WEBVTT 00:00:00.000 --> 00:00:01.820 Welcome to the Deep Dive, the place where we 00:00:01.820 --> 00:00:03.879 take the sources you share with us, strip away 00:00:03.879 --> 00:00:07.160 the noise, and hopefully hand you the foundational 00:00:07.160 --> 00:00:10.220 knowledge you need with a few aha moments thrown 00:00:10.220 --> 00:00:12.439 in. Yeah, our mission is really to give you, 00:00:12.500 --> 00:00:15.539 the listener, a genuine shortcut to being well 00:00:15.539 --> 00:00:17.839 -informed on these, you know, really deep topics. 00:00:18.100 --> 00:00:21.199 And today we are embarking on a deep dive into 00:00:21.199 --> 00:00:24.239 the life and, I guess, the seminal work of a 00:00:24.239 --> 00:00:27.780 figure whose name might not be, you know, a household 00:00:27.780 --> 00:00:31.379 name. Not like some CEOs, no. Yep. But his technical 00:00:31.379 --> 00:00:34.079 contributions are, and this is not an exaggeration, 00:00:34.179 --> 00:00:37.100 the literal bedrock of modern computing. We're 00:00:37.100 --> 00:00:39.060 talking about Kenneth Lane Thompson. Ken Thompson. 00:00:39.179 --> 00:00:41.719 We often discuss the massive influence of founders 00:00:41.719 --> 00:00:44.420 and visionaries, but Thompson is, he's different. 00:00:44.479 --> 00:00:47.500 He's the engineer's engineer, the architect behind 00:00:47.500 --> 00:00:49.880 the fundamental structures. He's widely considered 00:00:49.880 --> 00:00:51.700 one of the greatest computer programmers of all 00:00:51.700 --> 00:00:54.039 time. So our mission today is pretty straightforward. 00:00:54.600 --> 00:00:56.979 We want to synthesize his decades -long influence 00:00:56.979 --> 00:01:00.020 and show you how this one person really did lay 00:01:00.020 --> 00:01:02.119 the groundwork for the entire digital world. 00:01:02.259 --> 00:01:04.379 And when you say the entire digital world, you 00:01:04.379 --> 00:01:07.340 mean that quite literally. I do. If you use a 00:01:07.340 --> 00:01:10.459 cell phone, Android or iOS, if you open a web 00:01:10.459 --> 00:01:13.159 browser, if you interact with the cloud, you 00:01:13.159 --> 00:01:16.040 are relying on design choices he made 50 years 00:01:16.040 --> 00:01:18.120 ago. And what's fascinating to me from the source 00:01:18.120 --> 00:01:21.159 material is the sheer breadth of his work. Right. 00:01:21.219 --> 00:01:24.219 It wasn't just one great program. It's the operating 00:01:24.219 --> 00:01:26.420 system. It's the programming language. It's foundational 00:01:26.420 --> 00:01:29.299 security concepts. Even the text encoding that 00:01:29.299 --> 00:01:33.040 allows the global Internet to, you know, function. 00:01:33.560 --> 00:01:36.280 It's an astonishing portfolio. So we're going 00:01:36.280 --> 00:01:38.459 to trace his journey basically chronologically. 00:01:38.680 --> 00:01:41.140 We'll start at Bell Labs in the 60s, move through 00:01:41.140 --> 00:01:43.260 the Unix revolution and bring it right up to 00:01:43.260 --> 00:01:45.140 his current work at Google on the go language. 00:01:45.359 --> 00:01:48.599 So let's unpack this and start with the mind 00:01:48.599 --> 00:01:51.579 that drove it all. Sounds good. So Ken Thompson 00:01:51.579 --> 00:01:54.719 was born on February 4th, 1943 in New Orleans, 00:01:54.819 --> 00:01:57.500 Louisiana. But to really understand his contributions, 00:01:57.719 --> 00:01:59.900 it seems like we need to focus less on the geography 00:01:59.900 --> 00:02:04.379 and more on his innate drive. Exactly. This wasn't 00:02:04.379 --> 00:02:06.540 a man who fell in love with a specific gadget 00:02:06.540 --> 00:02:09.500 or computer. He fell in love with logic. And 00:02:09.500 --> 00:02:11.560 that's a key distinction, isn't it? It's vital. 00:02:11.699 --> 00:02:13.240 I mean, when you ask a lot of engineers what 00:02:13.240 --> 00:02:15.460 got them started, they'll mention an early Apple 00:02:15.460 --> 00:02:18.849 II or something. Thompson, though, he described 00:02:18.849 --> 00:02:22.569 this lifelong sort of internal fascination with 00:02:22.569 --> 00:02:24.669 systems. He said he was working on arithmetic 00:02:24.669 --> 00:02:28.169 problems in binary in grade school. In grade 00:02:28.169 --> 00:02:32.030 school. Yes. Just for the mental exercise. He 00:02:32.030 --> 00:02:33.830 summed it up perfectly. He said he did it just 00:02:33.830 --> 00:02:36.530 because I was fascinated. That tells you everything. 00:02:36.669 --> 00:02:39.189 His entire career was driven by solving these 00:02:39.189 --> 00:02:42.530 abstract foundational problems, not by, you know, 00:02:42.530 --> 00:02:44.930 the next hot product. So that fascination with 00:02:44.930 --> 00:02:47.469 logic took him to UC Berkeley, where he got both 00:02:47.469 --> 00:02:49.689 his bachelor's and master's degrees in electrical 00:02:49.689 --> 00:02:51.810 engineering and computer sciences. Right. In 00:02:51.810 --> 00:02:55.210 65 and 66. And armed with that academic foundation, 00:02:55.490 --> 00:02:58.930 he was hired by Bell Labs in 1966. And this is 00:02:58.930 --> 00:03:01.129 really where the drama of modern computing begins, 00:03:01.349 --> 00:03:04.009 in the middle of this massive, complex project 00:03:04.009 --> 00:03:06.449 called Multics. Right, Multics. Multics stood 00:03:06.449 --> 00:03:08.729 for Multiplexed Information and Computing Services, 00:03:08.789 --> 00:03:11.030 this huge, ambitious time -sharing operating 00:03:11.030 --> 00:03:14.409 system, a collaboration between MIT, Bell Labs, 00:03:14.689 --> 00:03:16.830 and General Electric. And the goal was what? 00:03:16.969 --> 00:03:19.090 To build something like a computing utility. 00:03:19.490 --> 00:03:22.689 Exactly. Like the power grid. A massive system 00:03:22.689 --> 00:03:24.969 that could serve hundreds of users at the same 00:03:24.969 --> 00:03:28.050 time reliably. And Thompson worked on this gargantuan 00:03:28.050 --> 00:03:30.990 project alongside his future collaborator, Dennis 00:03:30.990 --> 00:03:32.909 Ritchie. And while he was there, he developed 00:03:32.909 --> 00:03:36.330 his first significant programming language, which 00:03:36.330 --> 00:03:39.270 he called Bon. Yeah, Bon. A language designed 00:03:39.270 --> 00:03:41.430 to be useful inside the Multics environment. 00:03:41.550 --> 00:03:44.389 And it kind of showcased his knack for creating 00:03:44.389 --> 00:03:47.250 these concise, effective tools. But Multics itself. 00:03:48.110 --> 00:03:52.490 It became this notoriously complex, slow, resource 00:03:52.490 --> 00:03:55.409 -intensive project. A classic case of scope creep, 00:03:55.530 --> 00:03:57.389 it sounds like. Total scope creep. It was trying 00:03:57.389 --> 00:03:59.650 to solve every problem at once. Yeah. So actually, 00:03:59.830 --> 00:04:02.250 Bell Labs made a crucial decision, a decision 00:04:02.250 --> 00:04:05.110 that paradoxically created the entire modern 00:04:05.110 --> 00:04:07.219 computing landscape. They pulled out. They withdrew 00:04:07.219 --> 00:04:09.699 from the collaboration in 1969. They decided 00:04:09.699 --> 00:04:11.580 it was just too costly, too complex, and that 00:04:11.580 --> 00:04:13.759 withdrawal left a vacuum. And this is where the 00:04:13.759 --> 00:04:15.639 famous anecdote comes in, right? The story of 00:04:15.639 --> 00:04:17.879 how Unix was born. It's a perfect example of 00:04:17.879 --> 00:04:21.000 how personal passion drives invention. Thomson 00:04:21.000 --> 00:04:23.160 wasn't focused on the corporate roadmap. He was 00:04:23.160 --> 00:04:25.379 focused on his own project. And that project 00:04:25.379 --> 00:04:28.019 was a video game he'd written. A video game called 00:04:28.019 --> 00:04:30.819 Space Travel. It simulated the solar system, 00:04:30.959 --> 00:04:33.540 and once Bell Labs pulled the plug on Multics, 00:04:33.699 --> 00:04:36.839 Thomson had a very frustrating problem. He couldn't 00:04:36.839 --> 00:04:39.459 run his game anymore. He needed a platform. So 00:04:39.459 --> 00:04:43.379 he finds this old, forgotten, underpowered machine, 00:04:43.779 --> 00:04:47.379 a PDP -7. The PDP -7 was an ancient machine, 00:04:47.560 --> 00:04:51.199 even for 1969. And the key constraint, get this, 00:04:51.360 --> 00:04:54.420 it had only about 16 kilobytes of memory. 