody uses. Bjarne Stroustrup
59. C is quirky, flawed, and an enormous success. Dennis M Ritchie
60. Ada Bash C C# C++ COBOL Dart Delphi Fortran Go Java JavaScript Kotlin MATLAB Objective-C Perl PHP PowerShell Python R Ruby Rust Scala Scratch SQL Swift TypeScript Visual Basic
61. Fortran COBOL C SQL MATLAB Ada Objective-C C++ Perl Bash Python Visual Basic R PHP Ruby Delphi JavaScript Java C# Scala PowerShell Scratch Go Rust Dart TypeScript Kotlin Swift 1950s 1970s 1980s 1990s 2000s 2010s
62. Fortran COBOL C SQL MATLAB Ada Objective-C C++ Perl Bash Python Visual Basic R PHP Ruby Delphi JavaScript Java C# Scala PowerShell Scratch Go Rust Dart TypeScript Kotlin Swift Baby Boomer Generation X Millennial Generation Z Generation Alpha
63. Fortran COBOL C SQL MATLAB Ada Objective-C C++ Perl Bash Python Visual Basic R PHP Ruby Delphi JavaScript Java C# Scala PowerShell Scratch Go Rust Dart TypeScript Kotlin Swift Baby Boomer Generation X Millennial Generation Z Generation Alpha
64. We could, of course, use any notation we want; do not laugh at notations; invent them, they are powerful. In fact, mathematics is, to a large extent, invention of better notations. Richard Feynman
65. In fact, programming is, to a large extent, invention of better notations.
66. While languages such as Python, C++, Rust, and even newer languages continue to introduce modern language features and programming paradigms, MATLAB can increasingly feel like a language from another era. tiobe.com/tiobe-index
67. Each successive language incorporates, with a little cleaning up, all the features of its predecessors plus a few more.
68. Each successive version incorporates, with a little cleaning up, all the features of its peers plus a few more.
70. für
71. for
73. else
75. One of the most powerful mechanisms for program structuring [...] is the block and procedure concept. Ole-Johan Dahl and C A R Hoare “Hierarchical Program Structures”
76. begin end
77. A procedure which is capable of giving rise to block instances which survive its call will be known as a class; and the instances will be known as objects of that class. Ole-Johan Dahl and C A R Hoare “Hierarchical Program Structures”
79. Simula 67
81. long int elif fi void bool union char short struct
82. 0 1 1 2 3 5 8 13 21 34 55 ⋮
83. F0 = 0 F1 = 1 Fn = Fn-1 + Fn-2
84. Are you going to this year’s Fibonnaci conference? It’s going to be as big as the last two put together.
85. proc fibonacci = (int n) int: if n < 2 then n else fibonacci(n - 1) + fibonacci(n - 2) fi
86. Coroutines
87. Coroutines are subroutines all at the same level, each acting as if it were the master program when in fact there is no master program. The coroutine notion can greatly simplify the conception of a program when its modules do not communicate with each other synchronously.
89. Iterators are a form of coroutine; however, their use is sufficiently constrained that they are implemented using just the program stack.
91. for n: int in fibonacci() do print(n) end
92. fibonacci = iter () yields (int) Fn_2: int := 0 yield (Fn_2) Fn_1: int := 1 yield (Fn_1) while true do Fn: int := Fn_1 + Fn_2 yield (Fn) Fn_2, Fn_1 := Fn_1, Fn end end fibonacci
93. IEnumerable<int> fibonacci() { int Fn_2 = 0; yield return Fn_2; int Fn_1 = 1; yield return Fn_1; while (true) { int Fn = Fn_1 + Fn_2; yield return Fn; (Fn_2, Fn_1) = (Fn_1, Fn); } }
94. function* fibonacci() { let Fn_2 = 0; yield Fn_2; let Fn_1 = 1; yield Fn_1; while (true) { let Fn = Fn_1 + Fn_2; yield Fn; [Fn_2, Fn_1] = [Fn_1, Fn]; } }
95. function* fibonacci(): Iterator<number> { let Fn_2: number = 0; yield Fn_2; let Fn_1: number = 1; yield Fn_1; while (true) { let Fn: number = Fn_1 + Fn_2; yield Fn; [Fn_2, Fn_1] = [Fn_1, Fn]; } }
96. Conventional programming languages are basically high level, complex versions of the von Neumann computer.
97. Von Neumann programming languages use variables to imitate the computer’s storage cells; control statements elaborate its jump and test instructions; and assignment statements imitate its fetching, storing, and arithmetic.
