The Great Loops Debate — Dex Horthy, Geoff Huntley, Ian Livingstone, Greg Pstrucha, @insecure-agents

AI Engineer · 60 min · 760 sentences · from YouTube's caption track

Each timecode opens YouTube at the start of that sentence. Line anchors (#s42) are the cue ids in the WebVTT, and every line carries its start and end seconds. All transcripts has every talk, and the whole corpus as one file.

00:12Hi, welcome to the great loops debate.
00:14My name is Ali Howard, the host of the insecure agents podcast.
00:18I also am a member of technical staff at Key Card and I'm super excited to bring the loops debate to you here today.
00:24You might recognize some familiar faces on stage today that did the MCP debate with us at AIE code back in November.
00:30Um, that was so much fun.
00:31We thought maybe we would do it a second time.
00:34Um, but this time we would debate loops and instead of just Ian versus Jax, we uh, we recruited some more people for each of them to have on their team.
00:42Um, so here on stage with me today to debate loops, we have Ian Livingston, CEO and co-founder of Key Card.
00:48We've got Jeffrey Huntley, the creator of the Ralph loop.
00:51Um, Greg Kostruba, developer at Century.
00:54And we've got Dax Raad, who you all know, um, CEO of Human Layer.
00:58Um, so super excited to have all of these people here with us today that are very close to loops engineering and also um, software factories to debate this topic.
01:08Um, so you might be wondering what are we actually debating today?
01:11Like aren't loops in and engineering um, in software factories pretty well uh, promoted and understood and we all kind of agree that's like where the industry is headed.
01:20Like what is there to debate?
01:21Um, it's a great question.
01:22What we're here to debate here today, the core thesis is there is or is not a delta between the hype behind loops and what actually works in practice.
01:31Um, and we also can debate, you know, is now really truly the largest inflection point we've seen towards fully autonomous software factories.
01:39Um, so super excited to cover in the debate today uh, loop history.
01:43We'll debate that.
01:44We'll debate the loop anatomy, like what makes a good loop.
01:47We'll also debate the future of the loop, um, future of loop engineering and what we need to see from that in order to make software factories truly something that every organization is able to build.
01:57And today's um, is going to be a Oxford debate format.
02:02So we're adding a little more structure to the debate this time this round where every single person will have a timed window to give their response.
02:10And so we won't have people running over.
02:12And the way this works in practice to judge the winner, which we'll judge in real time, is the winner will be the team that changes the most minds.
02:21So I want everyone in the audience right now to take a minute to think about whose side you're on.
02:27Lock that in, remember that.
02:29And at the end I will ask you to raise your hand and say, "Okay, actually I was changed my mind.
02:34Actually I'm now team Dexter.
02:35I changed my mind.
02:36I'm actually now team Ian."
02:38So to present the two sides to help you make that decisions, I'll talk about the first team, which is Ian and Jeff's team.
02:45Their team no Delta.
02:47Loops are absolutely worth the the hype today.
02:50People can get up and running with them easily, build them.
02:53They're an important step up the autonomy curve and towards real software factories.
02:57Key points to look out for for team Ian and Jeff is that loops are core unit of engineering.
03:04With the right discipline, infra, and tests, loops are highly effective.
03:08And the best practices for those have emerged.
03:12For team Dex and Greg, their side believes there is a delta between the hype behind loops and what actually works in practice.
03:19The way we are doing loops today is wrong.
03:21Loops are not a silver bullet and there is no magic.
03:25Key points to look out for.
03:26The hype is out running the discipline and a software factory can run the mechanical spec'd out test covered slices unattended.
03:32It cannot on and honestly decide whether it built the right thing.
03:36So you still need engineers in the loop essentially.
03:40So now that you know a little bit about what what both sides are about, take a minute to internalize that.
03:46Think about, okay, like where do I stand?
03:48Am I team Dex or am I team Ian?
03:50And then end we'll ask you again.
03:53So, from here we'll go into the beginning of our debate.
03:56Um each member of each team is going to give a 4-minute monologue on why they're defending their side.