16 00:04:54.420 --> 00:04:57.019 kilobytes? That's nothing. My watch has more 00:04:57.019 --> 00:04:58.620 than that. It's nothing. You couldn't even store 00:04:58.620 --> 00:05:00.620 a decent photo with that. So Thompson realized 00:05:00.620 --> 00:05:02.639 that to get space travel running, he couldn't 00:05:02.639 --> 00:05:04.939 port the complexity of Multics. He needed something 00:05:04.939 --> 00:05:07.870 simple. efficient, and tiny. So the necessity 00:05:07.870 --> 00:05:10.089 of running a game on a cheap machine forced him 00:05:10.089 --> 00:05:12.610 to reject complexity and embrace minimalism. 00:05:12.810 --> 00:05:15.389 That's the core of it. He rewrote the game on 00:05:15.389 --> 00:05:18.490 the PDP -7, but to do that, he first had to build 00:05:18.490 --> 00:05:21.639 the tools. An assembler, a text editor, a tiny 00:05:21.639 --> 00:05:24.500 little kernel, and those foundational tools very 00:05:24.500 --> 00:05:27.079 quickly coalesced into the first version of the 00:05:27.079 --> 00:05:29.319 Unix operating system. It's a complete philosophical 00:05:29.319 --> 00:05:32.459 shift. It is. Multics was a cathedral. Unix was 00:05:32.459 --> 00:05:34.839 a bazaar of small, simple tools that each did 00:05:34.839 --> 00:05:38.079 one job well. The constraints of the PDP -7 forced 00:05:38.079 --> 00:05:41.019 that elegance, and that elegance became the defining 00:05:41.019 --> 00:05:43.540 DNA of modern computing. Okay, so let's unpack 00:05:43.540 --> 00:05:46.680 that. On this tiny PDP -7 Thompson with Dennis 00:05:46.680 --> 00:05:49.500 Ritchie and Red Canaday, they basically architected 00:05:49.500 --> 00:05:52.000 the core concepts we still use today. And the 00:05:52.000 --> 00:05:54.759 impact here, it just can't be overstated. They 00:05:54.759 --> 00:05:57.220 define concepts that every single major OS Linux, 00:05:57.399 --> 00:06:00.079 Windows, Mac OS still relies on. Let's start 00:06:00.079 --> 00:06:02.019 with maybe the most radical one, the concept 00:06:02.019 --> 00:06:04.139 of device files. So what was so revolutionary 00:06:04.139 --> 00:06:06.500 about treating a piece of hardware like a printer 00:06:06.500 --> 00:06:09.490 is just a file. Well, before Unix, if you wanted 00:06:09.490 --> 00:06:12.350 your program to talk to a printer or keyboard, 00:06:12.750 --> 00:06:15.290 you had to write a highly specialized driver 00:06:15.290 --> 00:06:19.149 that spoke directly to the hardware. The OS code 00:06:19.149 --> 00:06:22.709 was incredibly complex and completely tied to 00:06:22.709 --> 00:06:24.769 that specific hardware. So it was impossible 00:06:24.769 --> 00:06:27.810 to port. Almost impossible. Thompson and his 00:06:27.810 --> 00:06:29.730 team decided to just treat every I .O. device, 00:06:30.050 --> 00:06:32.029 your hard drive, your keyboard, your screen, 00:06:32.149 --> 00:06:34.250 as if it were just a simple file you could read 00:06:34.250 --> 00:06:36.420 from and write to. So the program doesn't need 00:06:36.420 --> 00:06:38.240 to know what it's talking to, just the file name. 00:06:38.439 --> 00:06:41.040 Exactly. It created this unified, simple interface 00:06:41.040 --> 00:06:43.720 for all I .O. If you want to print something, 00:06:43.819 --> 00:06:46.360 you just write data to the printer's device file. 00:06:46.500 --> 00:06:48.980 It was genius. It massively simplified the kernel 00:06:48.980 --> 00:06:52.360 and made Unix manageable and, crucially, portable. 00:06:52.600 --> 00:06:54.980 That's the first real step toward hardware abstraction, 00:06:55.279 --> 00:06:58.040 which is essential for any modern OS. It is. 00:06:58.139 --> 00:07:00.019 And they also defined the concept of computer 00:07:00.019 --> 00:07:02.930 processes, the command line shell. But the real 00:07:02.930 --> 00:07:04.670 technical breakthrough, the thing that defined 00:07:04.670 --> 00:07:07.610 Unix's power, was the concept of pipes. This 00:07:07.610 --> 00:07:09.250 is the centerpiece of the whole Unix philosophy, 00:07:09.569 --> 00:07:11.850 right? It's the ultimate rejection of the Multics 00:07:11.850 --> 00:07:15.189 monolith. Pipes enabled simple, inter -process 00:07:15.189 --> 00:07:19.050 communication. Before this, if Program A needed 00:07:19.050 --> 00:07:22.050 to pass its result to Program B, it was clumsy. 00:07:23.009 --> 00:07:25.069 Program A would save its data to a temporary 00:07:25.069 --> 00:07:27.189 file on the disk. And then Program B would have 00:07:27.189 --> 00:07:30.769 to open and read that file back in? Slow? Super 00:07:30.769 --> 00:07:35.180 slow, inefficient, wasted disk space. Thompson 00:07:35.180 --> 00:07:37.500 and his team realized they could just route the 00:07:37.500 --> 00:07:40.319 output stream of program A directly into the 00:07:40.319 --> 00:07:43.060 input stream of program B using an in -memory 00:07:43.060 --> 00:07:45.259 buffer, the pipe. So you completely bypass the 00:07:45.259 --> 00:07:47.519 disk. Completely. And this means you can combine 00:07:47.519 --> 00:07:50.220 tiny, single -purpose utilities to do complex 00:07:50.220 --> 00:07:52.680 tasks. You can take a program that searches text, 00:07:52.860 --> 00:07:55.240 pipe its output to a program that sorts it, and 00:07:55.240 --> 00:07:57.019 pipe that to a program that counts the lines. 00:07:57.220 --> 00:07:59.180 It's like a computational assembly line. Perfect 00:07:59.180 --> 00:08:01.740 analogy. Instead of one giant machine, you have 00:08:01.740 --> 00:08:03.980 specialized stations that are infinitely customizable. 00:08:04.560 --> 00:08:06.899 That philosophy small tools doing one thing well, 00:08:07.000 --> 00:08:09.480 connected by universal interfaces, that is the 00:08:09.480 --> 00:08:12.379 DNA of Unix. And that DNA needed a better language 00:08:12.379 --> 00:08:14.560 to be written in. The early work was on the PDP 00:08:14.560 --> 00:08:16.839 -7, but Thompson knew they needed a real system 00:08:16.839 --> 00:08:19.279 programming language. Especially to move to the 00:08:19.279 --> 00:08:21.500 more popular PDP -11 computer. And that's where 00:08:21.500 --> 00:08:24.279 the B language comes in. Yes. Thompson created 00:08:24.279 --> 00:08:26.839 B, which was a simplified, typeless language 00:08:26.839 --> 00:08:28.959 designed specifically for systems programming. 00:08:29.500 --> 00:08:32.980 And B was the direct precursor to Dennis Ritchie's 00:08:32.980 --> 00:08:36.990 C language. The history of UNIX and C is so intertwined. 00:08:37.070 --> 00:08:39.029 They're completely symbiotic. They grew up together. 00:08:39.570 --> 00:08:42.470 Thompson later said C grew up with one of the 00:08:42.470 --> 00:08:45.070 rewritings of the system, and as such, it became 00:08:45.070 --> 00:08:48.090 perfect for writing systems. When they rewrote 00:08:48.090 --> 00:08:50.570 UNIX and C, that process sharpened C itself, 00:08:50.850 --> 00:08:52.990 making sure it had the low -level control needed 00:08:52.990 --> 00:08:54.889 for a kernel. It's interesting to see the division 00:08:54.889 --> 00:08:57.070 of labor there. People often lump them together, 00:08:57.269 --> 00:08:59.230 but Thompson himself gave a bit more clarity 00:08:59.230 --> 00:09:02.149 on who did what. He did. In a 2011 interview, 00:09:02.309 --> 00:09:05.399 he was very... Clear. He said he did the first 00:09:05.399 --> 00:09:08.539 version of the kernel alone. Ritchie became the 00:09:08.539 --> 00:09:10.700 evangelist and worked mostly on the I .O. system 00:09:10.700 --> 00:09:12.840 and, of course, the C language during the later 00:09:12.840 --> 00:09:15.580 rewrites. Thompson was the primary architect 00:09:15.580 --> 00:09:18.000 of that minimalist kernel. And finally, the name 00:09:18.000 --> 00:09:19.700 itself. It wasn't even Thompson who named it. 00:09:19.840 --> 00:09:22.240 No, that was another Bell Labs colleague, Brian 00:09:22.240 --> 00:09:25.399 Kernighan, who suggested Unix in 1970. And it 00:09:25.399 --> 00:09:28.480 was a pun on Multics. A perfect pun. Multics 00:09:28.480 --> 00:09:31.779 implied multiplexed. Many things. Unix implied 00:09:31.779 --> 00:09:35.120 a simpler, single, unified structure that name 00:09:35.120 --> 00:09:37.740 just perfectly captured that philosophical shift. 