98. Ada Bash C C# C++ COBOL Dart Delphi Fortran Go Java JavaScript Kotlin MATLAB Objective-C Perl PHP PowerShell Python R Ruby Rust Scala Scratch SQL Swift TypeScript Visual Basic
99. Ada Bash C C# C++ COBOL Dart Delphi Fortran Go Java JavaScript Kotlin MATLAB Objective-C Perl PHP PowerShell Python R Ruby Rust Scala Scratch SQL Swift TypeScript Visual Basic
100. Ada Bash C C# C++ COBOL Dart Delphi Fortran Go Java JavaScript Kotlin MATLAB Objective-C Perl PHP PowerShell Python R Ruby Rust Scala Scratch SQL Swift TypeScript Visual Basic
101. Ada Bash C C# C++ COBOL Dart Delphi Fortran Go Java JavaScript Kotlin MATLAB Objective-C Perl PHP PowerShell Python R Ruby Rust Scala Scratch SQL Swift TypeScript Visual Basic
102. Excel
103. Excel is the world’s most popular functional language Simon Peyton-Jones
107. (define (fibonacci (lambda (n) (cond ((< n 2) n) (t (+ (fibonacci (- n 1)) (fibonacci (- n 2))))))))
108. (defun fibonacci (n) (cond ((< n 2) n) (t (+ (fibonacci (- n 1)) (fibonacci (- n 2))))))
109. (defun fibonacci (n) (if (< n 2) n (+ (fibonacci (- n 1)) (fibonacci (- n 2)))))
110. a i
111. Symbolic AI
112. Machine learning
113. Machine learning algorithm walks into a bar. Bartender asks, “What’ll you have?” Algorithm says, “What’s everyone else having?”
115. strawberry
116. strawberry
117. strawberry
118. [302, 1618, 19772]
124. AI’s ability to write code in a language is proportional to how much of that language it’s seen. AI has seen tons of JavaScript, Python, and TypeScript so it’s great at writing them. New languages are actually disadvantaged. Anders Hejlsberg github.blog/developer-skills/programming-languages-and-frameworks/typescripts-rise-in-the-ai-era-insights-from-lead-architect-anders-hejlsberg
126. Elon facebook.com/Rodenstock/posts/its-time-for-the-next-optical-illusion-how-many-legs-does-this-elephant-havewhoe/329176703106659
127. Things will move to where you don’t even bother doing coding. The AI just creates the binary directly. AI can create a much more efficient binary than can be done by any compiler. Elon Musk