04:03Um and just start us off, we'll have Jeff.
04:06Jeff, I'll give you 4 minutes on the floor.
04:07Why are you Why are you pro loops today?
04:10It's because it's somewhat inevitable.
04:12Um I I have first uh basically if we wind back time 2 and 1/2 years ago when I was a tech lead over at Canva and I was seeing all the engineers just prompting and prompting and prompting and they they they they were in the loop.
04:25And I'm like, "Wait a sec.
04:26This could be programmed.
04:28This is a programmable thing."
04:30And it just became a really inevitable.
04:31Um loops are uh somewhat uh whilst Ralph might be a bit of a meme and cetera, there was actually deep thought into it.
04:40Um it's essentially applying this as if this is a new form of CPU architecture and figuring out the behaviors of this and how to do it.
04:48And through that I was able to reduce it down to a bash loop.
04:52Now, it is not a complete silver bullet, folks.
04:56I have my deepest concerns next this time next year at the conference.
05:00We're going to see a whole bunch of talks saying how factories fail and how loops fail.
05:05Um these are things These are things that we are still yet to figure out.
05:13Do you remember the early days of Kubernetes where everyone was just doing Kubernetes?
05:17So, it's here.
05:19It's inevitable.
05:20It it It is here to stay.
05:22Like programming the machine and automating your job function is the expectation of employers.
05:29Categorically.
05:31And um I don't see myself going back to writing code by hand.
05:36It's been 2 and 1/2 years since I manually wrote code by hand.
05:40Um I autonomously factor code from one code base to another code base.
05:45[snorts]
05:45I find something in GoLang and I'm like, "Oh, I'm in TypeScript."
05:48So, I I I run a loop, and I just autonomously ported across.
05:52And like, even for product managers and product manager research, it's easy for us to index on what it means as software engineers.
05:59We can focus just on software engineering, but think about something like, you want to do product management research on all the linear tickets.
06:07Well, there is a termination.
06:09It's actually defined when you've enumerated all of those linear tickets.
06:14So, that's easy.
06:15So, there's a lot of nuance in here.
06:17But like, one If you ever done any product management research, and you started running these loops, and be able to like compress time and the amount of time to do that research, it it's inevitable.
06:28We got this new programmable substrate.
06:30We got to figure out how to use it, where it's good to use, and um I know I'm meant to debate like that it is the thing in all thing, but in all things, there is no silver bullet.
06:43And we're going to figure out how we're going to be using it effectively over the next year.
06:51Thank you.
06:52All right, next we'll have Dex.
06:55Tell us why there's a difference between the hype that exists between Loops today and Loops themselves.
07:00Okay, cool.
07:01And um what are the What are the rules around personal attack in this debate?
07:04Is this encouraged
07:05[laughter]
07:05or
07:06You can do whatever you want with your 4-minute monologue.
07:09That's That's the rules,
07:09[laughter]
07:09yeah.
07:10Um It's funny, this is reminding me of the debate last year, where we're we're all kind of arguing, and uh maybe at the end we all I'm going to be convinced that Jeff is right, but Jeff is going to be convinced that I'm right, and maybe we switch sides by the end.
07:25Um
07:26[clears throat]
07:27Yeah, I think uh the the the basic take here is is not whether Loops are good or bad.
07:34I think um it's funny you bring up Kubernetes, because Kubernetes was this thing that took us seven or eight years to get right.
07:40Yep.
07:40Uh and before that it was cloud infrastructure, and you could argue that like it took us seven or eight years of cloud infrastructure to get to Kubernetes to then seven or eight years to get that where it was really usable by everybody.
07:51Uh, and Kubernetes is actually built on loops.
07:53It's built on control loops, but they're deterministic loops.
07:57And we've actually figured out exactly what types of things that, uh, small isolated tasks that can be sort of owned by one system.
08:07I think this is actually the biggest value in loops is that you can pick a small, sort of desired end state and feed in the current state of the world
08:18and have an agent or a deterministic system kind of progress towards that desired end state.
08:24Um, the challenge I have with the hype is we were already in a world where it where the the the prevailing mantra was see if you can get to a point where you don't have to read the code anymore.