00:09:37.960 --> 00:09:40.740 So his legacy expands way beyond just the OS 00:09:40.740 --> 00:09:43.500 kernel. One of the most widespread fundamental 00:09:43.500 --> 00:09:46.659 tools in all of modern programming is also a 00:09:46.659 --> 00:09:49.419 Thompson invention. Regular expressions. Oh, 00:09:49.419 --> 00:09:52.509 yeah. Rejects. If you've ever written code to, 00:09:52.549 --> 00:09:54.889 you know, validate an email address or search 00:09:54.889 --> 00:09:56.830 through log files for a pattern, you're using 00:09:56.830 --> 00:09:59.309 his work. And this idea dates back even before 00:09:59.309 --> 00:10:02.070 Unix, right? It does. Back in the 1960s during 00:10:02.070 --> 00:10:04.230 the Multics era, he developed a version of the 00:10:04.230 --> 00:10:07.610 text editor QED. And QED was groundbreaking because 00:10:07.610 --> 00:10:09.850 it included this powerful way to search and manipulate 00:10:09.850 --> 00:10:12.870 text based on these patterns. It was a huge leap 00:10:12.870 --> 00:10:14.649 from simple keyword searching. But it was his 00:10:14.649 --> 00:10:17.289 later Unix editor, Nigg, that really made regular 00:10:17.289 --> 00:10:19.990 expressions a daily tool for programmers. It's 00:10:19.990 --> 00:10:23.230 exactly. Ayed was the standard Unix text editor. 00:10:23.450 --> 00:10:26.509 And Rejects was so deeply integrated that it 00:10:26.509 --> 00:10:28.789 just became part of the background of using Unix. 00:10:28.889 --> 00:10:31.750 And he didn't just invent the concept, he standardized 00:10:31.750 --> 00:10:34.690 the syntax. And that notation is still the foundation 00:10:34.690 --> 00:10:37.450 for almost every modern implementation, from 00:10:37.450 --> 00:10:40.929 Python to JavaScript. That's a huge legacy. But 00:10:40.929 --> 00:10:43.570 searching complex patterns can be slow, especially 00:10:43.570 --> 00:10:46.169 in big files. He also tackled the performance 00:10:46.169 --> 00:10:48.570 side of it. And this is where his mathematical 00:10:48.570 --> 00:10:52.309 and theoretical genius really shines. He invented 00:10:52.309 --> 00:10:54.169 what's called Thomson's construction algorithm. 00:10:54.450 --> 00:10:56.830 Okay, what does that do? The problem with rejects 00:10:56.830 --> 00:10:58.870 is that matching a pattern can involve a lot 00:10:58.870 --> 00:11:00.809 of backtracking and checking different paths, 00:11:00.990 --> 00:11:03.929 which is slow. His algorithm converts the regular 00:11:03.929 --> 00:11:06.590 expression pattern into a specific type of state 00:11:06.590 --> 00:11:09.350 machine called a non -deterministic finite automaton, 00:11:09.490 --> 00:11:12.269 or NFA. So it's like a flowchart for the search 00:11:12.269 --> 00:11:14.350 pattern. A perfect way to put it. Instead of 00:11:14.350 --> 00:11:16.250 the search engine constantly backtracking in 00:11:16.250 --> 00:11:18.970 the text, the NFA lets the search process the 00:11:18.970 --> 00:11:21.850 text character by character, efficiently tracking 00:11:21.850 --> 00:11:24.970 all possible matches at the same time. It dramatically 00:11:24.970 --> 00:11:27.289 speeds up pattern matching. So he invents the 00:11:27.289 --> 00:11:29.610 tool, and then he invents the mathematical engine 00:11:29.610 --> 00:11:31.830 that makes the tool fast enough to be useful 00:11:31.830 --> 00:11:34.809 at scale. Correct. And that engine is still fundamental 00:11:34.809 --> 00:11:38.549 to how many text parsers work today. And of course, 00:11:38.549 --> 00:11:40.669 the most famous utility built on all this is 00:11:40.669 --> 00:11:44.279 grep. Right. GREP, which stands for Global Regular 00:11:44.279 --> 00:11:47.580 Expression Print, the go -to tool for sifting 00:11:47.580 --> 00:11:49.480 through text. It's one of the pillars of the 00:11:49.480 --> 00:11:52.039 whole system. So now we've got the core OS, the 00:11:52.039 --> 00:11:54.019 foundational language and the fundamental text 00:11:54.019 --> 00:11:56.559 tools. But this was all still proprietary, locked 00:11:56.559 --> 00:11:59.740 up inside Bell Labs. How did it get out and become 00:11:59.740 --> 00:12:02.419 this global phenomenon? And that's where Thompson 00:12:02.419 --> 00:12:04.879 sort of becomes this intellectual emissary. It 00:12:04.879 --> 00:12:08.279 all started with his sabbatical in 1975. He went 00:12:08.279 --> 00:12:11.460 back to his alma mater, UC Berkeley. And while 00:12:11.460 --> 00:12:13.200 he was there, he didn't just give a lecture. 00:12:13.320 --> 00:12:16.639 He personally helped install version 6 Unix on 00:12:16.639 --> 00:12:19.840 a university computer, a PDP -1170. And that 00:12:19.840 --> 00:12:22.899 one act had just monumental consequences. It 00:12:22.899 --> 00:12:25.539 planted the most important seed. That Unix installation 00:12:25.539 --> 00:12:28.419 at Berkeley attracted a group of brilliant students, 00:12:28.620 --> 00:12:31.519 most famously Bill Joy. They started maintaining 00:12:31.519 --> 00:12:34.059 and developing it on their own, adding new utilities, 00:12:34.240 --> 00:12:36.480 better networking, and this became known as the 00:12:36.480 --> 00:12:39.720 Berkeley Software Distribution. BSD. BSD. And 00:12:39.720 --> 00:12:43.440 the lineage from BSD is just massive. Free BSD, 00:12:43.700 --> 00:12:46.659 open BSD, and crucially, the networking stack, 00:12:46.799 --> 00:12:50.019 the TCPIP code, developed at Berkeley. That became 00:12:50.019 --> 00:12:52.320 the gold standard for the internet. So when Apple 00:12:52.320 --> 00:12:56.139 built macOS and iOS, they chose a modified BSD 00:12:56.139 --> 00:12:59.200 kernel as their core. They did. So when you pull 00:12:59.200 --> 00:13:01.419 out your iPhone, you are running the intellectual 00:13:01.419 --> 00:13:03.879 descendant of a system that Ken Thompson personally 00:13:03.879 --> 00:13:06.419 hand -delivered to Berkeley in 1975. It's an 00:13:06.419 --> 00:13:08.360 incredible story of influence. And even during 00:13:08.360 --> 00:13:10.100 that sabbatical, he was still building things. 00:13:10.220 --> 00:13:12.919 Oh, yeah, in his spare time. In early 1976, he 00:13:12.919 --> 00:13:14.799 wrote the initial version of Berkeley Pascal. 00:13:15.340 --> 00:13:17.600 Now, others improved it later, but the fact that 00:13:17.600 --> 00:13:19.899 he was simultaneously shaping system architecture, 00:13:20.639 --> 00:13:23.100 text processing, and programming languages just 00:13:23.100 --> 00:13:25.379 speaks to his relentless curiosity. It's hard 00:13:25.379 --> 00:13:27.639 to reconcile that the same mind that built so 00:13:27.639 --> 00:13:30.440 much of the digital world also devised one of 00:13:30.440 --> 00:13:32.279 its most terrifying security vulnerabilities. 00:13:33.419 --> 00:13:36.320 Let's shift to his 1983 Turing Award speech, 00:13:36.679 --> 00:13:41.399 Reflections on Trusting Trust. This is a cornerstone 00:13:41.399 --> 00:13:43.700 of computer security. It's known as the Thompson 00:13:43.700 --> 00:13:47.720 hack or the trusting trust attack. It's a concept 00:13:47.720 --> 00:13:50.059 for a persistent, self -propagating compiler 00:13:50.059 --> 00:13:53.940 backdoor that is completely undetectable by reviewing 00:13:53.940 --> 00:13:56.159 the source code. Okay, so let's define the terms. 00:13:56.340 --> 00:13:58.639 The compiler is the program that takes human 00:13:58.639 --> 00:14:00.720 -readable source code and turns it into the machine 00:14:00.720 --> 00:14:03.480 code the computer actually runs. Right, and we 00:14:03.480 --> 00:14:06.039 have to, by definition, trust that the compiler 00:14:06.039 --> 00:14:08.639 is doing its job, honestly. And Thompson showed 00:14:08.639 --> 00:14:11.179 that trust is the ultimate vulnerability. He 00:14:11.179 --> 00:14:13.620 did. His thought experiment showed how a C compiler 00:14:13.620 --> 00:14:16.299 could be secretly modified to recognize when 00:14:16.299 --> 00:14:18.600 it was compiling two specific things. What were 00:14:18.600 --> 00:14:20.600 those two things? First, the source code for 00:14:20.600 --> 00:14:23.259 the unit's login program. When the compromised 00:14:23.259 --> 00:14:26.159 compiler saw that, it would secretly inject a 00:14:26.159 --> 00:14:28.460 backdoor, allowing a special password to grant 00:14:28.460 --> 00:14:31.600 access. A devastating attack on its own. But 00:14:31.600 --> 00:14:33.960 that's not the really chilling part, is it? No. 00:14:34.320 --> 00:14:36.860 The real genius, the part that gives the speech 00:14:36.860 --> 00:14:40.009 its title, is the second phase. He programmed 00:14:40.009 --> 00:14:42.649 the modified compiler to recognize when it was 00:14:42.649 --> 00:14:45.330 compiling itself. So when the compiler is compiling 00:14:45.330 --> 00:14:48.350 its own source code... It automatically injects 00:14:48.350 --> 00:14:50.870 the code that creates the login backdoor and 00:14:50.870 --> 00:14:53.750 the code that propagates the hack. So you could 00:14:53.750 --> 00:14:56.470 take the original, perfectly clean source code 00:14:56.470 --> 00:14:58.769 for the compiler, compile it with the infected 00:14:58.769 --> 00:15:01.509 compiler, and the new binary would also be infected. 