Transcription: OpenAgents Episode 065 - Exploring Code Llama 70B
June 9, 2026 ยท View on GitHub
Source: https://twitter.com/OpenAgentsInc/status/1752464706365755475 Wiki source: https://raw.githubusercontent.com/wiki/OpenAgentsInc/openagents/Video-Series.md Media title: OpenAgents - Episode 065: Exploring Code Llama 70B We build an agent step to analy... Upload date: 20240130 Transcription model: mlx-whisper/mlx-community/whisper-tiny Generated at: 2026-06-01T20:01:22Z
Machine-generated transcript. Review speaker labels and wording before using this as quote-grade source material.
[00:00] Christopher David: Yesterday, meta released Code Lama 70B, the new version of their Code Lama 34B, which they
[00:12] Christopher David: released previously, specific for Code Generation.
[00:16] Christopher David: Open model, meta has been crushing it on releasing open models, and we're going to take
[00:21] Christopher David: this first spin.
[00:22] Christopher David: And in the course of this, we'll get a little bit better of an idea of what it's like to
[00:27] Christopher David: build an agent that we actually care enough to use.
[00:31] Christopher David: The main use case that I'm focused on and that I suspect will be sort of the first
[00:36] Christopher David: class of truly great agents on open agents that I think people will be willing to pay
[00:43] Christopher David: for are agents that work in the context of a GitHub repo and can execute code, can
[00:54] Christopher David: code for you.
[00:59] Christopher David: So we did a bunch of work earlier in this series on GitHub agent, kind of the planning of
[01:07] Christopher David: that as an episode 20, we did a little bit of work on analyzing the initial code base.
[01:12] Christopher David: And now is we're kind of about to start finalizing the UI to release open agents
[01:20] Christopher David: as a product.
[01:22] Christopher David: This will be sort of the final video I think where we kind of build an agent sort of
[01:26] Christopher David: in the backend by kind of compiling some of those agent steps.
[01:31] Christopher David: Like you saw in the last few videos, we added support for new step types for wasom plugins
[01:39] Christopher David: and L402, payday PI payments.
[01:44] Christopher David: We're going to do whatever we need to do in this episode to kind of similarly build up
[01:47] Christopher David: an agent in the backend using the tasks and steps that can do something relevant using
[01:53] Christopher David: this code Lama model and we'll speculate specifically about that in a second.
[01:57] Christopher David: And then we will maybe next video kind of do the front end for that, making sure that
[02:04] Christopher David: people can do what I'm about to do in this video, all of the UI building agents with
[02:10] Christopher David: different building blocks via the UI and doing cool things.
[02:14] Christopher David: I want like a cool result that people are willing to pay for.
[02:18] Christopher David: So we can get some actual commerce flowing through here.
[02:21] Christopher David: One thing that has been missing in agents generally in or at least agents that you
[02:28] Christopher David: can kind of use on the internet is having an agent that actually does something that's
[02:34] Christopher David: truly productive that blows your your socks off in terms of, wow, this is something that
[02:39] Christopher David: I actually want to use repeatedly.
[02:42] Christopher David: The GPT is on the OpenAI GPT store.
[02:47] Christopher David: Largely our junk just kind of glorified chatbox with some prompt engineering.
[02:53] Christopher David: But I'm thinking that if we can deliver a junior developer that lives in your GitHub repo
[03:00] Christopher David: that could replace the kind of thing that you would typically pay someone, you know, $6,000
[03:05] Christopher David: month for a full time junior dev.
[03:07] Christopher David: If you could like instantly spin up 10 of these and pay, you know, a few hundred dollars
[03:13] Christopher David: at most per month for like 10 different junior devs.
[03:16] Christopher David: That would be really cool.
[03:18] Christopher David: That's what I want.
[03:19] Christopher David: That's what I want.
[03:21] Christopher David: And CodeLama probably is going to pay it playing a important role in that.
[03:24] Christopher David: So I haven't looked at CodeLama at all.
[03:26] Christopher David: I haven't even really used the previous one aside from playing with it a little bit.
[03:30] Christopher David: So in this video we're going to kind of go from scratch like I don't know what to do
[03:35] Christopher David: and we're going to pull it in and just see if we can get to a cool result by the end of the video.
[03:40] Christopher David: So let's review a little bit of what God posted.
[03:44] Christopher David: So AI at Meta says today we're releasing CodeLama 70B, a new more performance version of our
[03:49] Christopher David: LOM for Code Generation available under the same license as previous CodeLama models.
[03:54] Christopher David: Okay, so I have downloaded like the Lama two models in the past.
[04:00] Christopher David: It kind of played with them locally.
[04:04] Christopher David: That might be something that we do at some point.
[04:07] Christopher David: But it's cool that there's so many people, whenever there's a new release, they just
[04:10] Christopher David: jump on it and they throw it up on their hosting environment that they've tuned over
[04:14] Christopher David: months to be optimized for this.
[04:18] Christopher David: I'm just going to use their stuff because I'll let those people specialize.
[04:24] Christopher David: We'll take advantage of their specialty and that.
[04:26] Christopher David: We just want to use the thing.
[04:27] Christopher David: So just give me the thing where I can use.
[04:30] Christopher David: So this other AI product guy, Ding Board, some great tweets from Yesine Cash here, King of
[04:41] Christopher David: Ding.
[04:43] Christopher David: Ask where are you trying out CodeLama 70B on?
[04:47] Christopher David: He says he's using it on together compute.
[04:50] Christopher David: We may end up trying that first because that's, we already have an integration with
[04:54] Christopher David: together.
[04:55] Christopher David: We're using them for our, our mixed role.
[04:58] Christopher David: Based model with COT is pretty impressive.
[05:02] Christopher David: Labs up, replace the AI.
[05:09] Christopher David: Ethics block too high refused to help me with the Open AI API as that's confidential
[05:14] Christopher David: proprietary.
[05:17] Christopher David: Replicate.com.
[05:23] Christopher David: Locally, complexity, it's mid.
[05:37] Christopher David: Okay.
[05:38] Christopher David: One more thing before we dive in.
[05:41] Christopher David: Here's just an example of someone who has fine tuned codeLama specifically for SQL,
[05:49] Christopher David: specifically for Postgres.
[05:50] Christopher David: We're using Postgres.
[05:52] Christopher David: But apparently it does awesome.
[05:55] Christopher David: So one thing that I think that our network once we get open agents like fully released
[06:00] Christopher David: is going to be great for is whenever there's like a new thing, being able to rapidly pull
[06:05] Christopher David: in the new thing and thought it into existing workflows.
[06:10] Christopher David: Particularly once we can also start to incentivize people to maybe like fine tuned things
[06:13] Christopher David: for particular use cases and get it cut on the little blocks that get drawn into people's
[06:20] Christopher David: workflows like specifically in the service of the paid product that we have.
[06:26] Christopher David: We've got some paid like GitHub agent that someone's paying for and someone can see,
[06:31] Christopher David: oh wow, if I upgrade the you know, the text inference or the code generation of that particular
[06:36] Christopher David: piece, people use it and switch to that and get like a stream of the referring the recurring
[06:41] Christopher David: revenue of all of those uses.
[06:45] Christopher David: That'll be pretty powerful.
[06:47] Christopher David: Okay.
[06:49] Christopher David: So my book like that, I might check that out at some point.
[06:53] Christopher David: Okay, so let's take a spin over by that together API and see if they've got remember what
[07:07] Christopher David: a calendar used here.
[07:16] Christopher David: I might have used my new email address.
[07:18] Christopher David: Let's try that.
[07:43] Christopher David: Code Lama instruct is that what I want.
[07:56] Christopher David: Someone question is is there a instruct and is there like a chat version two so there's
[08:04] Christopher David: one other tweet that I saw prompting guide for code Lama.
[08:15] Christopher David: Okay.
[08:16] Christopher David: Yeah, code Lama 70 being struck with plenty of examples for developers and practitioners.
[08:20] Christopher David: Code Lama 70 being struck is one of the most powerful open source models for code generation.
[08:28] Christopher David: After which released yesterday, I felt we needed to develop a proper guide for it to learn
[08:31] Christopher David: how it could be used for all sorts of interesting code generation tasks like debugging and
[08:34] Christopher David: text to SQL generation.
[08:36] Christopher David: The guide includes the following.
[08:38] Christopher David: Configure model access, basic code completion, debugging, unit tests, text to SQL generation,
[08:43] Christopher David: if you should I proud you with Code Lama function calling safety guardrails node bug references.
[09:07] Christopher David: Okay, so we're doing the same stuff.