08:38And, uh, even before loops, it was like just prompt, just go.
08:42And this idea of I don't even prompt anymore, I'm even at a level higher up implies that I am like taking even more of a backseat to the architecture of my code.
08:53And I think my biggest like point in terms of like the hype is outrunning the discipline is that we are all looking for magic, you're all looking for a silver bullet, we're all looking for something
09:05that will take away that horrible part of our jobs that we all hate, which is like reviewing code.
09:11Uh, some people enjoy it.
09:12Really good pull requests are fun.
09:14But, that we can somehow prompt our way out of, um, this this challenge that models have of like, okay, the code is pretty If you've ever reviewed a fully,
09:26uh, lights off sort of, uh, no one read the code before they sent it to you PR, I'm sure you've had this, uh, probably not great experience.
09:34And I think I've seen lots of people try to apply AI to this problem of hey, we have review bots and we have all these things.
09:42But it it doesn't seem it doesn't feel to me like it's working.
09:45I haven't seen proof uh in any any of the discourse publicly or any of sort of more private conversations we've had with people trying to put this into practice
09:55that we are we are at a point where we can just kind of like step up an abstraction level.
10:00I actually think we need to step down an abstraction level if anything.
10:04Um so, I think loops are there are good things about loops and we should be doing them, but the hype is is making us feel like there's a magic answer to this
10:15and it requires a lot more thought and care than the uh the the Twitter sphere would have you believe.
10:22Excellent.
10:23Yeah, good points.
10:24All right, I'll kick it back over to you Ian for the pro loop side.
10:28Absolutely.
10:29So, I think first and foremost, I'm coming for you Dex.
10:31But in reality, um let let's let's take a step back.
10:35And let's talk about like what is software engineering in the first place and also just like remove the word engineering and talk about like development.
10:42In inherent in building a system, whether it was 50 years ago or it was a thousand years ago or it's today, it is a is a loop is at the core of I try something, I learn something, I apply something.
10:54And all we're really talking about is how quickly we can expedite that process, right?
10:59And really what we're doing is removing what used to be human judgment in that process.
11:04So, the speed at which it is to generate something and removing that from humans typing.
11:08And you know, tab completion is a version of of auto complete and what is this stuff other than like really like a much better version of tab complete.
11:16Except now we give a much higher level version of what intent instead of just typing tab, right?
11:21And so, I think the premise is loops are at the core of everything we build already.
11:25They were at the core of how we built software 30 years ago.
11:28What is CICD, PR pull requests, design review, feedback from customers other than just driving a loop.
11:34And the question is, how much of that process can of the which is deeply subjective and requires reasoning, can we move out of the human brain and into the brain into these like non-deterministic models?
11:45And the underlying question for all of us is more about verifiability.
11:49Software is one of the most unverifiable things in the world because ultimately at the end of the day, it is most things can become true or false one way or the other.
11:58And my premise and my my point I want to make here truly is as humans interact less with software, which is how does a human interpret what that software is doing,
12:07how does a human interact with that software, how does a human use judgment to navigate that software, the subjectivity of what that software needs to be through the human-computer interface reduces and becomes a much more verifiable problem because it becomes more constricted to specific APIs.
12:22And so over time, it's both the fact that at the core of software development is loop-driven anyways, what is lint than a feedback loop, and that creates a verifiable thing, and as more humans are less interacting with software, you have less UI and UX and less subjectivity and what and how we interpret those things,
12:38you will become much more loop-driven and you'll become much more verifiable in a way that wasn't previously not possible.
12:45Awesome.
12:46Yeah, good points for sure.
12:47All right, Greg, you want to close this out with the anti-loops or the there's a delta between the type.
12:53Absolutely.
12:53I do think that the way that I would start this is um there is a lot of hype.
12:59There is a lot of FOMO.
13:00There is a lot of I'm looking at Twitter.
13:02I'm seeing what people are talking about.
13:04Am I missing out?
13:04Am I doing something wrong?
13:05Am I holding AI wrong?
13:08Uh should I be catching up?