00:15:01.710 --> 00:15:04.429 Wait, so the malicious code is no longer in... 00:15:04.620 --> 00:15:06.779 Any source file you can look at. It's only in 00:15:06.779 --> 00:15:08.899 the binary of the tool you're using to build 00:15:08.899 --> 00:15:11.259 everything. Exactly. You are trapped in a recursive 00:15:11.259 --> 00:15:13.879 cycle. To verify the compiler, you need a trusted 00:15:13.879 --> 00:15:16.600 compiler. But how do you trust that one? It proved 00:15:16.600 --> 00:15:18.259 that if you cannot trust your software supply 00:15:18.259 --> 00:15:20.460 chain, you can never fully trust your programs. 00:15:20.759 --> 00:15:22.940 And the annotated source code for this actual 00:15:22.940 --> 00:15:26.580 hack was finally published in 2023. It was. Decades 00:15:26.580 --> 00:15:28.899 later. Now, the good news is that this theoretical 00:15:28.899 --> 00:15:31.360 threat led directly to modern defense strategies, 00:15:31.580 --> 00:15:33.659 like something called diverse double compilation. 00:15:34.159 --> 00:15:36.120 How does that work? The idea is you take the 00:15:36.120 --> 00:15:37.899 source code for a compiler you want to trust 00:15:37.899 --> 00:15:40.100 and you compile it with two completely different 00:15:40.100 --> 00:15:43.299 independently verified compilers. If the two 00:15:43.299 --> 00:15:45.600 resulting binaries are bit for bit identical, 00:15:45.940 --> 00:15:48.139 you can have a very high degree of certainty 00:15:48.139 --> 00:15:51.399 that no malicious payload has been secretly injected. 00:15:51.840 --> 00:15:55.039 It breaks the cycle of trust. That is just a 00:15:55.039 --> 00:15:57.600 mind bending security concept to come out of 00:15:57.600 --> 00:16:00.299 an award speech. It is. He was also very critical 00:16:00.299 --> 00:16:02.860 of how the media portrayed hackers at the time. 00:16:03.200 --> 00:16:05.259 He clearly saw the negative potential of the 00:16:05.259 --> 00:16:07.720 tools he helped create. Yeah, he was acutely 00:16:07.720 --> 00:16:10.940 aware of that duality. Now, let's pivot entirely 00:16:10.940 --> 00:16:13.899 to a field that seems completely different, but 00:16:13.899 --> 00:16:17.080 where he also had a profound influence. Computer 00:16:17.080 --> 00:16:19.340 chess. This is one of my favorite parts of his 00:16:19.340 --> 00:16:22.019 story. His work in chess spans his entire career 00:16:22.019 --> 00:16:24.379 at Bell Labs. He wrote a chess program for the 00:16:24.379 --> 00:16:27.659 very first version of Unix. He did, back in 1971. 00:16:28.039 --> 00:16:30.649 Just called chess. But he didn't stop there. 00:16:30.769 --> 00:16:33.129 He later partnered with Joseph Condon to create 00:16:33.129 --> 00:16:35.210 a hardware -assisted program called Bell. And 00:16:35.210 --> 00:16:37.590 Bell was a world champion. It was a world champion 00:16:37.590 --> 00:16:40.269 chess computer. It proved that combining his 00:16:40.269 --> 00:16:43.450 deep algorithms with custom hardware could outperform 00:16:43.450 --> 00:16:46.549 pure software. It was a huge proof of concept 00:16:46.549 --> 00:16:49.980 for specialized AI hardware. But his most lasting 00:16:49.980 --> 00:16:52.059 contribution here, the thing that mathematically 00:16:52.059 --> 00:16:55.600 redefined computer chess, was his work on endgame 00:16:55.600 --> 00:16:57.820 table bases. This is where he went from search 00:16:57.820 --> 00:17:01.820 algorithms to pure brute force mathematics. An 00:17:01.820 --> 00:17:04.440 endgame table base is a complete map of every 00:17:04.440 --> 00:17:07.299 possible legal position for a fixed number of 00:17:07.299 --> 00:17:09.880 pieces on the board. He wrote programs that generated 00:17:09.880 --> 00:17:12.140 these for all four, five, and eventually six 00:17:12.140 --> 00:17:14.259 -piece endings. So what does that mean in practice? 00:17:14.700 --> 00:17:16.940 It means that for those positions, chess is solved. 00:17:17.339 --> 00:17:19.319 The computer doesn't have to think or search 00:17:19.319 --> 00:17:21.640 anymore. If it reaches a position that's in the 00:17:21.640 --> 00:17:23.880 table base, it knows with mathematical certainty 00:17:23.880 --> 00:17:26.640 the perfect move to guarantee a win or a draw. 00:17:26.799 --> 00:17:28.799 So it makes perfect moves. Perfect moves every 00:17:28.799 --> 00:17:31.759 time. This work revolutionized how grandmasters 00:17:31.759 --> 00:17:34.240 and computers analyzed the late game. And it 00:17:34.240 --> 00:17:36.759 was so significant that the ICGA journal devoted 00:17:36.759 --> 00:17:39.539 almost an entire issue to his contributions in 00:17:39.539 --> 00:17:42.779 2001. And as a kind of historical footnote, he 00:17:42.779 --> 00:17:46.009 even helped implement Turochamp. Yeah, Alan Turing's 00:17:46.009 --> 00:17:48.849 chess program from 1948, which Turing designed 00:17:48.849 --> 00:17:51.670 on paper before a computer even existed that 00:17:51.670 --> 00:17:54.170 could run it. Thompson helped bring that piece 00:17:54.170 --> 00:17:56.269 of history to life. It's a beautiful connection. 00:17:56.670 --> 00:17:59.549 So after the initial Unix revolution, Thompson 00:17:59.549 --> 00:18:02.150 and Ritchie didn't stop. They kept revising research 00:18:02.150 --> 00:18:05.069 Unix, but their focus eventually shifted to building 00:18:05.069 --> 00:18:08.720 a true modern replacement. Right. And this brings 00:18:08.720 --> 00:18:11.579 us to Plan 9 from Bell Labs in the mid -1980s. 00:18:11.759 --> 00:18:14.460 Thompson was a key designer, and the core idea 00:18:14.460 --> 00:18:17.059 was to take that Unix philosophy of everything 00:18:17.059 --> 00:18:19.940 is a file and apply it radically to everything, 00:18:20.059 --> 00:18:23.119 especially networking. So how did Plan 9 extend 00:18:23.119 --> 00:18:25.599 that philosophy? It extended it across the network. 00:18:25.720 --> 00:18:28.619 In Plan 9, if you wanted to access a remote resource, 00:18:28.940 --> 00:18:32.059 say a printer on another machine, you didn't 00:18:32.059 --> 00:18:34.119 use complex network protocols. You would simply 00:18:34.119 --> 00:18:36.079 mount that remote printer into your local file 00:18:36.079 --> 00:18:37.819 system. So to the program, it just looks like 00:18:37.819 --> 00:18:40.380 another local file. Exactly. The complexity of 00:18:40.380 --> 00:18:42.240 the distributed network was completely hidden 00:18:42.240 --> 00:18:44.960 behind that simple, unified abstraction. It was 00:18:44.960 --> 00:18:47.660 incredibly elegant. While Plan 9 never really 00:18:47.660 --> 00:18:49.980 took off commercially, its influence is everywhere. 00:18:50.430 --> 00:18:52.730 And that includes one of his most important inventions 00:18:52.730 --> 00:18:55.670 from that era. Yes, the foundational glue of 00:18:55.670 --> 00:18:59.930 the modern Internet, UTF -8. In 1992, Thompson 00:18:59.930 --> 00:19:03.039 developed the UTF -8 encoding scheme. with Rob 00:19:03.039 --> 00:19:05.519 Pike. And before UTF -8, character encoding was 00:19:05.519 --> 00:19:08.220 just a total mess. A total mess. You had ASCII 00:19:08.220 --> 00:19:11.079 for English, but then every region had its own 00:19:11.079 --> 00:19:13.859 incompatible standard. If you opened a document 00:19:13.859 --> 00:19:15.799 from another country, you'd just see garbage 00:19:15.799 --> 00:19:18.460 characters. The world needed a unified standard 00:19:18.460 --> 00:19:20.920 to handle Unicode, which covers every writing 00:19:20.920 --> 00:19:23.579 system. And the genius of UTF -8 is that it solved 00:19:23.579 --> 00:19:25.880 this without breaking the old internet. That's 00:19:25.880 --> 00:19:28.380 the key. UTF -8 is a variable with encoding. 