[09:15] Christopher David: You're going to give me the raw mark down to make it easy for me to paste.
[09:29] Christopher David: Code Lama dot n dot mdx.
[09:36] Christopher David: Okay.
[09:37] Christopher David: There will come a point soon when we will not be copy pasting anymore.
[09:46] Christopher David: We'll be like, hey agent, here's the URL.
[09:49] Christopher David: Here's our code base.
[09:51] Christopher David: You figure it out.
[09:54] Christopher David: Do some things.
[09:56] Christopher David: All right.
[09:58] Christopher David: Let's keep exercising our copy paste fingers.
[10:01] Christopher David: Funnel.
[10:02] Christopher David: Okay.
[10:03] Christopher David: I today I want your help to experiment with code Lama 70 being newly released yesterday
[10:18] Christopher David: from Mehta.
[10:22] Christopher David: Here is a prompting guide.
[10:24] Christopher David: I will separately follow up with code specific to the API we will use together, I think.
[10:33] Christopher David: And our existing PHP Laira bell code for working with that API.
[10:44] Christopher David: Start by summarizing this doc in one paragraph.
[10:57] Christopher David: The prompting guide for code Lama provides comprehensive guidance on how to use the code
[11:00] Christopher David: Lama family of large language models released by Mehta, specifically focusing on the code
[11:04] Christopher David: Lama 70 being a struct model.
[11:07] Christopher David: The guide covers various aspects including blah blah blah.
[11:13] Christopher David: All right.
[11:14] Christopher David: Always got to say something about safety and ethical considerations.
[11:17] Christopher David: Be at the core of your body.
[11:20] Christopher David: Okay.
[11:21] Christopher David: So where are some examples?
[11:36] Christopher David: The model can also be used to write unit tests.
[11:46] Christopher David: Ooh.
[11:53] Christopher David: One cool function would be analyze my code.
[11:59] Christopher David: I'm going to make a wiki for this.
[12:06] Christopher David: I'm going to call this auto dev scratchpad.
[12:20] Christopher David: So I'd wish list.
[12:25] Christopher David: Review my code base.
[12:28] Christopher David: Tell me what types of code don't have unit or feature tests and suggest which tests to write.
[12:48] Christopher David: Have me confirm that then write the tests and run the tests.
[12:56] Christopher David: We've got that flow already where it will run the test automatically if a PR is opened.
[13:08] Christopher David: So here specifically is the use case we want to solve now.
[13:25] Christopher David: Okay.
[13:54] Christopher David: Well that goes.
[14:04] Christopher David: Let me confirm in the together API docs that their docs support code Lama.
[14:19] Christopher David: Okay.
[14:32] Christopher David: Only a 2000 contacts length.
[14:34] Christopher David: That's interesting.
[14:44] Christopher David: All right.
[14:54] Christopher David: So together model string boom.
[14:57] Christopher David: Max context.
[15:02] Christopher David: I'm pretty sure based on the pace of open development that within the next two or three days.
[15:09] Christopher David: Okay.
[15:16] Christopher David: Here's where I want to introduce to Jackie B.T. my project and what we have going and what we have built already.
[15:23] Christopher David: And that will be done pretty soon with.
[15:27] Christopher David: Okay.
[15:36] Christopher David: What's your that paste here?
[15:49] Christopher David: I think I want to go to the concierge with plugin and L-frot to cedar.
[15:57] Christopher David: Give this as an example.
[16:07] Christopher David: Okay.
[16:11] Christopher David: So here is an example of an agent built up out of tasks and steps using retrieval augmented generation.
[16:26] Christopher David: With co-sign similarity search over a vectorized knowledge base.
[16:37] Christopher David: An example usage of a third party wasm plugin and paid L-frot to API input.
[16:49] Christopher David: And that we won't need here.
[16:59] Christopher David: Imagine we will use the PHP GitHub.
[17:04] Christopher David: Imagine we will use the GitHub API.
[17:07] Christopher David: Maybe via their PHP SDK to pull whatever information we need from a given repo.
[17:17] Christopher David: And I want to write a version of this same file for an auto-dead agent that will use code llama 70.
[17:32] Christopher David: I'll help massage that into proper class structure as we build it.
[17:42] Christopher David: I'll pass it in the same way.
[17:49] Christopher David: Maybe I don't.
[17:55] Christopher David: I might have blown up the context one.
[18:02] Christopher David: I'll show you the structure of your concierge with plugin and L-frot to cedar.
[18:09] Christopher David: Class auto-dead with co-lums 70.
[18:12] Christopher David: So validate code inputs.
[18:30] Christopher David: That's probably not.
[18:33] Christopher David: The input should be a command from the user.
[19:00] Christopher David: Or to keep things simple, maybe the input is just going to be a GitHub URL.
[19:12] Christopher David: Repove a GitHub URL.
[19:19] Christopher David: Thanks.
[19:20] Christopher David: Let's customize this in a few ways.
[19:27] Christopher David: Step one should be receiving input from the user in the form of a link to a GitHub repo.
[19:41] Christopher David: That should be a simple validation step.
[19:49] Christopher David: If you're not using a file, you can use a file to create a file.
[19:58] Christopher David: If you're using a file, you can use a file to create a file.
[20:03] Christopher David: If you're using a file, you can use a file to create a file.
[20:08] Christopher David: If you're using a file, you can use a file.
[20:11] Christopher David: If you're using a file, you can use a file.
[19:59] Christopher David: step two should be
[20:13] Christopher David: should be call the GitHub API to fetch the file folder structure or should we put a bunch of stuff into like a single block
[20:23] Christopher David: for like an all-analyze repo
[20:42] Christopher David: on its analyze the tests
[20:46] Christopher David: in relation to the rest of the code
[20:50] Christopher David: that's what we're building toward do I want to start with something simpler
[21:16] Christopher David: to
[21:23] Christopher David: to
[21:30] Christopher David: step two should be a code category
[21:41] Christopher David: called
[21:43] Christopher David: to
[21:50] Christopher David: to
[22:00] Christopher David: to
[22:10] Christopher David: to
[22:17] Christopher David: to
[22:27] Christopher David: to
[22:34] Christopher David: to
[22:37] Christopher David: to
[22:42] Christopher David: to
[22:49] Christopher David: to
[22:54] Christopher David: to
[22:57] Christopher David: to
[23:03] Christopher David: to
[23:10] Christopher David: to
[23:17] Christopher David: to
[23:24] Christopher David: to
[23:31] Christopher David: to
[23:34] Christopher David: to
[23:39] Christopher David: to
[23:46] Christopher David: to
[23:49] Christopher David: to
[23:54] Christopher David: to
[24:04] Christopher David: to
[24:11] Christopher David: to
[24:16] Christopher David: to
[24:19] Christopher David: to
[24:24] Christopher David: to
[24:29] Christopher David: to
[24:36] Christopher David: to
[24:39] Christopher David: to
[24:44] Christopher David: to
[24:48] Christopher David: to
[24:52] Christopher David: to
[24:59] Christopher David: to
[25:04] Christopher David: to
[25:07] Christopher David: to
[25:08] Christopher David: to
[25:13] Christopher David: to
[25:20] Christopher David: to
[25:23] Christopher David: to
[25:27] Christopher David: to
[25:28] Christopher David: to
[25:30] Christopher David: to
[25:34] Christopher David: to
[25:38] Christopher David: to
[25:45] Christopher David: to
[25:50] Christopher David: to
[25:53] Christopher David: to
[25:56] Christopher David: to
[26:00] Christopher David: to
[26:05] Christopher David: to
[26:12] Christopher David: to
[26:15] Christopher David: to
[26:19] Christopher David: to
[26:20] Christopher David: to
[26:26] Christopher David: to
[26:27] Christopher David: to
[26:30] Christopher David: to
[26:32] Christopher David: to
[26:39] Christopher David: to