13:09And that is um very stressful.
13:14And I think it boils down to two points for me.
13:15One of them is when you are generating code with AI in any manner, loops or no loops, are you happy with the output?
13:23Do you think the output qualitatively is what you need it to be to do whatever you are trying to do to get to the desired state.
13:31Um and if so, I would love to learn from you.
13:34I have not get there.
13:35Uh I think that the best way that we are improving that is both with model intelligence, but also as everybody here seems to agree, semantic verification as much as we can do statically, we should do statically.
13:48That's that's sort of one big thing where I lean on and in practice end up reading uh code that is sort of post semantic verification and it's still crap.
14:00I still have to do a lot of iteration on that uh on my own and still have to steer it towards the right architecture, tell it where it should be simplified.
14:09Um And so that's one big step.
14:11And there is a lot of things you can help yourself with by throwing more tokens at the problem, but one of the things that the current sort of hype-based
14:19discourse uh leads you to believe is that you can just have loops on top of loops on top of loops and orchestrate that or orchestrate your problems of quality away by more tokens.
14:29And that brings me to my second point, which is the economic viability of of the way that we are using agents today.
14:37And I don't believe that this is sustainable.
14:41I don't believe that like you when you are at the company, especially a larger company, you have to ask yourself what is a good budget for an engineer?
14:48Is it 10K a month?
14:50100K?
14:511 million dollar a month for for a token spend?
14:53At some point that that just starts cracking and it's not sustainable in the way that we are doing it today.
14:59That said, I am also writing code with agents and I also use some loops um for some specific flows.
15:06It just depends.
15:07There is nuance.
15:08You know, if you go to Twitter, Twitter has no nuance.
15:11Um but there is actual nuance to the conversation and there are specific tasks and and jobs that uh you can already loop on and be and be getting pretty reasonable results.
15:22Thank you.
15:23All right, Now that we've heard from each of the debaters on stage more about their stance and their opinion, we'll move into the main debate portion.
15:32The first section of our debate is going to focus on the history of the loop and why now is a major inflection point or not for loops engineering and also software factories.
15:42I know people will have said that you know, vision models have improved a lot and so they'll be able to verify work the agents have done was not possible previously.
15:52Context windows have improved, therefore memory is improved and we can now track work in a loop that maybe we couldn't have before.
15:59So now we look at this like loops history, where did the loop start from?
16:02Was it you know, Jeff Huntley's like Ralph loop?
16:05We'll get into some of those questions and debate here.
16:09It'll be the questions we targeted now at a very specific like single person and only they will get to respond and they'll have 2 minutes and 30 seconds to respond.
16:18So our first question is Anthropic took Jeff's concept of the Ralph loop, absorbed it into their platform and created a series of three commands, loop, batch and goal.
16:28The goal command is designed to keep going until a condition is true.
16:32Agents are very determined and the whole point of this command is to keep going and iterating through ways to solve a task until it's done.
16:40Ian is a security expert in the room.
16:42How are you confident agents can stay aligned to their task and not overstep their intended permissions while ruthlessly pursuing their goals?
16:50I mean I think if there's any evidence, I'm totally not convinced that that's possible.
16:54In fact, I think what we've seen is as we scale with these models and as we use reinforcement learning, they're inherently incredibly goal-seeking.
17:03And so we're now we're seeing them finding exploits and vulnerabilities and escapes that you know, humans through hundreds and thousands and thousands and thousands of hours and attempts and attacks have never been able to find.
17:14So I I don't think inherently the model itself in any capacity can keep itself aligned Um to aligned and safe, right?
17:22And safe is a word that I don't love to use because it implies a bunch of things.
17:27So I'm I'm not I'm not I don't have good belief that the model itself can actually do that.
17:32I don't think it can reason.
17:33I also don't believe holistically that a model can tell good from bad and I can't tell whether it's doing something malicious or unaligned.
17:42It is not alive and it doesn't actually self if it doesn't have air.
17:46It doesn't deal with the fact that hey, if I do something wrong, no one's going to love me or or want to be my friend.