00:19:29.000 --> 00:19:31.480 Any character from the old ASCII standard is 00:19:31.480 --> 00:19:33.680 represented by a single byte, just like it was 00:19:33.680 --> 00:19:36.599 before. So all existing English text files and 00:19:36.599 --> 00:19:38.759 protocols just worked. But it could also represent 00:19:38.759 --> 00:19:41.259 more complex characters, like from Asian languages, 00:19:41.380 --> 00:19:44.180 using two, three, or four bytes. This brilliant 00:19:44.180 --> 00:19:46.779 backward compatibility and efficiency made it 00:19:46.779 --> 00:19:49.480 the runaway global winner. The impact is just 00:19:49.480 --> 00:19:53.680 impossible to overstate. By 2019, over 90 % of 00:19:53.680 --> 00:19:57.019 all web pages were encoded in UTF -8. It's the 00:19:57.019 --> 00:19:59.720 silent infrastructure that holds the global multilingual 00:19:59.720 --> 00:20:03.220 Internet together. A massive, quiet victory for 00:20:03.220 --> 00:20:05.880 clean design. Now, before we leave Bell Labs, 00:20:05.960 --> 00:20:09.140 we have to revisit his famous and, let's say, 00:20:09.160 --> 00:20:12.980 colorful critique of the C++ programming language. 00:20:13.359 --> 00:20:17.200 Ah, yes. Yeah. The decisive rejection. He actually 00:20:17.200 --> 00:20:19.720 tested early versions of C++ for its creator, 00:20:19.880 --> 00:20:22.700 Bjorn Strostrup, so he knew it well. But he reached 00:20:22.700 --> 00:20:24.519 a point where he just flatly refused to work 00:20:24.519 --> 00:20:27.049 in it. Why? What was the main issue? She saw 00:20:27.049 --> 00:20:29.609 it as excessively complex, with frequent unnecessary 00:20:29.609 --> 00:20:32.250 incompatibilities between versions. It was a 00:20:32.250 --> 00:20:35.309 fundamental clash of worldviews. C++ tried to 00:20:35.309 --> 00:20:37.769 be everything to everyone. Thompson valued precision 00:20:37.769 --> 00:20:40.009 and minimalism. And he summed up his feelings 00:20:40.009 --> 00:20:42.869 in that famous 2009 interview. He did not mince 00:20:42.869 --> 00:20:45.849 words. He said C++ does a lot of things half 00:20:45.849 --> 00:20:47.890 well, and it's just a garbage heap of ideas that 00:20:47.890 --> 00:20:50.589 are mutually exclusive. A garbage heap. That 00:20:50.589 --> 00:20:52.890 quote really summarizes everything he spent his 00:20:52.890 --> 00:20:56.309 life fighting against. From Multics onward. And 00:20:56.309 --> 00:20:58.750 this rejection of complexity becomes the driving 00:20:58.750 --> 00:21:00.690 force behind the final chapter of his career. 00:21:00.849 --> 00:21:03.670 Absolutely. So in late 2000, Thompson retired 00:21:03.670 --> 00:21:06.490 from Bell Labs after almost 35 years. After a 00:21:06.490 --> 00:21:09.089 short stint elsewhere, he made a pivotal move 00:21:09.089 --> 00:21:11.809 to Google in 2006. Where he's been ever since, 00:21:11.890 --> 00:21:14.450 first as a distinguished engineer and now as 00:21:14.450 --> 00:21:17.069 a Google advisor. And this gave him a new environment 00:21:17.069 --> 00:21:19.990 and new challenges. Wow. The massive scale of 00:21:19.990 --> 00:21:22.369 modern cloud infrastructure. And he was joined 00:21:22.369 --> 00:21:24.509 there by his old colleague, Rob Pike. And it 00:21:24.509 --> 00:21:26.869 was at Google with Pike and Robert Griesmer that 00:21:26.869 --> 00:21:29.190 he co -developed the Go programming language. 00:21:29.650 --> 00:21:32.099 This feels like a der - correct spiritual successor 00:21:32.099 --> 00:21:34.539 to B and C, but for the modern world. That's 00:21:34.539 --> 00:21:36.960 exactly what it is. And the motivation, according 00:21:36.960 --> 00:21:39.140 to Thompson, was that deep dissatisfaction we 00:21:39.140 --> 00:21:40.859 just talked about. He said they got together 00:21:40.859 --> 00:21:44.480 and decided that we hated C++. But Go was born 00:21:44.480 --> 00:21:46.720 out of a desire for a systems language that was 00:21:46.720 --> 00:21:49.799 simple, efficient, and could handle modern concurrency. 00:21:50.220 --> 00:21:53.380 Right. They felt C++ was too slow to compile, 00:21:53.579 --> 00:21:56.319 too hard to debug, and its complexity led to 00:21:56.319 --> 00:21:58.720 all kinds of bugs. They wanted a language that 00:21:58.720 --> 00:22:01.160 compiled instantly, had built -in concurrency, 00:22:01.220 --> 00:22:03.660 and simplified memory management. They started 00:22:03.660 --> 00:22:07.700 it as pure research. So, to avoid creating another 00:22:07.700 --> 00:22:11.079 garbage heap, what was their core design rule? 00:22:11.579 --> 00:22:15.619 It was a democratic anti -complexity rule. Thompson 00:22:15.619 --> 00:22:18.259 said the key was that All three of us had to 00:22:18.259 --> 00:22:20.000 be talked into every feature in the language 00:22:20.000 --> 00:22:22.700 so there was no extraneous garbage put into the 00:22:22.700 --> 00:22:25.059 language for any reason. That three -way consensus 00:22:25.059 --> 00:22:27.980 must have resulted in a brutally lean language. 00:22:28.359 --> 00:22:30.700 It did. Go is famous for the features it doesn't 00:22:30.700 --> 00:22:33.039 have. But what it does have, fast compilation 00:22:33.039 --> 00:22:35.980 and an elegant model for concurrency using goroutines 00:22:35.980 --> 00:22:38.440 and channels, makes it perfect for building cloud 00:22:38.440 --> 00:22:40.640 infrastructure. And that concurrency model, that 00:22:40.640 --> 00:22:43.079 seems to echo his earlier work, doesn't it? It 00:22:43.079 --> 00:22:45.940 absolutely does. Gortines are like lightweight, 00:22:46.099 --> 00:22:48.660 independent processes echoing the modularity 00:22:48.660 --> 00:22:51.259 of Unix. And channels, which are how they communicate, 00:22:51.500 --> 00:22:54.519 are the spiritual successors to Unix pipes. It's 00:22:54.519 --> 00:22:57.259 his signature principles, simplicity, modularity, 00:22:57.299 --> 00:23:00.380 efficiency spanning six decades, and culminating 00:23:00.380 --> 00:23:02.480 in a language for today's distributed world. 00:23:02.779 --> 00:23:05.220 Given the incredible impact we've covered, it's 00:23:05.220 --> 00:23:07.480 no surprise that Thompson has received basically 00:23:07.480 --> 00:23:10.299 every major award in computer science. And often 00:23:10.299 --> 00:23:12.740 jointly with Dennis Ritchie. They received the 00:23:12.740 --> 00:23:16.240 Turing Award in 1983. The Nobel Prize of Computing. 00:23:16.279 --> 00:23:18.759 Yeah. The citation was for their development 00:23:18.759 --> 00:23:21.539 of generic operating systems theory and specifically 00:23:21.539 --> 00:23:24.000 for the implementation of the UNIX operating 00:23:24.000 --> 00:23:26.740 system. And it's striking how often the OS and 00:23:26.740 --> 00:23:30.200 the language are paired. In 1990, the IE Hamming 00:23:30.200 --> 00:23:33.019 Medal was for the origination of the UNIX operating 00:23:33.019 --> 00:23:35.900 system and the C programming language. The two 00:23:35.900 --> 00:23:37.980 are just inseparable. And the recognition came 00:23:37.980 --> 00:23:39.519 early. He was elected to the National Academy 00:23:39.519 --> 00:23:42.500 of Engineering in 1980 just for designing UNIX. 00:23:42.799 --> 00:23:44.079 And then there's the National Medal. medalist 00:23:44.079 --> 00:23:46.099 technology from President Bill Clinton in 1999. 00:23:46.599 --> 00:23:48.819 Yeah, that citation really captures the economic 00:23:48.819 --> 00:23:51.640 impact. It said they stimulated growth of an 00:23:51.640 --> 00:23:54.380 entire industry, thereby enhancing American leadership 00:23:54.380 --> 00:23:57.200 in the information age. That's not just a technical 00:23:57.200 --> 00:23:59.460 award. That's recognizing the creation of an 00:23:59.460 --> 00:24:02.259 entire economic sector. He also received major 00:24:02.259 --> 00:24:04.819 international honors like the Japan Prize in 00:24:04.819 --> 00:24:07.960 2011. The entire industry just recognized that 00:24:07.960 --> 00:24:10.140 he provided the template for all future systems. 00:24:10.319 --> 00:24:13.220 It's a... Truly unparalleled record of recognition. 00:24:13.559 --> 00:24:15.839 So let's end with a small personal tidbit that 00:24:15.839 --> 00:24:18.200 I think kind of humanizes this giant. We know 00:24:18.200 --> 00:24:21.880 he's married, has a son, but his recent choice 00:24:21.880 --> 00:24:24.539 in hardware feels like a nice full circle moment. 