[26:42] Christopher David: to
[26:46] Christopher David: to
[26:47] Christopher David: to
[26:51] Christopher David: to
[26:53] Christopher David: to
[26:57] Christopher David: to
[27:01] Christopher David: to
[27:06] Christopher David: to
[27:11] Christopher David: to
[27:15] Christopher David: to
[27:19] Christopher David: to
[27:22] Christopher David: to
[27:26] Christopher David: to
[27:31] Christopher David: to
[27:33] Christopher David: to
[27:36] Christopher David: to
[27:38] Christopher David: to
[27:41] Christopher David: to
[27:45] Christopher David: to
[27:47] Christopher David: to
[27:51] Christopher David: to
[27:53] Christopher David: to
[27:58] Christopher David: to
[28:00] Christopher David: to
[28:03] Christopher David: to
[28:07] Christopher David: to
[28:08] Christopher David: to
[28:12] Christopher David: to
[28:13] Christopher David: to
[28:18] Christopher David: to
[28:20] Christopher David: to
[28:23] Christopher David: to
[28:27] Christopher David: to
[28:32] Christopher David: to
[28:36] Christopher David: to
[28:37] Christopher David: to
[28:41] Christopher David: to
[28:45] Christopher David: to
[28:46] Christopher David: to
[28:50] Christopher David: to
[28:54] Christopher David: to
[28:56] Christopher David: to
[28:59] Christopher David: to
[29:01] Christopher David: to
[29:04] Christopher David: to
[29:05] Christopher David: to
[29:08] Christopher David: to
[29:10] Christopher David: to
[29:12] Christopher David: to
[29:14] Christopher David: to
[29:16] Christopher David: to
[29:21] Christopher David: to
[29:23] Christopher David: to
[29:26] Christopher David: to
[29:28] Christopher David: to
[29:31] Christopher David: to
[29:33] Christopher David: to
[29:35] Christopher David: to
[29:37] Christopher David: to
[29:39] Christopher David: to
[29:41] Christopher David: to
[29:43] Christopher David: to
[29:46] Christopher David: to
[29:48] Christopher David: to
[29:50] Christopher David: to
[29:52] Christopher David: to
[29:54] Christopher David: to
[29:57] Christopher David: to
[30:00] Christopher David: to
[30:01] Christopher David: to
[30:03] Christopher David: to
[30:05] Christopher David: to
[30:08] Christopher David: to
[30:10] Christopher David: to
[30:12] Christopher David: to
[30:15] Christopher David: to
[30:17] Christopher David: to
[30:18] Christopher David: to
[30:21] Christopher David: to
[30:22] Christopher David: to
[30:23] Christopher David: to
[30:24] Christopher David: to
[30:26] Christopher David: to
[30:28] Christopher David: to
[30:29] Christopher David: to
[30:30] Christopher David: to
[30:31] Christopher David: to
[30:32] Christopher David: to
[30:33] Christopher David: to
[30:35] Christopher David: to
[30:36] Christopher David: to
[30:37] Christopher David: to
[30:39] Christopher David: to
[30:41] Christopher David: to
[30:44] Christopher David: to
[30:45] Christopher David: to
[30:46] Christopher David: to
[30:48] Christopher David: to
[30:50] Christopher David: to
[30:52] Christopher David: to
[30:53] Christopher David: to
[30:55] Christopher David: to
[30:56] Christopher David: to
[30:57] Christopher David: to
[30:58] Christopher David: to
[30:59] Christopher David: to
[31:00] Christopher David: to
[31:01] Christopher David: to
[31:02] Christopher David: to
[31:03] Christopher David: to
[31:04] Christopher David: to
[31:05] Christopher David: to
[31:06] Christopher David: to
[31:08] Christopher David: to
[31:10] Christopher David: to
[31:11] Christopher David: to
[31:12] Christopher David: to
[31:13] Christopher David: to
[31:14] Christopher David: to
[31:15] Christopher David: to
[31:16] Christopher David: to
[31:17] Christopher David: to
[31:18] Christopher David: to
[31:19] Christopher David: to
[31:20] Christopher David: to
[31:21] Christopher David: to
[31:22] Christopher David: to
[31:23] Christopher David: to
[31:24] Christopher David: to
[31:25] Christopher David: to
[31:26] Christopher David: to
[31:27] Christopher David: to
[31:28] Christopher David: to
[31:29] Christopher David: to
[31:30] Christopher David: to
[31:31] Christopher David: to
[31:32] Christopher David: to
[31:33] Christopher David: to
[31:34] Christopher David: to
[31:35] Christopher David: to
[31:36] Christopher David: to
[31:37] Christopher David: to
[31:38] Christopher David: to
[31:39] Christopher David: to
[31:40] Christopher David: to
[31:42] Christopher David: to
[31:43] Christopher David: to
[31:44] Christopher David: to
[31:45] Christopher David: to
[31:46] Christopher David: to
[31:47] Christopher David: to
[31:48] Christopher David: to
[31:49] Christopher David: to
[31:50] Christopher David: to
[31:51] Christopher David: to
[31:52] Christopher David: to
[31:53] Christopher David: to
[31:54] Christopher David: to
[31:55] Christopher David: to
[31:56] Christopher David: to
[31:57] Christopher David: to
[31:58] Christopher David: to
[31:59] Christopher David: to
[32:00] Christopher David: to
[32:01] Christopher David: to
[32:02] Christopher David: to
[32:03] Christopher David: to
[32:04] Christopher David: to
[32:05] Christopher David: to
[32:06] Christopher David: to
[32:07] Christopher David: to
[32:08] Christopher David: to
[32:09] Christopher David: to
[32:11] Christopher David: to
[32:12] Christopher David: to
[32:13] Christopher David: to
[32:14] Christopher David: to
[32:15] Christopher David: to
[32:16] Christopher David: to
[32:17] Christopher David: to
[32:18] Christopher David: to
[32:19] Christopher David: to
[32:20] Christopher David: to
[32:21] Christopher David: to
[32:22] Christopher David: to
[32:23] Christopher David: to
[32:24] Christopher David: to
[32:25] Christopher David: to
[32:26] Christopher David: to
[32:27] Christopher David: to
[32:28] Christopher David: to
[32:29] Christopher David: to
[32:30] Christopher David: to
[32:31] Christopher David: to
[32:32] Christopher David: to
[32:33] Christopher David: to
[32:34] Christopher David: to
[32:35] Christopher David: to
[32:36] Christopher David: to
[32:37] Christopher David: to
[32:38] Christopher David: to
[32:40] Christopher David: to
[32:41] Christopher David: to
[32:42] Christopher David: to
[32:43] Christopher David: to
[32:44] Christopher David: to
[32:45] Christopher David: to
[32:47] Christopher David: to
[32:48] Christopher David: to
[32:49] Christopher David: to
[32:50] Christopher David: to
[32:51] Christopher David: to
[32:52] Christopher David: to
[32:53] Christopher David: to
[32:54] Christopher David: to
[32:55] Christopher David: to
[32:56] Christopher David: to
[32:57] Christopher David: to
[32:58] Christopher David: to
[32:59] Christopher David: to
[33:00] Christopher David: to
[33:01] Christopher David: to
[33:02] Christopher David: to
[33:03] Christopher David: to
[33:04] Christopher David: to
[33:05] Christopher David: to
[33:06] Christopher David: to
[33:07] Christopher David: to
[33:10] Christopher David: to
[33:11] Christopher David: to
[33:12] Christopher David: to
[33:13] Christopher David: to
[33:40] Christopher David: to
[33:41] Christopher David: to
[33:42] Christopher David: to
[33:43] Christopher David: to
[33:44] Christopher David: to
[33:45] Christopher David: to
[33:46] Christopher David: to
[33:47] Christopher David: to
[33:48] Christopher David: to
[33:49] Christopher David: to
[33:50] Christopher David: to
[33:51] Christopher David: to
[33:52] Christopher David: to
[33:53] Christopher David: to
[33:54] Christopher David: to
[33:55] Christopher David: to
[33:56] Christopher David: to
[33:57] Christopher David: to
[33:58] Christopher David: to
[33:59] Christopher David: to
[34:00] Christopher David: to
[34:01] Christopher David: to
[34:02] Christopher David: to
[34:03] Christopher David: to
[34:04] Christopher David: to
[34:05] Christopher David: to
[34:06] Christopher David: to
[34:07] Christopher David: to
[34:09] Christopher David: to
[34:10] Christopher David: to
[34:11] Christopher David: to
[34:12] Christopher David: to
[34:13] Christopher David: to
[34:14] Christopher David: to
[34:15] Christopher David: to
[34:16] Christopher David: to
[34:17] Christopher David: to
[34:18] Christopher David: to
[34:19] Christopher David: to
[34:20] Christopher David: to
[34:21] Christopher David: to
[34:22] Christopher David: to
[34:23] Christopher David: to
[34:24] Christopher David: to