17:52And if I do something good that someone's going to praise me, it may seem that way, but it is just a probability distribution at scale.
17:57So I'll still I'll still be your friend even if you I know Dex will hug me after this even though it seems like it's not true.
18:03So broadly speaking, I don't think that comes from the model and it doesn't come from the loop.
18:08It's about the infrastructure you build around it and how you enable that infrastructure to actually enable you to take advantage of these loops.
18:15And as the models get better and as the underlying infrastructure and platform we build to enable these these feedback loops and loop automation and software development, whatever word software factor whatever word word we want to use for this conjunction
18:27of stuff that sits on top of a probability distribution, we will be able to have better guarantees, but we certainly are not going to be able to I do not fundamentally believe or ever believe that some type of alignment or reinforcement learning is going to result in a model ever being 100%
18:42safe in any capacity.
18:43And if there's any evidence, it's that as these models get better, the most important thing to remember is they actually become higher goal seeking and higher capable in terms of finding exploits to achieve their ultimate goal.
18:55No, I would concur with that completely.
18:57Um if you the the most concrete thing you can do to secure your environment is just not have secrets as files.
19:03Yep.
19:04Um if you ever seen the behavior where it wants to deploy a web service or what else have you and the token's not privileged enough, it'll start goal seeking
19:12on the file system looking for higher privileged tokens credentials.
19:16You do not want to get in the way of an agent wanting to do its a goal.
19:21Okay, so next question is directed to Jeff.
19:26Jeff, your original post from last year said Ralph was best for greenfield work.
19:30Today it seems engineers are running loops on existing code bases to improve latency, eval's, or refactor parts of their back-end code.
19:38What's changed that suddenly makes loops more broadly usable today?
19:42Um it doesn't matter how good models get, folks.
19:47The models have been good enough for at least the last year.
19:51Um what has changed is people's understanding of that.
19:56So society is only able to adjust at a rate.
20:00For example, I hypothesize it's it's Christmas breaks.
20:05Because the models back in November were released in November last year.
20:10In December, there was no real new models.
20:12What the difference was people had time.
20:14They could actually sit down and play with it.
20:15They they had the realization that these have actually gotten really good.
20:20So the reason why is it because it works, folks.
20:25These LLMs generate code better than you can hire you can actually hire for.
20:31If you think in the broad mass of software developers or coders, these LLMs generate code better than any software developer in the mass market that most founders can actually hire for.
20:43It it's it's sad but true.
20:45Um now why loops?
20:48It's really simple.
20:50Cuz if you run it in a loop, it works out to $10.42 an hour.
20:55Calculation index did back about a year ago now.
20:59Yeah, August we did the we did the hackathon where we we we copied all of the sponsor tools.
21:04We rewrote a bunch of Python libraries in TypeScript.
21:07Yeah.
21:08So, like concretely, like loops.
21:10I've come across so many engineering managers and founders and they've got these complex tech stack.
21:15They're running on four or five different programming languages, etc. And they run a loop and they've got good tests on these type of tech stack.
21:20And all of a sudden their complexity is they're just managing one tech stack.
21:24It works.
21:25Go to YouTube.
21:26You've got all these software developers now who are now software developers because it's software development as a profession has been commoditized.
21:34Some deep thinking to be there.
21:36And they're just on YouTube and they're like, "Yo, check out rough loop because it's a I went to sleep and I woke up and it works."
21:43Like boss, it's meme-y, it's punchy, it works.
21:47But there are there are problems with folks.
21:50Like I originally described it to should be only be used for green field.
21:55Because the models were pretty bad back then in science 35 days.
21:59But uh it is it is kind of inevitable at least with software.
22:03Because it's so easy to be verified.
22:06And the quality uh the quality of the code generated is better than most people can actually hire for or buy.
22:15Now, on your topic about architecture and taste, that's what the word engineering means in loop engineering, folks.
22:24Like your job now is to actually encode a file your domain to prevent the agent from doing a commit, for example, pre-commit hooks.
22:35They're fantastic.
22:37As a human, I hate them because they slow down the ability [clears throat] to do commits.
22:41But agents don't care.
