text
stringlengths
0
2.46k
[565.82 --> 569.82] And it can also generate code that's kind of tuned to the specific patterns in your code
[569.82 --> 574.26] base, not just the kind of like median stack overflow or open source code.
[574.70 --> 575.54] How did you get there?
[575.60 --> 576.70] How did you think, wow?
[576.78 --> 578.88] I mean, obviously, LLMs are a big deal.
[579.18 --> 579.36] Yep.
[579.36 --> 579.88] Right.
[579.88 --> 579.92] Right.
[580.02 --> 583.44] This new wave of intelligence that we have access to.
[584.24 --> 586.62] How far back is this in the making?
[586.74 --> 587.90] Has this been years?
[588.16 --> 590.72] Or has it been like, wow, Chet GPT is crazy.
[590.88 --> 591.18] November.
[591.38 --> 593.56] Chet GPT-3 is in November.
[593.56 --> 594.04] Okay.
[594.20 --> 594.88] We got to move.
[594.88 --> 596.38] How far back does this go?
[596.54 --> 596.70] Yeah.
[596.80 --> 597.22] Good question.
[597.36 --> 597.50] Yeah.
[597.54 --> 600.64] So for me personally, it's kind of a bit of a homecoming.
[600.84 --> 605.24] So like my first interest in computer science actually was machine learning and artificial
[605.24 --> 605.72] intelligence.
[605.72 --> 607.96] That's what I did a lot of my undergrad doing.
[607.96 --> 612.74] It was actually part of the Stanford AI lab doing vision research in those days under
[612.74 --> 614.12] Professor Daphne Kohler.
[614.22 --> 614.92] She's my advisor.
[615.70 --> 617.12] And so I did a lot of work there.
[617.22 --> 618.14] It was super interesting.
[618.56 --> 620.20] And I felt really passionate about it.
[620.26 --> 623.18] There's just a lot of elegant math that goes into things.
[623.18 --> 626.58] And it feels like you're kind of like poking at some of the hidden truths of the universe
[626.58 --> 627.44] a little bit.
[628.40 --> 632.86] But the technology at that point was just, it was nowhere near commercializable.
[633.58 --> 637.88] And so I decided to pursue my other passion, which is developing productivity and dev tools
[637.88 --> 642.40] and kind of like stayed on top of the research as it was coming along.
[642.58 --> 647.88] And I think one of the inflection points for us was the release of GPT-3 because that was
[647.88 --> 651.38] kind of a step function increase in the quality of the language models.
[651.38 --> 656.38] And we started to see some potential applications to developer tools and code.
[657.06 --> 662.50] And we really started in earnest maybe a little over a year ago, maybe 12 to 18 months ago,
[662.80 --> 667.76] experimenting with the kind of like internal representations of language models as a way
[667.76 --> 668.98] to enhance code search.
[668.98 --> 676.72] So we actually put out an experiment called code search.ai that uses embeddings to enhance
[676.72 --> 680.54] the quality of code search results that you get.
[680.74 --> 682.66] And that was pretty successful as an experiment.
[683.14 --> 687.28] I think we released that probably middle of last year, so about a year ago.
[687.82 --> 689.56] And that kind of started us down the road.
[689.56 --> 694.48] And then, of course, when ChatGPT came out, that was also another big inflection point.
[694.98 --> 700.52] And that's when we started to think very seriously about kind of like a chat-based interaction
[700.52 --> 705.72] that could happen in your editor, have all the advantages of ChatGPT, but know about the
[705.72 --> 706.96] specific context of your code.
[707.24 --> 713.18] And so for Cody specifically, I think first commit was December 1 or something like that.
[713.18 --> 718.14] By February, we basically had a version that we're having users and customers try.
[718.40 --> 721.30] And then March was when we rolled out to our first enterprise customer.
[721.52 --> 724.18] So it's just been like this whirlwind of development activity.
[725.56 --> 731.82] And I don't know, I cannot remember a time where I've been more excited and just eager
[731.82 --> 735.54] to build stuff because we're living through interesting times right now.
[735.64 --> 735.90] It is.
[736.02 --> 736.12] Yeah.
[736.20 --> 740.96] This is the eureka moment that we've all been waiting for, basically, right?
[740.96 --> 745.84] I mean, this is the invention of the internet all over again, potentially the iPhone level
[745.84 --> 746.82] invention.
[746.98 --> 752.02] I think it's a dramatic paradigm shift in how we think as engineers and software developers.
[752.30 --> 753.70] Like, how do we learn?
[753.98 --> 754.98] How do we leverage?
[755.22 --> 756.00] How do we augment?
[756.36 --> 756.44] Yeah.
[756.74 --> 762.26] You know, it's just insane what is available to somebody who doesn't have an understanding
[762.26 --> 766.92] to quickly get understanding and then be, you know, performant in a certain task or whatever
[766.92 --> 770.02] because of the LLMs that are available and how it works.
[770.02 --> 770.64] It's so crazy.
[770.64 --> 774.12] The chat interface is pretty simple though, right?
[774.22 --> 776.50] Like the simplicity of a chat interface.
[777.06 --> 781.24] Did you expect this eureka moment to be simply chat?
[781.80 --> 783.12] Like as you've been, you know what I mean?
[783.40 --> 785.18] Like it's a web app.
[785.42 --> 785.56] Yeah.
[785.68 --> 787.16] It's not something else.
[787.22 --> 788.30] It's a web interface.
[788.42 --> 789.42] It's a chat interface.
[789.84 --> 793.62] I think, so, you know, I'm a programmer by background.
[793.62 --> 798.34] So I've been like pushing, I've been trying to spread the gospel of textual based input
[798.34 --> 800.68] for, you know, as long as I can remember.
[800.86 --> 804.78] Obviously, it's mostly fallen on deaf ears because, you know, the non-programming world
[804.78 --> 806.92] is like, you know, command line.
[807.04 --> 808.22] That's what are we in?
[808.30 --> 809.56] Like the 1980s.
[809.56 --> 809.84] Right.
[811.04 --> 817.66] But I actually think philosophically, like textual input, the reason I like it is because
[817.66 --> 822.28] if you think about just like the IO, like bit rate of human computer interaction, it's
[822.28 --> 829.72] like we live in a time where like we have 4K screens running at, you know, 60 or 120
[829.72 --> 830.10] hertz.
[830.28 --> 835.98] Like the sheer amount of like data that computers can feed into us through our eyeballs is huge.
[836.10 --> 840.00] Whereas in kind of like the point and click, you know, mouse world, it's like how many bits
[840.00 --> 843.40] per second can you really feed into the computer as a human?
[843.62 --> 843.94] Right.
[843.94 --> 848.20] And now textual input, you know, it doesn't get us all the way there to, you know, 4K times,
[848.34 --> 849.62] you know, 60 hertz.
[849.62 --> 857.06] But it does, it basically like 10Xs or more like the input bit rate of what we can do to
[857.06 --> 857.74] instruct machines.