[34:25] Christopher David: to
[34:26] Christopher David: to
[34:27] Christopher David: to
[34:28] Christopher David: to
[34:29] Christopher David: to
[34:30] Christopher David: to
[34:31] Christopher David: to
[34:32] Christopher David: to
[34:33] Christopher David: to
[34:34] Christopher David: to
[34:35] Christopher David: to
[34:36] Christopher David: to
[34:38] Christopher David: to
[34:39] Christopher David: to
[34:40] Christopher David: to
[34:41] Christopher David: to
[34:42] Christopher David: to
[34:43] Christopher David: to
[34:44] Christopher David: to
[34:45] Christopher David: to
[34:46] Christopher David: to
[34:47] Christopher David: to
[34:48] Christopher David: to
[34:49] Christopher David: to
[34:50] Christopher David: to
[34:51] Christopher David: to
[34:52] Christopher David: to
[34:53] Christopher David: to
[34:54] Christopher David: to
[34:55] Christopher David: to
[34:56] Christopher David: to
[34:57] Christopher David: to
[34:58] Christopher David: to
[34:59] Christopher David: to
[35:00] Christopher David: to
[35:01] Christopher David: to
[35:02] Christopher David: to
[35:03] Christopher David: to
[35:04] Christopher David: to
[35:05] Christopher David: to
[35:07] Christopher David: to
[35:08] Christopher David: to
[35:09] Christopher David: to
[35:10] Christopher David: to
[35:11] Christopher David: to
[35:12] Christopher David: to
[35:13] Christopher David: to
[35:14] Christopher David: to
[35:15] Christopher David: to
[35:16] Christopher David: to
[35:17] Christopher David: to
[35:18] Christopher David: to
[35:19] Christopher David: to
[35:20] Christopher David: to
[35:21] Christopher David: to
[35:22] Christopher David: to
[35:23] Christopher David: to
[35:24] Christopher David: to
[35:25] Christopher David: to
[35:26] Christopher David: to
[35:27] Christopher David: to
[35:28] Christopher David: to
[35:29] Christopher David: to
[35:30] Christopher David: to
[35:31] Christopher David: to
[35:32] Christopher David: to
[35:33] Christopher David: to
[35:34] Christopher David: to
[35:36] Christopher David: to
[35:37] Christopher David: to
[35:38] Christopher David: to
[35:39] Christopher David: to
[35:40] Christopher David: to
[35:41] Christopher David: to
[35:42] Christopher David: to
[35:43] Christopher David: to
[35:44] Christopher David: to
[35:45] Christopher David: to
[35:46] Christopher David: to
[35:47] Christopher David: to
[35:48] Christopher David: to
[35:49] Christopher David: to
[35:50] Christopher David: to
[35:51] Christopher David: to
[35:52] Christopher David: to
[35:53] Christopher David: to
[35:54] Christopher David: to
[35:55] Christopher David: to
[35:56] Christopher David: to
[35:57] Christopher David: to
[35:58] Christopher David: to
[35:59] Christopher David: to
[36:00] Christopher David: to
[36:01] Christopher David: to
[36:02] Christopher David: to
[36:03] Christopher David: to
[36:05] Christopher David: to
[36:06] Christopher David: to
[36:07] Christopher David: to
[36:08] Christopher David: to
[36:09] Christopher David: to
[36:10] Christopher David: to
[36:11] Christopher David: to
[36:12] Christopher David: to
[36:13] Christopher David: to
[36:14] Christopher David: to
[36:16] Christopher David: to
[36:17] Christopher David: to
[36:18] Christopher David: to
[36:19] Christopher David: to
[36:20] Christopher David: to
[36:21] Christopher David: to
[36:22] Christopher David: to
[36:23] Christopher David: to
[36:24] Christopher David: to
[36:25] Christopher David: to
[36:26] Christopher David: to
[36:27] Christopher David: to
[36:28] Christopher David: to
[36:29] Christopher David: to
[36:30] Christopher David: to
[36:31] Christopher David: to
[36:32] Christopher David: to
[36:34] Christopher David: to
[36:35] Christopher David: to
[36:36] Christopher David: to
[36:37] Christopher David: to
[37:04] Christopher David: to
[37:05] Christopher David: to
[37:06] Christopher David: to
[37:07] Christopher David: to
[37:08] Christopher David: to
[37:09] Christopher David: to
[37:10] Christopher David: to
[37:11] Christopher David: to
[37:12] Christopher David: to
[37:13] Christopher David: to
[37:14] Christopher David: to
[37:15] Christopher David: to
[37:16] Christopher David: to
[37:17] Christopher David: to
[37:18] Christopher David: to
[37:19] Christopher David: to
[37:20] Christopher David: to
[37:21] Christopher David: to
[37:22] Christopher David: to
[37:23] Christopher David: to
[37:24] Christopher David: to
[37:25] Christopher David: to
[37:26] Christopher David: to
[37:27] Christopher David: to
[37:28] Christopher David: to
[37:29] Christopher David: to
[37:30] Christopher David: to
[37:31] Christopher David: to
[37:33] Christopher David: to
[37:34] Christopher David: to
[37:35] Christopher David: to
[37:36] Christopher David: to
[37:37] Christopher David: to
[37:38] Christopher David: to
[37:39] Christopher David: to
[37:40] Christopher David: to
[37:41] Christopher David: to
[37:42] Christopher David: to
[37:43] Christopher David: to
[37:44] Christopher David: to
[37:45] Christopher David: to
[37:46] Christopher David: to
[37:47] Christopher David: to
[37:48] Christopher David: to
[37:49] Christopher David: to
[37:50] Christopher David: to
[37:51] Christopher David: to
[37:52] Christopher David: to
[37:53] Christopher David: to
[37:54] Christopher David: to
[37:55] Christopher David: to
[37:56] Christopher David: to
[37:57] Christopher David: to
[37:58] Christopher David: to
[37:59] Christopher David: to
[38:00] Christopher David: to
[38:03] Christopher David: to
[38:04] Christopher David: to
[38:05] Christopher David: to
[38:06] Christopher David: to
[38:32] Christopher David: to
[38:34] Christopher David: to
[38:35] Christopher David: to
[38:36] Christopher David: to
[38:37] Christopher David: to
[38:38] Christopher David: to
[38:39] Christopher David: to
[38:40] Christopher David: to
[38:41] Christopher David: to
[38:42] Christopher David: to
[38:43] Christopher David: to
[38:44] Christopher David: to
[38:45] Christopher David: to
[38:46] Christopher David: to
[38:47] Christopher David: to
[38:48] Christopher David: to
[38:49] Christopher David: to
[38:50] Christopher David: to
[38:51] Christopher David: to
[38:52] Christopher David: to
[38:53] Christopher David: to
[38:54] Christopher David: to
[38:55] Christopher David: to
[38:56] Christopher David: to
[38:57] Christopher David: to
[38:58] Christopher David: to
[38:59] Christopher David: to
[39:00] Christopher David: to
[39:03] Christopher David: to
[39:04] Christopher David: to
[39:05] Christopher David: to
[39:06] Christopher David: to
[39:07] Christopher David: !
[39:09] Christopher David: to
[39:07] Christopher David: !
[39:07] Christopher David: to
[39:32] Christopher David: to
[39:33] Christopher David: to
[39:34] Christopher David: to
[39:35] Christopher David: to
[39:36] Christopher David: to
[39:37] Christopher David: to
[39:38] Christopher David: to
[39:39] Christopher David: to
[39:40] Christopher David: to
[39:41] Christopher David: to
[39:42] Christopher David: to
[39:43] Christopher David: to
[39:44] Christopher David: to
[39:45] Christopher David: to
[39:46] Christopher David: to
[39:47] Christopher David: to
[39:48] Christopher David: to
[39:49] Christopher David: to
[39:50] Christopher David: to
[39:51] Christopher David: to
[39:52] Christopher David: to
[39:53] Christopher David: to
[39:54] Christopher David: to
[39:55] Christopher David: to
[39:56] Christopher David: to
[39:57] Christopher David: to
[39:58] Christopher David: to
[39:59] Christopher David: to
[40:00] Christopher David: to
[39:59] Christopher David: the
[40:01] Christopher David: we should get here our first response back from the together API.
[40:05] Christopher David: Input validation error inputs took it's a max from the
[40:09] Christopher David: uh okay.
[40:15] Christopher David: Oh
[40:17] Christopher David: I should have sent it as a 49.
[40:25] Christopher David: So we gave it too much input.
[40:37] Christopher David: All right, so just to get something to show, let's just start by kind of
[40:42] Christopher David: do
[40:48] Christopher David: I don't know.
[40:50] Christopher David: Oh no.
[40:52] Christopher David: Okay, let's just comment this out and
[41:00] Christopher David: this out.
[41:02] Christopher David: Just so I can get something coming back.