22:43So, you can make a pre-commit hook that echoes out essentially a prompt that tells it say that this boundary here can't depend upon this and that and that's just feedback.
22:54That's a feedback loop on it.
22:56So, the engineering here is to prevent the loop from actually closing until it satisfies your engineering certification and your your requirements in the domain.
23:07So, it could be code formatting, it could be static language analyzers, it could be deterministic system testing, simulators.
23:15Like, let's put our engineering hats on.
23:16Like, we're kind of like locomotive engineers now.
23:20And it's our job to keep the locomotive on the rails.
23:23Because, um, to be frank, the model the model's a drunk.
23:27Right?
23:28You can't trust them.
23:29But, like, we accept that, but we engineer away those failure domains.
23:34We engineer away the failure way these failure domains.
23:37So, now is an inflection point.
23:40I guess Boris, when he first posted about Ralph back in November last year, I was like, "What the heck is Ralph?"
23:48Ralph is now, uh, it's essentially almost what, a year and a half old now?
23:54I saw it in June 19th.
23:57It's a a year and a year and 2 weeks.
24:00Yeah.
24:00But you had been working on it for months at that point.
24:03So, it was kind of weird because we had all the YC startups all just like autonomously compressing time to build their start to build up their MVPs.
24:13And that's also something that's quite scary if you're a business founder as well.
24:17Like, if you've got a incumbent startup coming and they're building autonomously, um, and they're running much leaner and the quality it's very easy for them to actually achieve those outcomes.
24:30That adds to some of these hysteria as well because it's the topic of in business competition being at your doors faster.
24:38All right, so I'll move on to our next question, which will be for Greg.
24:42Um, it seems like now is a large inflection point for Loops, like I said before, and compared to Jeff's announcement of the Ralph Loop a year ago, and even the widespread adoption we saw in late 2025,
24:52early 2026, um, the reason um that that may be like caught on was because maybe this is a new like capability stack where models can now process images better.
25:01They needed better verification.
25:03Um context windows got bigger and reasoning models improved.
25:07Greg, with all of these advancements, why is the way we're using loops today still wrong in your opinion?
25:12I mean, I don't think that model intelligence matters a lot anymore.
25:16I think it boils down to I agree with you the semantic verification, the actual ability to uh close the feedback loop, or however you call it, to actually verify that the outputs of the agent are correct.
25:28And you can do it to an extent, I think.
25:32I don't think you can do it holistically, at least not at this point.
25:35I think you can do it to an extent to things that are deterministically verifiable.
25:39Uh you can get better typing in your system.
25:42You can get better linters.
25:43You can get um simulation testing and all of that.
25:45You can start keep adding that.
25:47And as long as you keep those cheap, I think that's fine.
25:50Um the moment you start adding even more non-determinism as your verification process, I think that becomes less and less correct.
25:59It starts contributing more like you know how if you prompt agent with one thing and there is a 5% chance it's going to have an error in it.
26:09And then you start looping that.
26:11Then suddenly after 10 20 loops, it's going to be 50% chance it's correct or maybe less.
26:16And that's what I mean.
26:17And it just costed you so much money to do that.
26:20I'm going to be keep coming back to the economic viability of all of that.
26:23Um And but but to base it a little bit in like evidence, I'm pretty sure that majority of large AI companies are still using Sentry.
26:32But why is that?
26:34They they are using that just to catch simple bugs as well.
26:36It's not security bugs.
26:38It's not performance regressions, et cetera.
26:40Those problems still exist in the way that we are looping now.
26:43And we haven't solved those problems uh yet.
26:47Thank you.
26:50For Dex, the Raffle pioneered the idea to feed fresh contacts into each iteration to avoid context rot.
26:56This has become even more manageable now that context windows have gotten much larger.
27:01Decks, are we out of the woods regarding context rot and context engineering?
27:06I'm going to answer your question, but is there going to be like an open floor part cuz I have more questions for Jeff.
27:12We were trying to do Oxford debate style for this one to keep it like more structured and then prevent like just
27:17You don't want it to just turn into a chaotic yapfest.
