Why Software Engineers Disagree About Everything

Why are there are so many disagreements in software? Why don’t we all converge on the same beliefs or technologies? It might sound obvious that people shouldn't agree, but I want to convince you it’s weird that we don't. This talk will be a philosophical exploration of how knowledge converges within subcultures, as I explore this question through the worlds of software, online fraud, and poker.

https://www.youtube.com/watch?v=x07q6V4VXC8

Transcript

1. WHY SOFTWARE ENGINEERS DISAGREE ABOUT EVERYTHING HASEEB QURESHI SOFTWARE ENGINEER, AIRBNB @hosseeb
2. LET’S TALK PHILOSOPHY.
3. EPISTEMOLOGY Epistemology, from Greek ἐπιστήμη, (epistēmē), meaning “unnecessary”, and λόγος, (logos), meaning “etymology.” C. 1809, Phrenology diagram by Joseph Gall
4. EPISTEMOLOGY Epistemology, from Greek ἐπιστήμη, (epistēmē), meaning “unnecessary”, and λόγος, (logos), meaning “etymology.” Epistemology studies the nature of knowledge, justification, and the rationality of belief. C. 1809, Phrenology diagram by Joseph Gall
5. HOW DOES ANYONE KNOW ANYTHING? AND HOW DO THEY KNOW THAT THEY KNOW IT?
6. A LITTLE BIT ABOUT ME: Studied English & Philosophy Professional poker player Programming instructor Risk engineer
7. KNOWLEDGE IS DEEPLY CULTURAL.
8. Full-stack JavaScript is the future.
9. Full-stack JavaScript is the future. Everyone should know C.
10. Full-stack JavaScript is the future. Everyone should know C. Rust is the best language for systems programming.
11. Full-stack JavaScript is the future. Everyone should know C. Rust is the best language for systems programming. Relational databases don’t scale.
12. Full-stack JavaScript is the future. Everyone should know C. Rust is the best language for systems programming. Relational databases don’t scale. TDD is a fantasy, nobody serious does it.
13. I’M NOT INTERESTED IN CONVINCING YOU OF THESE CLAIMS.
14. I’M NOT INTERESTED IN CONVINCING YOU OF THESE CLAIMS. I’M INTERESTED IN WHY WE DISAGREE ABOUT THEM.
15. No-limit is dying, you have to learn mixed games. IN POKER, IT WAS…
16. No-limit is dying, you have to learn mixed games. Everyone needs to use a HUD. IN POKER, IT WAS…
17. No-limit is dying, you have to learn mixed games. Everyone needs to use a HUD. Only fish play loose-passive styles. IN POKER, IT WAS…
18. No-limit is dying, you have to learn mixed games. Everyone needs to use a HUD. Only fish play loose-passive styles. GTO is a fantasy, nobody actually plays like that. IN POKER, IT WAS…
19. WHEN MOST PEOPLE HEAR STUFF LIKE THIS, THEY ASK, “IS THAT TRUE? WHAT SHOULD I DO ABOUT IT?”
20. WHEN I HEAR THIS, I ASK, “WHY DO THEY THINK THEY KNOW THAT?”
21. THE WEIRD THING ABOUT PROGRAMMING IS THAT NOBODY AGREES.
22. YOU MIGHT THINK IT’S OBVIOUS THAT PEOPLE WON’T AGREE.
23. I THINK IT’S WEIRD.
24. IN A WAY, IT MAKES MORE SENSE FOR PEOPLE TO AGREE THAN TO DISAGREE.
25. WHAT I MEAN IS: SYSTEMS TEND TO CONVERGE.
26. WHEN YOU LOOK AT A SYSTEM, YOU SHOULD ASSUME IT WILL CONVERGE ON WHAT’S OPTIMAL.
27. LET ME GIVE YOU AN EXAMPLE.
28. IN POKER, THERE’S A STRATEGY CALLED “SET MINING”