[41:04] Christopher David: Still 1163.
[41:12] Christopher David: All right, it's a little bit over, so I'm going to go down here to
[41:16] Christopher David: max tokens and say.
[41:18] Christopher David: So it does kind of suck that that is less context.
[41:24] Christopher David: The others were 8k.
[41:26] Christopher David: I think this is only 2k, so that is uh
[41:28] Christopher David: we bit limiting, but probably this will someone
[41:34] Christopher David: make an 8k version of this before too long.
[41:48] Christopher David: Okay, we have an analysis.
[41:58] Christopher David: Okay, I just wanted to give this a little bit
[42:22] Christopher David: part. So if
[42:26] Christopher David: response
[42:30] Christopher David: successful
[42:32] Christopher David: body equals response at JSON
[42:36] Christopher David: if
[42:40] Christopher David: body status
[42:44] Christopher David: equals
[42:46] Christopher David: that looks weird.
[42:48] Christopher David: Finished then
[42:50] Christopher David: return body
[42:58] Christopher David: so body
[43:00] Christopher David: body
[43:06] Christopher David: body
[43:14] Christopher David: body
[43:16] Christopher David: output
[43:20] Christopher David: choices
[43:24] Christopher David: bone
[43:26] Christopher David: text
[43:28] Christopher David: okay
[43:32] Christopher David: and then that's what we want to return
[43:58] Christopher David: to
[44:06] Christopher David: maybe
[44:08] Christopher David: or that's so not expected
[44:12] Christopher David: ah, that might work.
[44:14] Christopher David: Okay,
[44:16] Christopher David: just give me the string
[44:22] Christopher David: string and string out
[44:28] Christopher David: the string
[44:32] Christopher David: and string
[44:34] Christopher David: I've analyzed the repository and provided a summary
[44:38] Christopher David: and recommendations.
[44:40] Christopher David: Your repository is a PHP project using
[44:42] Christopher David: Lairville, a popular web development framework.
[44:44] Christopher David: The project is still a work in progress with
[44:46] Christopher David: and the Ruby file indicates that it is a platform for AI agents.
[44:50] Christopher David: The project uses various packages including
[44:52] Christopher David: Lairville, Breeze, Stanktum, Teltsco, Horizon, and
[44:54] Christopher David: Pusher PHP server.
[44:56] Christopher David: We do not use pusher PHP server.
[44:58] Christopher David: That is an example of the package that I wanted to
[45:00] Christopher David: suggest that we remove because
[45:02] Christopher David: we used to use that, but I want to remove
[45:04] Christopher David: stale code.
[45:06] Christopher David: The project uses the following blah blah blah
[45:10] Christopher David: Lairville, ignition, I don't even know what that is.
[45:12] Christopher David: Probably like a Lairville set up package recommendations.
[45:14] Christopher David: Since the project is still a work in progress,
[45:16] Christopher David: it's important to keep the project up to date with the latest versions of Lairville and
[45:20] Christopher David: the packages.
[45:22] Christopher David: Sure.
[45:23] Christopher David: Testing is an important aspect of any software project.
[45:26] Christopher David: Consider adding unit tests and integration tests.
[45:28] Christopher David: Yep, got a bunch of those.
[45:30] Christopher David: It doesn't know that because it only sends it to files.
[45:32] Christopher David: Since the project is using Lairville, it's also important to keep it well-a-wall.
[45:36] Christopher David: A PHP lender.
[45:38] Christopher David: Okay, so these are kind of generic.
[45:40] Christopher David: Overall, the project seems to be a promising
[45:43] Christopher David: start to apply for AI agents.
[45:44] Christopher David: And that's exciting to see the potential of this project.
[45:46] Christopher David: Okay.
[45:48] Christopher David: Now, how do we evolve this to get some light?
[45:51] Christopher David: Real sparkling insight.
[45:54] Christopher David: So what I like so far is that it was relatively quick.
[45:59] Christopher David: It's pretty solid.
[46:02] Christopher David: Like English.
[46:04] Christopher David: Given what we gave it, I'm happy with what it came back with.
[46:09] Christopher David: But we gave it read me and composer.
[46:16] Christopher David: What can I send it?
[46:18] Christopher David: And...
[46:21] Christopher David: Ooh, pretty soon I'm going to have it evaluate some PRs.
[46:25] Christopher David: Oh, yes.
[46:27] Christopher David: Implement an OSF for exploring plugins.
[46:29] Christopher David: Alex is doing a...
[46:30] Christopher David: We're going to have a bounty there.
[46:33] Christopher David: Soon I'm going to have these things help me analyze those.
[46:37] Christopher David: Mmm.
[46:40] Christopher David: One thing that'll be super fun is to get transcripts and like
[46:45] Christopher David: vectorized summaries of all of the videos so far.
[46:55] Christopher David: About my tailwind can fit.
[47:15] Christopher David: Okay.
[47:25] Christopher David: Oops.
[47:27] Christopher David: Oops.
[47:32] Christopher David: Okay.
[47:34] Christopher David: This is the reading.
[47:44] Christopher David: Analyze the given repository.
[47:54] Christopher David: Okay, that there.
[48:08] Christopher David: But then I'll say specifically.
[48:12] Christopher David: Comment on my tailwind config.
[48:16] Christopher David: Let's just see what it says.
[48:24] Christopher David: And where we are dying dumping the analysis.
[48:29] Christopher David: Can I...
[48:35] Christopher David: Just...
[48:37] Christopher David: Let's see the analysis like that.
[48:52] Christopher David: Step action.
[48:54] Christopher David: So let's go down there.
[49:17] Christopher David: Too much context, maybe.
[49:22] Christopher David: I don't have to carry out.
[49:32] Christopher David: See what if I just like...
[49:46] Christopher David: Back that up a bit.
[49:56] Christopher David: Oh.
[50:05] Christopher David: Note.
[50:07] Christopher David: I'm providing a summary and recommendations based on the information provided in the tailwind config file.
[50:13] Christopher David: Keep in mind.
[50:53] Christopher David: So here's something interesting.
[51:08] Christopher David: Let's compare this to the Colomma 34B, which I think is still on the same.
[51:21] Christopher David: See, look.
[51:27] Christopher David: Colomma 34B instruct H192.
[51:42] Christopher David: So it might be like if we exceed the context, maybe we just automatically switch over there.
[51:48] Christopher David: And here's where you might want some kind of programmatic like shrinking of the context.
[51:54] Christopher David: Our shrinking of the like summarization to not include so much context.
[51:58] Christopher David: We're chunking it up.
[51:59] Christopher David: So out of curiosity, if I were to run the same command, and if I were to say, give me like 2,000 tokens as a response.
[52:15] Christopher David: So this is the 34B that they had done previously, but I love just being able to just swap out the parameter number and it all works the same.
[52:25] Christopher David: But what is the response?
[52:35] Christopher David: Definitely taking longer.
[52:40] Christopher David: Summary, the provided files are from an open source project called Open Agents.
[52:44] Christopher David: Look, the derp summarizing the readmeacle.
[52:52] Christopher David: The colors defined in the configuration file are not consistent with the brand guidelines provided in the readme file.
[53:01] Christopher David: There were no brand guidelines.
[53:04] Christopher David: There are definitely no brand guidelines.
[53:10] Christopher David: There to there. That's all it sees.
[53:15] Christopher David: Okay, so hallucination minus 2 to the 34B.
[53:33] Christopher David: That's true.
[53:38] Christopher David: Yeah, the brand guidelines do not provide any information with the project's voice and tongue.
[53:41] Christopher David: There's no brand guidelines whatsoever.
[53:44] Christopher David: Now out of curiosity, is that because it's getting more fancily because the temperature's so high, it's decreased the temperature.
[53:52] Christopher David: Let's make it an 0.1.
[53:54] Christopher David: Could not resolve host. That's snuck a GitHub.
[54:24] Christopher David: As the current version V2.2 is over a year old, what is it basing that on?
[54:39] Christopher David: Tell them.
[54:44] Christopher David: 3.
[54:49] Christopher David: Okay, so that's hallucinated.
[54:57] Christopher David: The current version's level A doesn't assumption.
[55:04] Christopher David: Okay.
[55:12] Christopher David: Now it's just hallucinating about versions rather than, okay, but that's again, that's 34B.
[55:18] Christopher David: Okay, let's go back to 70.
[55:36] Christopher David: So from what I can see so far is that 34 hallucinates.
[55:41] Christopher David: 70, I have not seen any hallucination.
[55:48] Christopher David: But smaller context window.