00:24:24.680 --> 00:24:27.579 It does. The sources say that despite the ubiquity 00:24:27.579 --> 00:24:29.910 of Apple products, which. ironically, are built 00:24:29.910 --> 00:24:33.029 on his BSD lineage. Thompson later switched from 00:24:33.029 --> 00:24:36.029 Apple to using Raspberry Pi OS. A move from the 00:24:36.029 --> 00:24:39.230 polished, closed ecosystem of Apple to this tiny, 00:24:39.309 --> 00:24:41.890 lightweight, open -sourced platform for hobbyists, 00:24:41.910 --> 00:24:43.890 that feels incredibly fitting. It's a perfect 00:24:43.890 --> 00:24:46.589 fit. It's a return to the source of his obsession. 00:24:47.450 --> 00:24:50.329 Small, simple, modular systems where the user 00:24:50.329 --> 00:24:53.509 has total control. The Raspberry Pi runs a derivative 00:24:53.509 --> 00:24:56.990 of Linux, which is a spiritual successor to Unix. 00:24:57.839 --> 00:25:00.220 It just shows that his lifelong commitment to 00:25:00.220 --> 00:25:02.940 elegant simplicity is still his guiding principle. 00:25:03.119 --> 00:25:05.559 So when we look back at the scope of Ken Thompson's 00:25:05.559 --> 00:25:07.640 career, the sheer density of his foundational 00:25:07.640 --> 00:25:11.200 work is, it's almost hard to believe. His work 00:25:11.200 --> 00:25:13.759 isn't just in the technology you use, it is the 00:25:13.759 --> 00:25:16.470 foundation. It's true infrastructure. He designed 00:25:16.470 --> 00:25:18.349 the architecture of Unix, which spawned Linux, 00:25:18.549 --> 00:25:21.970 macOS, Android. He created B, which led to C. 00:25:22.170 --> 00:25:24.950 He defined regular expressions in grep. He co 00:25:24.950 --> 00:25:27.589 -invented UTF -8. And he gave us one of the most 00:25:27.589 --> 00:25:29.970 terrifying concepts in security, the Thompson 00:25:29.970 --> 00:25:32.450 hack. It's a legacy of elegance and minimalism. 00:25:32.509 --> 00:25:34.670 He taught us that complex tasks should be done 00:25:34.670 --> 00:25:37.190 by combining small, honest tools. And it brings 00:25:37.190 --> 00:25:39.029 us right back to that idea of foundational trust. 00:25:39.519 --> 00:25:41.640 The Thompson hack proved that trust in technology 00:25:41.640 --> 00:25:43.700 is recursive. You have to trust the tool that 00:25:43.700 --> 00:25:46.200 built the tool. And his own creations, Unix, 00:25:46.299 --> 00:25:48.839 see Go form the base layer of that digital trust 00:25:48.839 --> 00:25:51.339 for so much of the world. So, as you go about 00:25:51.339 --> 00:25:54.619 your day, relying on systems built on Unix, using 00:25:54.619 --> 00:25:57.839 software compiled by C, consider this. The complexity 00:25:57.839 --> 00:26:00.279 of modern software is exponentially greater than 00:26:00.279 --> 00:26:03.200 anything Thompson faced on that PDP -7. We're 00:26:03.200 --> 00:26:05.559 relying on layers upon layers of compiled code. 00:26:05.779 --> 00:26:08.740 Layers built by tools we rarely, if ever, inspect. 00:26:09.180 --> 00:26:11.640 So given the lesson of the Thompson hack, what 00:26:11.640 --> 00:26:14.059 are the modern, subtle compiler backdoors or 00:26:14.059 --> 00:26:16.019 supply chain vulnerabilities that we might be 00:26:16.019 --> 00:26:18.740 trusting today without even realizing it? What 00:26:18.740 --> 00:26:21.259 is the next Thompson hack waiting to be discovered 00:26:21.259 --> 00:26:24.460 inside the very tools we assume are honest? A 00:26:24.460 --> 00:26:26.200 profound thought on the architecture of trust 00:26:26.200 --> 00:26:28.640 itself. Thank you for joining us on this deep 00:26:28.640 --> 00:26:30.859 dive into the indispensable foundation of modern 00:26:30.859 --> 00:26:33.980 computing. Welcome to the debate. Today we are 00:26:33.980 --> 00:26:36.799 diving into the intellectual output of, well, 00:26:37.099 --> 00:26:39.619 one of the 20th century's most profound computer 00:26:39.619 --> 00:26:42.660 scientists, Kenneth Lane Thompson. His career 00:26:42.660 --> 00:26:46.359 is just immense. It spans everything from foundational 00:26:46.359 --> 00:26:49.160 operating systems to, you know, revolutionary 00:26:49.160 --> 00:26:53.220 security theory and modern language design. Exactly. 00:26:53.440 --> 00:26:56.059 And that sheer scope forces a really critical 00:26:56.059 --> 00:26:59.339 question. Which body of work truly defines his 00:26:59.339 --> 00:27:01.960 most significant legacy? Is it the foundational 00:27:01.960 --> 00:27:04.920 systems and languages he first architected? I 00:27:04.920 --> 00:27:07.000 mean, the very bedrock of modern computing for 00:27:07.000 --> 00:27:09.640 which he got global acclaim? Or is it his later 00:27:09.640 --> 00:27:12.220 more critical evolutionary contributions that 00:27:12.220 --> 00:27:14.799 actually sought to refine, or in some cases outright 00:27:14.799 --> 00:27:17.180 replace, those very foundations? And that's the 00:27:17.180 --> 00:27:19.619 heart of it. I hold that Thomson's foundational 00:27:19.619 --> 00:27:23.079 work, anchored by Unix and its precursors, constitutes 00:27:23.079 --> 00:27:26.519 his immovable, defining legacy. And I come at 00:27:26.519 --> 00:27:29.420 it from a different way. While we obviously can't 00:27:29.420 --> 00:27:32.140 deny the initial success of those early systems, 00:27:32.670 --> 00:27:35.250 I'm going to argue that his enduring intellectual 00:27:35.250 --> 00:27:38.750 significance is really defined by his willingness 00:27:38.750 --> 00:27:41.450 to challenge them, to critique their paradigms 00:27:41.450 --> 00:27:43.789 and offer intellectually superior alternatives. 00:27:44.069 --> 00:27:46.690 His critical contributions, whether it's the 00:27:46.690 --> 00:27:49.589 Thompson hack or the clean slate approach of 00:27:49.589 --> 00:27:52.710 Plan 9 and Go, they reveal a deeper evolutionary 00:27:52.710 --> 00:27:56.890 genius. The sheer scale and pervasiveness of 00:27:56.890 --> 00:27:59.930 Thompson's initial works firmly established them 00:27:59.930 --> 00:28:02.130 as the unavoidable centerpiece of his career. 00:28:02.589 --> 00:28:05.710 He's the systems engineer who fundamentally redefined 00:28:05.710 --> 00:28:08.390 how people interact with computers. I mean, you 00:28:08.390 --> 00:28:11.690 have to consider the context. After working on 00:28:11.690 --> 00:28:14.890 Multics, Thompson needed a simple platform. So 00:28:14.890 --> 00:28:18.589 he designed and implemented the original Unix 00:28:18.589 --> 00:28:21.529 operating system on an aging PDP -7 machine. 00:28:22.009 --> 00:28:24.769 This seemingly small project provided the conceptual 00:28:24.769 --> 00:28:27.390 model for, well, all modern operating systems. 00:28:27.569 --> 00:28:30.390 The hierarchical file system, treating devices 00:28:30.390 --> 00:28:33.150 as files, and most crucially, the elegance of 00:28:33.150 --> 00:28:35.849 pipes for inter -process communication. This 00:28:35.849 --> 00:28:38.089 wasn't just a feature, it was a philosophy, you 00:28:38.089 --> 00:28:40.450 know? It drove the shell -scripting culture that 00:28:40.450 --> 00:28:42.829 persists today. He also invented the B programming 00:28:42.829 --> 00:28:45.650 language, the direct and necessary precursor 00:28:45.650 --> 00:28:49.000 to C. This pioneering status is... Well, it's 00:28:49.000 --> 00:28:51.380 validated by the highest professional honors, 00:28:51.519 --> 00:28:54.500 including the 1983 Turing Award and the National 00:28:54.500 --> 00:28:57.160 Medal of Technology, all specifically citing 00:28:57.160 --> 00:29:00.720 his essential work on Unix and that related language 00:29:00.720 --> 00:29:03.859 infrastructure. That foundational period set 00:29:03.859 --> 00:29:07.099 the irreversible trajectory of computing. I agree 00:29:07.099 --> 00:29:10.420 that early work was essential, of course, but 00:29:10.420 --> 00:29:14.099 sometimes the highest acclaim precedes the deepest 00:29:14.099 --> 00:29:17.450 insight. I'm going to argue that Thompson's intellectual 00:29:17.450 --> 00:29:20.789 zenith is really defined by his willingness to 00:29:20.789 --> 00:29:23.690 challenge those very essential paradigms, even 00:29:23.690 --> 00:29:26.970 the ones he created. The most sophisticated contribution 00:29:26.970 --> 00:29:31.190 he ever made is arguably his 1983 Turning Award 00:29:31.190 --> 00:29:34.309 acceptance speech, Reflections on Trusting Trust. 