128. Tell me you don’t understand something without telling me you don’t understand something.
129. something  { programming languages, LLMs, natural language, physics, compilers, hardware, VMs, security, platforms, optimisation, toolchains, ... }
130. English 3GL 2GL 1GL
131. NL 3GL 2GL 1GL
132. 1GL 011111110100010101001100010001100010000100010000 000000000000000000000000000000000001000000111110 000000010000000000000000000000000000000000000000 000000000000000000000000000000000000011100001111 111111111111111111111000111100000000000000000000 000000000000000000000100000000000000000000000000 010000000000111111111111111111111111100010110000 111111111111111111111111100010100000000100000000 000000101110000000000000001011110000000000000001 000000000000001100000000000000000011000100000000 000000010000000000000011010000000000000000110101 000000000000000100000000000000110110000000000000 001101110000000000000001000000000000001110000000 000000000011100100000000000000010000000000000011 101000000000000000111011000000000000000100000000 000000111100000000000000001111010000000000000001 000000000000001111100000000000000011111100000000 000000010000000000000100000000000000000001000001 000000000000000100000000000001000010000000000000 010000110000000000000001000000000000010001000000 000000000100010100000000000000010000000000000100 011000000000000001000111000000000000000100000000 000001001000000000000000010010010000000000000001
133. 2GL 1GL stack::top() const [clone .cold]: push r14 mov edi,0x10 push rbx push rax call 401030 <__cxa_allocate_exception@plt> mov esi,0x402004 mov rdi,rax mov rbx,rax call 401080 <std::logic_error::logic_error(char const*)@plt> mov edx,0x401070 mov esi,0x403da0 mov rdi,rbx call 4010a0 <__cxa_throw@plt> mov r14,rax mov rdi,rbx call 401040 <__cxa_free_exception@plt> mov rdi,r14 call 4010b0 <_Unwind_Resume@plt> nop stack::pop() [clone .cold]: push r14 mov edi,0x10 push rbx push rax call 401030 <__cxa_allocate_exception@plt> mov esi,0x40201c mov rdi,rax mov rbx,rax call 401080 <std::logic_error::logic_error(char const*)@plt> mov edx,0x401070 mov esi,0x403da0 mov rdi,rbx call 4010a0 <__cxa_throw@plt> mov r14,rax mov rdi,rbx call 401040 <__cxa_free_exception@plt> mov rdi,r14 call 4010b0 <_Unwind_Resume@plt> nop WORD PTR [rax+rax*1+0x0] stack::push(void* const&): push rbp mov rbp,rsi push rbx mov rbx,rdi mov edi,0x10 sub rsp,0x8 call 401050 <operator new(unsigned long)@plt> movq xmm0,QWORD PTR [rbx] movhps xmm0,QWORD PTR [rbp+0x0] mov QWORD PTR [rbx],rax movups XMMWORD PTR [rax],xmm0 add rsp,0x8 pop rbx pop rbp ret nop nop DWORD PTR [rax+0x0]
134. 3GL 2GL 1GL template<typename T> class stack { public: std::size_t depth() const { return items.size(); } const T & top() const { if (items.empty()) throw std::logic_error("stack::top: empty"); return items.back(); } void push(const T & item) { items.push_back(item); } void pop() { if (items.empty()) throw std::logic_error("stack::pop: empty"); items.pop_back(); } private: std::vector<T> items; };
135. A stack is a LIFO abstract data type with the following public behaviour: NL 3GL 2GL 1GL * A stack is initially empty * depth() returns the number of elements in the stack * top() returns the top element if present, otherwise it fails with an exception * push(item) adds a new element, item, to the top of the stack * pop() removes and returns the top element of the stack if present, otherwise it fails with an exception * Elements of a given stack are all of the same type A stack's internal representation is private. * The representation should be memory efficient * The representation should support efficient end addition and removal
136. new MDA 3GL 2GL 1GL Empty pop [depth = 1] push(item) Non-Empty push pop [depth > 1] top
137. Maths  T •  Stack { new: Stack[T], push: Stack[T]  T → Stack[T], pop: Stack[T] ⇀ Stack[T], depth: Stack[T] → Integer, top: Stack[T] ⇀ T 3GL 2GL 1GL }
138. Maths English
139. English Maths
140. 5GL 4GL 3GL 2GL 1GL
141. 5GL DSL 3GL 2GL 1GL
142. DSL+ DSL 3GL 2GL 1GL
143. DSL 3GL 2GL 1GL
144. DSL 3GL 2GL 1GL
145. 3GL 2GL 1GL
146. 3GL 2GL 1GL Ada Bash C C# C++ COBOL Dart Delphi Fortran Go Java JavaScript Kotlin MATLAB Objective-C Perl PHP PowerShell Python R Ruby Rust Scala Scratch SQL Swift TypeScript Visual Basic
147. 3GL NL
148. 3GL NL
149. 3GL NL
150. We could, of course, use any notation we want. Richard Feynman
151. But we probably won’t.
154. spectrum.ieee.org/top-programming-languages-2025
156. Betteridge’s law, noun ▪ The maxim that any headline that ends in a question mark can be answered “No” facebook.com/WordFriday/posts/383363405084956
157. The past is never dead. It’s not even past. William Faulkner