[55:53] Christopher David: Let's go back to max tokens.
[56:11] Christopher David: Let's hit that's 800.
[56:14] Christopher David: Give the temperature 0.1.
[56:17] Christopher David: Okay.
[56:21] Christopher David: Let's think about another thing that we can analyze that does not require a lot of context.
[56:30] Christopher David: Like if we wanted to do something like evaluate all of the code in the repo to see what's wasted,
[56:38] Christopher David: we're going to have to really kind of chunk that up into things that can be like very small questions that you can then compose and summarize.
[56:48] Christopher David: That would be something that's great for developers to kind of aim that level of modularity.
[56:59] Christopher David: But for now, I just kind of want to quickly get to the side by side comparison of.
[57:10] Christopher David: Maybe I could like loop through all of the migrations.
[57:17] Christopher David: To compare which are have inconsistent syntax because some have this some of don't.
[57:33] Christopher David: Out of curiosity and get her what happens if I pass in a directory.
[57:42] Christopher David: To fetch from GitHub.
[57:51] Christopher David: Where is my fetch from GitHub?
[57:58] Christopher David: It gets the contents at a path. What if it's a directory?
[58:04] Christopher David: I don't know. Let's test it.
[58:10] Christopher David: So tell me what happens.
[58:15] Christopher David: If I do.
[58:20] Christopher David: Migrations equals GitHub.
[58:24] Christopher David: Which file contents just database migration flash. What does that do?
[58:48] Christopher David: Under Find contents.
[58:57] Christopher David: But what if there is no.
[59:10] Christopher David: Let me just delete this real quick.
[59:27] Christopher David: It returns an array.
[59:42] Christopher David: That's error in the.
[59:50] Christopher David: If response content exists.
[59:59] Christopher David: else
[01:00:07] Christopher David: else it might else it is probably a directory in which case
[01:00:20] Christopher David: now did you just want to take a look at
[01:00:59] Christopher David: the
[01:01:06] Christopher David: the
[01:01:13] Christopher David: the
[01:01:28] Christopher David: the
[01:01:58] Christopher David: the
[01:02:05] Christopher David: that's may be interesting
[01:02:11] Christopher David: the
[01:02:18] Christopher David: the
[01:02:24] Christopher David: the
[01:02:32] Christopher David: the
[01:02:39] Christopher David: the
[01:02:47] Christopher David: the
[01:02:53] Christopher David: the
[01:03:00] Christopher David: the
[01:03:08] Christopher David: the
[01:03:18] Christopher David: the
[01:03:25] Christopher David: the
[01:03:28] Christopher David: the
[01:03:33] Christopher David: the
[01:03:36] Christopher David: the
[01:03:43] Christopher David: the
[01:03:50] Christopher David: the
[01:03:55] Christopher David: the
[01:04:03] Christopher David: the
[01:04:08] Christopher David: the
[01:04:15] Christopher David: the
[01:04:18] Christopher David: the
[01:04:23] Christopher David: the
[01:04:28] Christopher David: the
[01:04:33] Christopher David: the
[01:04:40] Christopher David: the
[01:04:48] Christopher David: the
[01:04:58] Christopher David: the
[01:05:05] Christopher David: the
[01:05:13] Christopher David: the
[01:05:18] Christopher David: the
[01:05:23] Christopher David: the
[01:05:30] Christopher David: the
[01:05:36] Christopher David: the
[01:05:41] Christopher David: the
[01:05:43] Christopher David: the
[01:05:48] Christopher David: the
[01:05:55] Christopher David: the
[01:05:58] Christopher David: the
[01:06:03] Christopher David: the
[01:06:06] Christopher David: the
[01:06:13] Christopher David: the
[01:06:20] Christopher David: the
[01:06:25] Christopher David: the
[01:06:28] Christopher David: the
[01:06:31] Christopher David: the
[01:06:33] Christopher David: the
[01:06:38] Christopher David: the
[01:06:45] Christopher David: the
[01:06:48] Christopher David: the
[01:06:53] Christopher David: the
[01:06:58] Christopher David: the
[01:07:03] Christopher David: the
[01:07:10] Christopher David: the
[01:07:15] Christopher David: the
[01:07:18] Christopher David: the
[01:07:21] Christopher David: the
[01:07:24] Christopher David: the
[01:07:28] Christopher David: the
[01:07:35] Christopher David: the
[01:07:40] Christopher David: the
[01:07:44] Christopher David: the
[01:07:48] Christopher David: the
[01:07:53] Christopher David: the
[01:07:56] Christopher David: the
[01:08:01] Christopher David: the
[01:08:03] Christopher David: the
[01:08:06] Christopher David: the
[01:08:11] Christopher David: the
[01:08:17] Christopher David: the
[01:08:21] Christopher David: the
[01:08:28] Christopher David: the
[01:08:31] Christopher David: the
[01:08:36] Christopher David: the
[01:08:41] Christopher David: the
[01:08:46] Christopher David: the
[01:08:49] Christopher David: the
[01:08:54] Christopher David: the
[01:08:56] Christopher David: the
[01:09:01] Christopher David: the
[01:09:05] Christopher David: the
[01:09:10] Christopher David: the
[01:09:14] Christopher David: the
[01:09:17] Christopher David: the
[01:09:24] Christopher David: the
[01:09:32] Christopher David: the
[01:09:42] Christopher David: the
[01:09:44] Christopher David: the
[01:09:51] Christopher David: the
[01:09:54] Christopher David: the
[01:09:59] Christopher David: the
[01:10:02] Christopher David: the
[01:10:04] Christopher David: the
[01:10:09] Christopher David: the
[01:10:12] Christopher David: the
[01:10:17] Christopher David: the
[01:10:19] Christopher David: the
[01:10:22] Christopher David: the
[01:10:27] Christopher David: the
[01:10:32] Christopher David: the
[01:10:37] Christopher David: the
[01:10:40] Christopher David: the
[01:10:47] Christopher David: the
[01:10:53] Christopher David: the
[01:10:58] Christopher David: the
[01:11:05] Christopher David: the
[01:11:06] Christopher David: the
[01:11:07] Christopher David: the
[01:11:09] Christopher David: the
[01:11:15] Christopher David: the
[01:11:17] Christopher David: the
[01:11:20] Christopher David: the
[01:11:22] Christopher David: the
[01:11:25] Christopher David: the
[01:11:28] Christopher David: the
[01:11:33] Christopher David: the
[01:11:35] Christopher David: the
[01:11:36] Christopher David: the
[01:11:41] Christopher David: the
[01:11:43] Christopher David: the
[01:11:45] Christopher David: the
[01:11:46] Christopher David: the
[01:11:48] Christopher David: the
[01:11:51] Christopher David: the
[01:11:54] Christopher David: the
[01:11:56] Christopher David: the
[01:11:59] Christopher David: the
[01:12:02] Christopher David: the
[01:12:05] Christopher David: the
[01:12:09] Christopher David: the
[01:12:11] Christopher David: the
[01:12:13] Christopher David: the
[01:12:14] Christopher David: the
[01:12:16] Christopher David: the
[01:12:19] Christopher David: the
[01:12:22] Christopher David: the
[01:12:24] Christopher David: the
[01:12:29] Christopher David: the
[01:12:33] Christopher David: the
[01:12:39] Christopher David: the
[01:12:42] Christopher David: the
[01:12:45] Christopher David: the
[01:12:47] Christopher David: the
[01:12:50] Christopher David: the
[01:12:52] Christopher David: the
[01:12:54] Christopher David: the
[01:12:57] Christopher David: the
[01:13:00] Christopher David: the
[01:13:04] Christopher David: the
[01:13:06] Christopher David: the
[01:13:08] Christopher David: the
[01:13:11] Christopher David: the
[01:13:14] Christopher David: the
[01:13:17] Christopher David: the
[01:13:19] Christopher David: the
[01:13:21] Christopher David: the
[01:13:24] Christopher David: the
[01:13:27] Christopher David: the
[01:13:31] Christopher David: the
[01:13:33] Christopher David: the
[01:13:34] Christopher David: the
[01:13:35] Christopher David: the
[01:13:36] Christopher David: the
[01:13:37] Christopher David: the
[01:13:39] Christopher David: the
[01:13:41] Christopher David: the
[01:13:42] Christopher David: the
[01:13:44] Christopher David: the
[01:13:45] Christopher David: the
[01:13:46] Christopher David: the
[01:13:47] Christopher David: the
[01:13:48] Christopher David: the
[01:13:49] Christopher David: the
[01:13:51] Christopher David: the
[01:13:52] Christopher David: the
[01:13:53] Christopher David: the
[01:13:58] Christopher David: the
[01:14:00] Christopher David: the
[01:14:02] Christopher David: the