00:29:34.589 --> 00:29:37.710 So he used the platform recognizing his implementation 00:29:37.710 --> 00:29:41.029 achievements to introduce the conceptual impossibility 00:29:41.029 --> 00:29:44.019 of perfect trust in the compiler chain. what 00:29:44.019 --> 00:29:46.700 we now call the Thompson hack. This isn't just 00:29:46.700 --> 00:29:49.200 system design. It's seminal computer security 00:29:49.200 --> 00:29:52.339 philosophy. It demonstrates a profound, almost 00:29:52.339 --> 00:29:54.859 abstract understanding of systemic vulnerability 00:29:54.859 --> 00:29:57.900 that goes far beyond any single kernel implementation. 00:29:58.359 --> 00:30:01.160 And look, his subsequent career is a cycle of 00:30:01.160 --> 00:30:04.259 refinement. Plan 9 from Bell Labs was designed 00:30:04.259 --> 00:30:07.500 as an explicit replacement for Unix, applying 00:30:07.500 --> 00:30:09.680 its principles more broadly and consistently. 00:30:10.329 --> 00:30:12.730 And his motivation for co -developing the modern 00:30:12.730 --> 00:30:16.109 Go language, it stemmed from his powerful and 00:30:16.109 --> 00:30:20.069 highly critical view of C++, which he notoriously 00:30:20.069 --> 00:30:23.009 called a garbage heap of ideas that are mutually 00:30:23.009 --> 00:30:26.890 exclusive. I mean, his later work consistently 00:30:26.890 --> 00:30:30.170 drove computing forward by eliminating what he 00:30:30.170 --> 00:30:32.869 saw as intellectual flaws in the dominant systems 00:30:32.869 --> 00:30:35.509 of the day. OK, let's delve into that Turing 00:30:35.509 --> 00:30:38.049 Award moment, because I think it acts as a perfect 00:30:38.049 --> 00:30:40.940 fault line for this debate. I maintain that the 00:30:40.940 --> 00:30:42.940 award recognized the functional implementation 00:30:42.940 --> 00:30:46.039 of Unix and generic operating systems theory. 00:30:46.400 --> 00:30:48.839 The fact that his greatest security insight, 00:30:49.039 --> 00:30:51.200 the Thompson hack, was delivered in the acceptance 00:30:51.200 --> 00:30:54.259 speech for Unix, well, that just proves that 00:30:54.259 --> 00:30:56.299 the foundational work provided the necessary 00:30:56.299 --> 00:30:59.400 cultural and architectural platform for the critical 00:30:59.400 --> 00:31:02.819 insight to even happen. The success of Unix was 00:31:02.819 --> 00:31:05.859 a prerequisite. The achievement is the robust, 00:31:05.880 --> 00:31:08.440 elegant system that enabled a generation of computing. 00:31:08.960 --> 00:31:12.680 The hack is conceptually a profound commentary 00:31:12.680 --> 00:31:15.839 on the limitations and unavoidable security assumptions 00:31:15.839 --> 00:31:18.299 in that system. It can't really stand without 00:31:18.299 --> 00:31:21.079 the implementation it critiques. I see why you 00:31:21.079 --> 00:31:23.980 think that. I do. And it's certainly true that 00:31:23.980 --> 00:31:26.720 the hack requires a compiler ecosystem to exist. 00:31:27.039 --> 00:31:30.940 But framing it as merely commentary, I think 00:31:30.940 --> 00:31:33.259 that diminishes the intellectual weight of trusting 00:31:33.259 --> 00:31:36.480 trust. The essay represents a conceptual breakthrough 00:31:36.480 --> 00:31:40.039 in computer security. that entirely transcends 00:31:40.039 --> 00:31:43.140 the specific technology of Unix or C. It's an 00:31:43.140 --> 00:31:45.680 insight into the supply chain paradox. How do 00:31:45.680 --> 00:31:48.059 you trust the tools that build the tools? That 00:31:48.059 --> 00:31:50.319 problem isn't limited to a Bell Labs kernel. 00:31:50.599 --> 00:31:53.539 It challenged the fundamental assumption of bootstrapping 00:31:53.539 --> 00:31:55.839 and trust in every contemporary software system. 00:31:56.440 --> 00:31:59.240 This abstract critical insight is a deeper and, 00:31:59.299 --> 00:32:01.559 I'd argue, more lasting intellectual contribution 00:32:01.559 --> 00:32:04.319 than the implementation of any single kernel, 00:32:04.440 --> 00:32:06.720 no matter how influential. I mean, the fact that 00:32:06.720 --> 00:32:08.920 the hack is consistently taught in university 00:32:08.920 --> 00:32:12.559 security courses today, 30 years after the PDP 00:32:12.559 --> 00:32:15.579 -7 became a museum piece, demonstrates its complete 00:32:15.579 --> 00:32:18.039 independence from the initial foundational work 00:32:18.039 --> 00:32:20.720 you prize. It's applied theoretical computer 00:32:20.720 --> 00:32:23.740 science. It was a conceptual bomb dropped on 00:32:23.740 --> 00:32:27.160 the entire field. But that theoretical bomb only 00:32:27.160 --> 00:32:30.259 became potent because Thompson first demonstrated 00:32:30.259 --> 00:32:33.660 the power of a highly permeable, extensible architecture. 00:32:33.880 --> 00:32:36.980 You highlight Plan 9 as an intellectual correction, 00:32:37.259 --> 00:32:40.319 but it failed to achieve ubiquity. The intellectual 00:32:40.319 --> 00:32:43.319 success of a systems architect has to be measured, 00:32:43.359 --> 00:32:45.819 at least in part, by successful proliferation. 00:32:46.259 --> 00:32:50.200 Ah, but see, ubiquity is a weak measure of intellectual 00:32:50.200 --> 00:32:54.019 rigor. particularly for a mind as uncompromising 00:32:54.019 --> 00:32:56.880 as Thomson's. I'm sorry, but I just don't buy 00:32:56.880 --> 00:33:00.380 that. Thomson's later projects, Plan 9 and Go, 00:33:00.619 --> 00:33:03.920 were explicit intellectual corrections motivated 00:33:03.920 --> 00:33:07.079 by a commitment to philosophical purity and rigor. 00:33:07.579 --> 00:33:10.240 Plan 9 aimed to improve upon the architectural 00:33:10.240 --> 00:33:13.740 failings of Unix, specifically by replacing its 00:33:13.740 --> 00:33:16.400 monolithic kernel with a microkernel philosophy. 00:33:17.069 --> 00:33:19.269 and applying the everything is a file concept 00:33:19.269 --> 00:33:22.289 to network services, making it a true distributed 00:33:22.289 --> 00:33:25.130 operating system. This was a move toward architectural 00:33:25.130 --> 00:33:28.470 sophistication. Similarly, the Go language was 00:33:28.470 --> 00:33:30.890 born out of a profound intellectual distaste 00:33:30.890 --> 00:33:34.160 for the complexity in languages like C++. the 00:33:34.160 --> 00:33:36.799 headers, the implicit conversions, the difficulty 00:33:36.799 --> 00:33:39.440 in parallel processing. Thompson himself said 00:33:39.440 --> 00:33:42.299 the goal of Go was to ensure there was no extraneous 00:33:42.299 --> 00:33:44.839 garbage put into the language for any reason, 00:33:44.980 --> 00:33:47.619 simplifying concurrency with Go routines and 00:33:47.619 --> 00:33:51.519 channels. This commitment to conceptual clarity, 00:33:51.740 --> 00:33:55.640 to anti -complexity, even at the cost of immediate 00:33:55.640 --> 00:33:59.259 ubiquity. Well, that demonstrates a more mature 00:33:59.259 --> 00:34:02.279 and rigorous intellectual standard than simply 00:34:02.279 --> 00:34:05.079 building something that works and spreads. That's 00:34:05.079 --> 00:34:07.660 a compelling argument regarding purity. I'll 00:34:07.660 --> 00:34:10.420 give you that. But ubiquity is the measure of 00:34:10.420 --> 00:34:13.360 elegance in systems engineering. When a principle 00:34:13.360 --> 00:34:16.440 spreads globally, it proves its resilience and 00:34:16.440 --> 00:34:19.260 its inherent correctness. His early work wasn't 00:34:19.260 --> 00:34:21.659 just functional. It defined elegance through 00:34:21.659 --> 00:34:24.099 pervasive simplicity. Look at his work on the 00:34:24.099 --> 00:34:28.030 CTSS editor. QED, which led directly to the editor 00:34:28.030 --> 00:34:31.050 Ed on Unix. This work established the notation 00:34:31.050 --> 00:34:33.650 for regular expressions, a system of pattern 00:34:33.650 --> 00:34:35.809 matching that is now just woven into the fabric 00:34:35.809 --> 00:34:38.750 of text processing across every single platform. 00:34:39.130 --> 00:34:42.889 Unix and its derivative, BSD, demonstrably took 00:34:42.889 --> 00:34:45.530 over the academic community, defining how multiple 00:34:45.530 --> 00:34:47.510 generations of programmers learned to think. 00:34:47.670 --> 00:34:50.690 His core legacy is tied to the successful proliferation 00:34:50.690 --> 00:34:53.610 of his initial brilliant simple design principles, 00:34:53.809 --> 00:34:57.030 principles that, crucially, even Plan 9 and Inferno 00:34:57.030 --> 00:35:00.070 adopted, proving their foundational nature. So 00:35:00.070 --> 00:35:03.000 his critiques, while sharp... have not in fact 00:35:03.000 --> 00:35:05.719 replaced the foundational architecture he established. 