29. IT WAS PRETTY SIMPLE AND STUPID.
30. IT WAS PRETTY SIMPLE AND STUPID. IT WORKED.
31. PRETTY SOON ALMOST EVERYONE WAS SET-MINING AT LOW-MID STAKES FULL RING GAMES.
32. THE GAME CONVERGED ON SET-MINING. EVERYONE SAW IT WAS THE HIGH GROUND AND MOVED IN THAT DIRECTION.
36. ONLINE POKER CONVERGED IN A WAY LIVE POKER NEVER DID.
37. NATURE IS FULL OF CONVERGENT SYSTEMS.
42. WHEN I STARTED WORKING AS A RISK ENGINEER AT AIRBNB, I STARTED LOOKING FOR CONVERGENCE.
44. HOW DO FRAUDSTERS FIGURE OUT WHAT TO DO? HOW DO THEY COMMUNICATE THIS KNOWLEDGE WITH EACH OTHER?
45. IS THERE CONVERGENCE IN FRAUD? (The new controversial Macbook Pro)
48. PATCH THE VULNERABILITY, WATCH THEM DISPERSE.
49. AND THE FRAUDSTERS HEAD TO THE NEXT HIGHEST PEAKS.
50. THIS LOOKS CONVERGENT. IT MAKES SENSE.
51. BUT…
52. WHY ARE THEY DEFRAUDING US AT ALL?
54. THERE IS SOME TOPOLOGY OF FRAUD
55. YOU’D THINK EVERYONE WOULD CONVERGE ON THE PEAKS.
56. INSTEAD, THE FRAUD IS SPREAD OUT EVERYWHERE.
57. THIS IS WEIRD.
58. DO THOSE OTHER FRAUDSTERS NOT CARE ABOUT OPTIMIZING?
59. YOU SEE THE SAME THING IN SOFTWARE.
60. SOFTWARE DOESN’T SEEM TO CONVERGE.
61. WHEN IT DOES, IT’S GENUINELY SURPRISING.
62. ONE OBVIOUS ANSWER: MULTIPLE TERRAINS CRUD apps Payments backends Social networks
63. … BUT EVEN WITHIN A SINGLE TERRAIN, WE DON’T SEE CONVERGENCE. CRUD apps Payments backends Social networks
64. YOU SHOULD WANT CONVERGENCE!
65. CONVERGENCE MEANS WE ALL SEE THE UNDERLYING TERRAIN, AND WE ALL DO THE BEST THING.
66. SO WHY DO THESE SYSTEMS NOT CONVERGE?
67. I THINK THERE ARE FOUR REASONS.
68. #1
69. UNSTABLE TERRAIN
70. UNSTABLE TERRAIN
73. #2
74. HIGH SWITCHING COSTS
75. HIGH SWITCHING COSTS
76. THIS IS EXACERBATED BY SPECIALIZATION.
77. #3
78. INFORMATION SHARING
79. Open Closed
80. Open Closed
81. Open Closed
82. Open Closed
83. Open Closed ???
84. #4
85. GROUP IDENTITY
86. , Java-lovers Pythonistas Scalaists Rubyists PROGRAMMING
87. , Uber Google Square Facebook FRAUD RINGS
88. SOCIAL IDENTITY THEORY
89. SOCIAL IDENTITY THEORY Social categorization
90. SOCIAL IDENTITY THEORY Social categorization
91. SOCIAL IDENTITY THEORY Social categorization Social identification
92. SOCIAL IDENTITY THEORY Social categorization Social identification
93. SOCIAL IDENTITY THEORY Social categorization Social identification Social comparison
94. SOCIAL IDENTITY THEORY Social categorization Social identification Social comparison
95. DIFFERENTIAL PSYCHOLOGY
97. Rails mountain
100. WE SHOULD WANT TO FIND THE TRUE GLOBAL OPTIMUM.
101. WE SHOULD WANT TO FIND THE TRUE GLOBAL OPTIMUM. SO WHAT SHOULD WE DO ABOUT THIS AS SOFTWARE ENGINEERS?
102. TWO PIECES OF ADVICE.
103. KEEP YOUR IDENTITY SMALL. FIRST:
104. DON’T THINK OF YOURSELF AS A RAILS DEVELOPER OR A B DEVELOPER.
105. YOU ARE A SOFTWARE ENGINEER.
106. SECOND:
107. EXPLORE THE TERRAIN
108. DO THINGS YOU’VE NEVER DONE BEFORE
109. TAKE REAL RISKS.
110. GO TO DJANGOCON.
111. GO TO DJANGOCON. (NOT REALLY, IT PROBABLY SUCKS.)
113. I AM HASEEB QURESHI You can find me on Twitter: @hosseeb You can read my blog at: haseebq.com