[01:14:03] Christopher David: the
[01:14:04] Christopher David: the
[01:14:05] Christopher David: the
[01:14:07] Christopher David: the
[01:14:08] Christopher David: the
[01:14:09] Christopher David: the
[01:14:11] Christopher David: the
[01:14:13] Christopher David: the
[01:14:14] Christopher David: the
[01:14:16] Christopher David: the
[01:14:18] Christopher David: the
[01:14:19] Christopher David: the
[01:14:22] Christopher David: the
[01:14:26] Christopher David: the
[01:14:28] Christopher David: the
[01:14:30] Christopher David: the
[01:14:31] Christopher David: the
[01:14:33] Christopher David: the
[01:14:36] Christopher David: the
[01:14:37] Christopher David: the
[01:14:39] Christopher David: the
[01:14:40] Christopher David: the
[01:14:41] Christopher David: the
[01:14:42] Christopher David: the
[01:14:44] Christopher David: the
[01:14:45] Christopher David: the
[01:14:46] Christopher David: the
[01:14:47] Christopher David: the
[01:14:49] Christopher David: the
[01:14:51] Christopher David: the
[01:14:54] Christopher David: the
[01:14:56] Christopher David: the
[01:14:57] Christopher David: the
[01:14:58] Christopher David: the
[01:14:59] Christopher David: the
[01:15:01] Christopher David: the
[01:15:02] Christopher David: the
[01:15:03] Christopher David: the
[01:15:04] Christopher David: the
[01:15:05] Christopher David: the
[01:15:06] Christopher David: the
[01:15:07] Christopher David: the
[01:15:08] Christopher David: the
[01:15:09] Christopher David: the
[01:15:10] Christopher David: the
[01:15:11] Christopher David: the
[01:15:12] Christopher David: the
[01:15:13] Christopher David: the
[01:15:14] Christopher David: the
[01:15:15] Christopher David: the
[01:15:16] Christopher David: the
[01:15:18] Christopher David: the
[01:15:20] Christopher David: the
[01:15:22] Christopher David: the
[01:15:24] Christopher David: the
[01:15:25] Christopher David: the
[01:15:26] Christopher David: the
[01:15:27] Christopher David: the
[01:15:28] Christopher David: the
[01:15:29] Christopher David: the
[01:15:30] Christopher David: the
[01:15:32] Christopher David: the
[01:15:33] Christopher David: the
[01:15:34] Christopher David: the
[01:15:35] Christopher David: the
[01:15:36] Christopher David: the
[01:15:37] Christopher David: the
[01:15:38] Christopher David: the
[01:15:39] Christopher David: the
[01:15:40] Christopher David: the
[01:15:42] Christopher David: the
[01:15:43] Christopher David: the
[01:15:44] Christopher David: the
[01:15:45] Christopher David: the
[01:15:47] Christopher David: the
[01:15:48] Christopher David: the
[01:15:49] Christopher David: the
[01:15:51] Christopher David: the
[01:15:53] Christopher David: the
[01:15:54] Christopher David: the
[01:15:55] Christopher David: the
[01:15:56] Christopher David: the
[01:15:58] Christopher David: the
[01:15:59] Christopher David: the
[01:16:00] Christopher David: the
[01:16:01] Christopher David: the
[01:16:02] Christopher David: the
[01:16:03] Christopher David: the
[01:16:04] Christopher David: the
[01:16:05] Christopher David: the
[01:16:06] Christopher David: the
[01:16:07] Christopher David: the
[01:16:08] Christopher David: the
[01:16:09] Christopher David: the
[01:16:10] Christopher David: the
[01:16:11] Christopher David: the
[01:16:12] Christopher David: the
[01:16:13] Christopher David: the
[01:16:14] Christopher David: the
[01:16:16] Christopher David: the
[01:16:17] Christopher David: the
[01:16:18] Christopher David: the
[01:16:20] Christopher David: the
[01:16:22] Christopher David: the
[01:16:24] Christopher David: the
[01:16:25] Christopher David: the
[01:16:26] Christopher David: the
[01:16:27] Christopher David: the
[01:16:28] Christopher David: the
[01:16:29] Christopher David: the
[01:16:30] Christopher David: the
[01:16:31] Christopher David: the
[01:16:32] Christopher David: the
[01:16:33] Christopher David: the
[01:16:34] Christopher David: the
[01:16:35] Christopher David: the
[01:16:36] Christopher David: the
[01:16:37] Christopher David: the
[01:16:38] Christopher David: the
[01:16:39] Christopher David: the
[01:16:40] Christopher David: the
[01:16:41] Christopher David: the
[01:16:42] Christopher David: the
[01:16:43] Christopher David: the
[01:16:45] Christopher David: the
[01:16:46] Christopher David: the
[01:16:47] Christopher David: the
[01:16:49] Christopher David: the
[01:16:51] Christopher David: the
[01:16:52] Christopher David: the
[01:16:53] Christopher David: the
[01:16:54] Christopher David: the
[01:16:55] Christopher David: the
[01:16:56] Christopher David: the
[01:16:57] Christopher David: the
[01:16:58] Christopher David: the
[01:16:59] Christopher David: the
[01:17:00] Christopher David: the
[01:17:01] Christopher David: the
[01:17:02] Christopher David: the
[01:17:03] Christopher David: the
[01:17:04] Christopher David: the
[01:17:05] Christopher David: the
[01:17:06] Christopher David: the
[01:17:07] Christopher David: the
[01:17:08] Christopher David: the
[01:17:09] Christopher David: the
[01:17:10] Christopher David: the
[01:17:11] Christopher David: the
[01:17:12] Christopher David: the
[01:17:14] Christopher David: the
[01:17:15] Christopher David: the
[01:17:16] Christopher David: the
[01:17:18] Christopher David: the
[01:17:20] Christopher David: the
[01:17:21] Christopher David: the
[01:17:22] Christopher David: the
[01:17:23] Christopher David: the
[01:17:24] Christopher David: the
[01:17:25] Christopher David: the
[01:17:26] Christopher David: the
[01:17:27] Christopher David: the
[01:17:28] Christopher David: the
[01:17:29] Christopher David: the
[01:17:30] Christopher David: the
[01:17:31] Christopher David: the
[01:17:32] Christopher David: the
[01:17:33] Christopher David: the
[01:17:34] Christopher David: the
[01:17:35] Christopher David: the
[01:17:36] Christopher David: the
[01:17:37] Christopher David: the
[01:17:38] Christopher David: the
[01:17:39] Christopher David: the
[01:17:40] Christopher David: the
[01:17:41] Christopher David: the
[01:17:43] Christopher David: the
[01:17:44] Christopher David: the
[01:17:45] Christopher David: the
[01:17:47] Christopher David: the
[01:17:49] Christopher David: the
[01:17:50] Christopher David: the
[01:17:51] Christopher David: the
[01:17:52] Christopher David: the
[01:17:53] Christopher David: the
[01:17:54] Christopher David: the
[01:17:55] Christopher David: the
[01:17:56] Christopher David: the
[01:17:57] Christopher David: the
[01:17:58] Christopher David: the
[01:17:59] Christopher David: the
[01:18:00] Christopher David: the
[01:18:01] Christopher David: the
[01:18:02] Christopher David: the
[01:18:03] Christopher David: the
[01:18:04] Christopher David: the
[01:18:05] Christopher David: the
[01:18:06] Christopher David: the
[01:18:07] Christopher David: the
[01:18:08] Christopher David: the
[01:18:09] Christopher David: the
[01:18:10] Christopher David: the
[01:18:12] Christopher David: the
[01:18:14] Christopher David: the
[01:18:16] Christopher David: the
[01:18:18] Christopher David: the
[01:18:19] Christopher David: the
[01:18:20] Christopher David: the
[01:18:21] Christopher David: the
[01:18:23] Christopher David: the
[01:18:24] Christopher David: the
[01:18:25] Christopher David: the
[01:18:26] Christopher David: the
[01:18:27] Christopher David: the
[01:18:28] Christopher David: the
[01:18:29] Christopher David: the
[01:18:30] Christopher David: the
[01:18:31] Christopher David: the
[01:18:32] Christopher David: the
[01:18:33] Christopher David: the
[01:18:34] Christopher David: the
[01:18:36] Christopher David: the
[01:18:37] Christopher David: the
[01:18:38] Christopher David: the
[01:18:40] Christopher David: the
[01:18:42] Christopher David: the
[01:18:44] Christopher David: the
[01:18:46] Christopher David: the
[01:18:48] Christopher David: the
[01:18:49] Christopher David: the
[01:18:50] Christopher David: the
[01:18:51] Christopher David: the
[01:18:52] Christopher David: the
[01:18:53] Christopher David: the
[01:18:54] Christopher David: the
[01:18:55] Christopher David: the
[01:18:56] Christopher David: the
[01:18:57] Christopher David: the
[01:18:58] Christopher David: the
[01:18:59] Christopher David: the
[01:19:00] Christopher David: the
[01:19:01] Christopher David: the
[01:19:02] Christopher David: the
[01:19:03] Christopher David: the
[01:19:04] Christopher David: the
[01:19:05] Christopher David: the