00:35:06.019 --> 00:35:09.019 The concepts of the Unix shell, the file system, 00:35:09.260 --> 00:35:12.260 pipes, they remain the dominant conceptual model 00:35:12.260 --> 00:35:15.300 globally. I think you're confusing the success 00:35:15.300 --> 00:35:19.239 of a few key ideas, which certainly persisted, 00:35:19.239 --> 00:35:22.719 I agree, with the comprehensive triumph of the 00:35:22.719 --> 00:35:25.980 entire Unix kernel model. The industry didn't 00:35:25.980 --> 00:35:29.260 just stop evolving in 1980. We have to look beyond 00:35:29.260 --> 00:35:32.119 the Bell Labs operating systems to see his continued 00:35:32.119 --> 00:35:35.119 critical evolution. Agreed. So let's look at 00:35:35.119 --> 00:35:37.599 his independent technical contributions that 00:35:37.599 --> 00:35:40.000 demonstrate that same pioneering foundational 00:35:40.000 --> 00:35:43.480 drive. We have to include his seminal work on 00:35:43.480 --> 00:35:46.440 computer chess. Thompson wrote the program Chess 00:35:46.440 --> 00:35:49.980 for the first version of Unix back in 1971. More 00:35:49.980 --> 00:35:53.099 importantly, he and Joseph Condon later created 00:35:53.099 --> 00:35:56.280 Bell, a dedicated hardware -assisted chess computer 00:35:56.280 --> 00:35:59.380 that became a world champion program. He also 00:35:59.380 --> 00:36:02.460 undertook this monumental, just exhaustive computational 00:36:02.460 --> 00:36:05.719 task of generating the comprehensive endgame 00:36:05.719 --> 00:36:08.519 table bases for all four-, five-, and six -piece 00:36:08.519 --> 00:36:11.360 endings. This allowed chess programs to achieve 00:36:11.820 --> 00:36:15.280 perfect play in those scenarios. These are tangible, 00:36:15.380 --> 00:36:18.440 self -contained, high -impact technical achievements, 00:36:18.739 --> 00:36:21.360 unprecedented explorations into computational 00:36:21.360 --> 00:36:24.420 limits, separate entirely from the OS debate. 00:36:24.659 --> 00:36:27.679 They cement his status as an unparalleled technical 00:36:27.679 --> 00:36:31.179 pioneer whose ability to build robust, high -performance 00:36:31.179 --> 00:36:33.599 systems defined the first half of his career. 00:36:33.920 --> 00:36:36.739 While the chess work and the sheer computational 00:36:36.739 --> 00:36:39.920 mastery it required are, yeah, highly impressive, 00:36:40.460 --> 00:36:42.880 I'm not convinced by that line of reasoning because 00:36:42.880 --> 00:36:46.820 his work on UTF -8 in 1992 developed with Rob 00:36:46.820 --> 00:36:50.059 Pike is perhaps his single most dominant and 00:36:50.059 --> 00:36:53.780 impactful contribution in the 21st century. The 00:36:53.780 --> 00:36:56.440 creation of Bell demonstrated mastery in a niche. 00:36:56.579 --> 00:37:00.739 UTF -8 solved a critical global modern computing 00:37:00.739 --> 00:37:04.119 problem of character encoding in a backward compatible 00:37:04.119 --> 00:37:12.110 and highly efficient way. for over 90 % of the 00:37:12.110 --> 00:37:16.150 world wide web by the late 2010s. This late career 00:37:16.150 --> 00:37:19.610 creation of a global system standard proves his 00:37:19.610 --> 00:37:22.429 capacity for evolutionary critical contribution 00:37:22.429 --> 00:37:26.590 long after the initial Bell Labs period. It shows 00:37:26.590 --> 00:37:29.429 that his intellectual peak was sustained and 00:37:29.429 --> 00:37:33.510 evolved to address new global challenges, influencing 00:37:33.510 --> 00:37:37.250 systems used by billions today far more directly 00:37:37.250 --> 00:37:40.869 than that foundational PDP -7 code. UTF -8 is 00:37:40.869 --> 00:37:43.769 certainly ubiquitous, no question, but it is 00:37:43.769 --> 00:37:46.130 ultimately a standard born out of necessity. 00:37:46.449 --> 00:37:49.170 It was a fix for limitations imposed by earlier 00:37:49.170 --> 00:37:52.690 character encodings. Unix, however, was a creative 00:37:52.690 --> 00:37:55.690 synthesis of fundamental ideas that defined the 00:37:55.690 --> 00:37:58.070 environment in which that necessity later arose. 00:37:58.630 --> 00:38:01.309 Without the foundational platform, the flexible 00:38:01.309 --> 00:38:04.530 file system, the modular design, Thompson would 00:38:04.530 --> 00:38:06.849 never have had the cultural space to offer his 00:38:06.849 --> 00:38:09.730 later critiques or refinements. The initial design 00:38:09.730 --> 00:38:12.389 was the prerequisite for the entire body of work 00:38:12.389 --> 00:38:15.190 that followed. See, I think the real prerequisite 00:38:15.190 --> 00:38:17.849 was Thompson's singular intellectual clarity. 00:38:18.389 --> 00:38:21.289 The fact that he was willing to accept the failures 00:38:21.289 --> 00:38:24.130 of his own foundational work is the mark of a 00:38:24.130 --> 00:38:27.110 true intellectual giant. He didn't rest on the 00:38:27.110 --> 00:38:29.869 laurels of Unix. He sought its structural limitations 00:38:29.869 --> 00:38:32.570 in a distributed world, which led to Plan 9, 00:38:32.710 --> 00:38:35.269 and its linguistic inadequacies for modern development, 00:38:35.429 --> 00:38:38.679 which led to Go. The elegance you praise in eunuchs 00:38:38.679 --> 00:38:41.659 is undeniable, but the courage required to look 00:38:41.659 --> 00:38:43.760 at the highest achievement of your career and 00:38:43.760 --> 00:38:46.900 say that is no longer good enough? That is where 00:38:46.900 --> 00:38:49.619 the lasting significance truly lies. It's the 00:38:49.619 --> 00:38:52.079 critical evolutionary impulse that drove his 00:38:52.079 --> 00:38:54.820 most sophisticated work. Thompson's incredible 00:38:54.820 --> 00:38:58.000 range makes defining a single legacy incredibly 00:38:58.000 --> 00:39:01.840 difficult, absolutely, but the sheer volume and 00:39:01.840 --> 00:39:03.840 the fundamental nature of his early successes 00:39:04.480 --> 00:39:08.079 Unix, the B language, the birth of regular expressions, 00:39:08.380 --> 00:39:11.059 and the subsequent avalanche of awards recognizing 00:39:11.059 --> 00:39:14.340 this work set the irreversible path for modern 00:39:14.340 --> 00:39:17.579 computing. He is primarily the pioneer who built 00:39:17.579 --> 00:39:20.599 the ground on which all subsequent systems, even 00:39:20.599 --> 00:39:23.380 his own later refinements, were forced to stand. 00:39:23.500 --> 00:39:26.219 He laid the foundation. I would argue his career 00:39:26.219 --> 00:39:29.480 is best characterized not by a foundation, but 00:39:29.480 --> 00:39:32.420 as a continuous intellectual process of building. 00:39:32.960 --> 00:39:36.019 Critiquing and correcting. His willingness to 00:39:36.019 --> 00:39:38.579 expose fundamental security paradoxes in his 00:39:38.579 --> 00:39:41.119 own system with the Thompson hack, his explicit 00:39:41.119 --> 00:39:43.400 rejection of complex dominant languages like 00:39:43.400 --> 00:39:46.639 C++, and his ongoing quest for architectural 00:39:46.639 --> 00:39:50.400 improvements embodied in Plan 9, Go, and UTF 00:39:50.400 --> 00:39:53.340 -8. They represent an uncompromising intellectual 00:39:53.340 --> 00:39:56.599 commitment to truth and precision. That relentless 00:39:56.599 --> 00:39:58.920 pursuit of conceptual clarity and system evolution 00:39:58.920 --> 00:40:01.960 is arguably more important than the successful 00:40:01.960 --> 00:40:04.579 proliferation of any single early implementation. 00:40:05.059 --> 00:40:07.860 His most profound work lies in the conceptual 00:40:07.860 --> 00:40:11.239 critiques and evolutionary designs that continually 00:40:11.239 --> 00:40:13.960 force the industry to mature and improve upon 00:40:13.960 --> 00:40:16.760 his own initial brilliance. We can certainly 00:40:16.760 --> 00:40:19.260 agree that Kenneth Lane Thompson is one of the 00:40:19.260 --> 00:40:21.820 greatest computer programmers of all time, whose 00:40:21.820 --> 00:40:24.139 career demands continued scrutiny and appreciation 00:40:24.139 --> 00:40:26.780 for its dual nature, this foundational creation 00:40:26.780 --> 00:40:29.820 on one hand and relentless intellectual refinement 00:40:29.820 --> 00:40:33.059 on the other. Right. The material shows that 00:40:33.059 --> 00:40:35.219 whether through his pioneering systems or his 00:40:35.219 --> 00:40:38.480 later, more critical contributions, his influence 00:40:38.480 --> 00:40:41.360 remains absolutely pervasive, regardless of which 00:40:41.360 --> 00:40:43.519 period we deem most significant in isolation.