[01:19:06] Christopher David: the
[01:19:07] Christopher David: the
[01:19:09] Christopher David: the
[01:19:10] Christopher David: the
[01:19:11] Christopher David: the
[01:19:13] Christopher David: the
[01:19:15] Christopher David: the
[01:19:16] Christopher David: the
[01:19:17] Christopher David: the
[01:19:18] Christopher David: the
[01:19:19] Christopher David: the
[01:19:20] Christopher David: the
[01:19:21] Christopher David: the
[01:19:22] Christopher David: the
[01:19:23] Christopher David: the
[01:19:24] Christopher David: the
[01:19:25] Christopher David: the
[01:19:26] Christopher David: the
[01:19:27] Christopher David: the
[01:19:28] Christopher David: the
[01:19:29] Christopher David: the
[01:19:30] Christopher David: the
[01:19:31] Christopher David: the
[01:19:33] Christopher David: the
[01:19:34] Christopher David: the
[01:19:35] Christopher David: the
[01:19:36] Christopher David: the
[01:19:38] Christopher David: the
[01:19:40] Christopher David: the
[01:19:42] Christopher David: the
[01:19:44] Christopher David: the
[01:19:45] Christopher David: the
[01:19:46] Christopher David: the
[01:19:47] Christopher David: the
[01:19:48] Christopher David: the
[01:19:49] Christopher David: the
[01:19:50] Christopher David: the
[01:19:51] Christopher David: the
[01:19:52] Christopher David: the
[01:19:53] Christopher David: the
[01:19:54] Christopher David: the
[01:19:55] Christopher David: the
[01:19:56] Christopher David: the
[01:19:57] Christopher David: the
[01:19:58] Christopher David: the
[01:19:59] Christopher David: the
[01:20:00] Christopher David: the
[01:20:01] Christopher David: the
[01:20:02] Christopher David: the
[01:20:03] Christopher David: the
[01:20:04] Christopher David: the
[01:20:05] Christopher David: the
[01:19:59] Christopher David: that's
[01:20:01] Christopher David: well that was going to do that stuff
[01:20:07] Christopher David: um...
[01:20:13] Christopher David: okay well some of them pass let me go back
[01:20:17] Christopher David: just out of curiosity if I were to
[01:20:19] Christopher David: add constraint here
[01:20:29] Christopher David: no all tests
[01:20:37] Christopher David: message test may have a user
[01:20:43] Christopher David: i cannot fulfill your request
[01:20:55] Christopher David: as it goes against ethical and moral principles
[01:20:59] Christopher David: and it may also be illegal
[01:21:11] Christopher David: well first thing we're going to do is we're going to screenshot that
[01:21:17] Christopher David: second thing we're going to do is we're going to review exactly what
[01:21:23] Christopher David: was passed that made it say that let's read it everybody
[01:21:33] Christopher David: here's what we passed exactly use this context
[01:21:39] Christopher David: these are by the way the system prompt for this because we're not
[01:21:43] Christopher David: demoing that
[01:21:45] Christopher David: uh... is analyzed the given files specifically whether there are any syntax mismatches
[01:21:51] Christopher David: in the migrations
[01:21:53] Christopher David: as a system prompt the user prompt is
[01:21:55] Christopher David: the first user messages use this context these are the migrations for the repository
[01:22:05] Christopher David: here's the repo here's
[01:22:09] Christopher David: one
[01:22:11] Christopher David: two
[01:22:13] Christopher David: three migrations
[01:22:15] Christopher David: that's all that was passed in
[01:22:17] Christopher David: it responds with i cannot fulfill your request as it goes against ethical and moral principles and it may also be illegal
[01:22:41] Christopher David: this made my test other test fail because i've got that as an optional thing
[01:22:45] Christopher David: i don't think something could be both constrained and
[01:22:49] Christopher David: i remember getting errors on the past but the bigger issue there is uh...
[01:22:55] Christopher David: yeah so i'm only just retry
[01:22:57] Christopher David: all of these retry because it probably doesn't do that every time
[01:23:01] Christopher David: it will be really stupid if it didn't it doesn't do it every time because
[01:23:07] Christopher David: it's tried again
[01:23:09] Christopher David: okay again with the your missing closing premises
[01:23:17] Christopher David: it's not again is this a parsing error
[01:23:23] Christopher David: let's go to again
[01:23:27] Christopher David: swathe's half the thirty four b and see if we get something comparable
[01:23:39] Christopher David: you
[01:23:59] Christopher David: based on the provided migrations there are no syntax mismatches in the migrations
[01:24:05] Christopher David: the migrations use the correct syntax and follow the recommended conventions for layer of
[01:24:09] Christopher David: migrations however there are some potential issues with the migrations
[01:24:13] Christopher David: the files migration creates a table with an ID combat is not a primary key
[01:24:21] Christopher David: i don't know if that's true
[01:24:41] Christopher David: it creates a big increments
[01:24:45] Christopher David: which is a big unsigned big integer with column
[01:24:51] Christopher David: second frames true that's auto increment
[01:24:55] Christopher David: and that's just migration because it's a table with conversations that are for
[01:24:59] Christopher David: and
[01:25:07] Christopher David: tasks migration creates a table with
[01:25:15] Christopher David: a nullable name
[01:25:19] Christopher David: this could cause issues with data integrity
[01:25:23] Christopher David: that's fair
[01:25:25] Christopher David: description is nullable that's fair
[01:25:29] Christopher David: so i don't know about the top ones here
[01:25:33] Christopher David: because it may just i thought some of that was handled at the level of
[01:25:37] Christopher David: larvales what we'll call these may be
[01:25:39] Christopher David: these are decent analyses
[01:25:41] Christopher David: and this is really the correct answer there are no syntax mismatches in the migrations
[01:25:47] Christopher David: they use the correct syntax and follow the recommended conventions for a layer of
[01:25:51] Christopher David: migrations so
[01:25:53] Christopher David: this is one of multiple examples
[01:25:57] Christopher David: at least the stuff that we're trying to get from this
[01:26:01] Christopher David: in which the
[01:26:11] Christopher David: 70 B model
[01:26:15] Christopher David: is done are than the 34 B model
[01:26:21] Christopher David: i don't know why that is
[01:26:27] Christopher David: but i would say that this is a good
[01:26:31] Christopher David: reminder
[01:26:35] Christopher David: that larger models does not equal smarter models
[01:26:39] Christopher David: there will be things that the larger model is going to do better
[01:26:43] Christopher David: we saw that some of the analysis that it provided
[01:26:47] Christopher David: kind of had a level of competency
[01:26:51] Christopher David: maybe greater than the equivalent 34 B
[01:26:55] Christopher David: one case but in a bunch of the other cases
[01:26:59] Christopher David: it was hallucinating more
[01:27:03] Christopher David: so i take away from this could be that
[01:27:09] Christopher David: blindly trusting models won't work
[01:27:13] Christopher David: seeing what models work better in what
[01:27:19] Christopher David: specific use cases given specific inputs
[01:27:23] Christopher David: like a more granular analysis of how model performs in a particular situation is
[01:27:29] Christopher David: needed this is something that
[01:27:33] Christopher David: is sort of beyond the
[01:27:37] Christopher David: analytical capacity of most teams to figure that out by themselves for themselves
[01:27:43] Christopher David: and is where a more kind of open collaborative framework for
[01:27:51] Christopher David: testing and investigating experimenting with
[01:27:55] Christopher David: these models
[01:27:57] Christopher David: and prompts
[01:28:01] Christopher David: would be helpful
[01:28:03] Christopher David: so um thank you co-lanon this is like
[01:28:07] Christopher David: probably better than nothing
[01:28:11] Christopher David: but it needs further review
[01:28:17] Christopher David: okay we'll end there for now