27:21Yeah.
27:21to prevent what you and I do where we just start yapping.
27:23Yeah.
27:24Yeah.
27:24It's already started.
27:25I yeah, I was trying to control like both of you guys this time.
27:28Okay, I'm going to do this answer as quickly as I possible and then I'm going to start busting Jeff's balls.
27:31Yeah, you can say whatever you want with your time.
27:32You can just yeah, that's that's all good.
27:34Okay.
27:35Uh so yeah, the the coolest thing about Ralph back in the day was like, okay, you keep clearing the context window and like is it completely efficient?
27:43Like probably not from a token perspective, but it meant you could leave a thing running overnight and it would never like if you just kept stuffing messages in, you would overflow the context window.
27:51But if you just relaunch it, say, here's my desired state of the world, go check the code and see what we have and do the one next step to get us there.
27:58Uh it was a very clean way to keep most of your work in what we call the smart zone of the context window.
28:04If you tell just do one thing and then we're going to clear and restart.
28:08Um context windows have gotten longer um and I will like give an update.
28:12I think I gave this in Miami, but that that video is still in production.
28:15Um The the dumb zone is really as as much as anything else is is a it's more like training wheels.
28:23Like if you have been talking to Claude for 70 70 hours a week and for two to three months, you probably don't need to think about the smart zone versus the dumb zone cuz you've built your intuition.
28:33It's a guideline.
28:34If you're just getting This is why we teach people this is like it's a guideline if you're just getting started with AI, try to keep it around 100,000 tokens.
28:41For larger million context window, we probably don't revise the revise this up to like 200,000 tokens, but I've regularly tried to keep it under 60 for the hardest problems.
28:50I've regularly gone over 300k for things where I'm just like kind of like riffing with the agent and I'm just like too lazy to compact it and and and move on and do a new one.
29:00Um but this is your intuition.
29:02Like one of the telltale signs that you're in the that you're in the dumb zone is like uh there's certain cases where the model, you know, you're 200,000 tokens in and the model's like finished some work and it's trying to get the test to pass and it's like not working and it's like doing all these weird hacks
29:15and you read the thinking traces and it's like oh that's a test, but that's from something else and I don't need to fix that and that's a preexisting thing and you're like well no, it's not.
29:23And that that is the moment, that frustration where you're like okay, it's it's flailing trying to make something happen.
29:28That's the instinct that you a lot of people I think cultivate after a couple months working with these models.
29:33Uh but if you don't have that yet, then you know, then then this is our guideline.
29:36So yeah, context windows are getting better.
29:38I think they're getting bigger.
29:39Um and so like the core like Ralph loop of like do as little as possible in every single iteration is like less of the motivation here than, you know, the more feedback you can pipe into the system,
29:49the more you can do autonomously and if you can have deterministic things making decisions and building small prompts to give to an agent and you don't have to remember to do that.
29:59You don't have to tell it, "Hey, go check the PR comments and fix them."
30:02and then wait and then someone makes another comment and you come back 3 hours later and say, "Oh, check the comments again."
30:07If you can automate that process, that's great.
30:10And that's kind of the core of I think what is like loops stuff that works today.
30:14Um
30:16Don't ask Jeff about that.
30:17That's all I have time for.
30:18I'm sorry.
30:18I have to start like really keeping this on schedule.
30:21Um okay, so now we're going to get into the anatomy of what makes a good loop a good loop.
30:26So okay, part of what makes a loop good is verification.
30:31However, it seems contradictory that people are saying our job is to stop writing prompts and start writing loops when the loops with bad prompts result in agents cheating and meeting its goal by by the test instead of working to pass them.
30:43Jeff, how do you keep the model from cheating when verifying its own work?
30:47Um I heavily exploit pre-commit hooks, folks.
30:50Um and I engineer in that back pressure uh by analyzing the work that is done.
30:57Um the other thing I do is um as Dex mentioned that with Ralph, it was one of the things was everyone was trying to do compaction.
31:06Think about compaction is kind of like a lossy function, like uploading a a video to YouTube and then downloading and uploading it 100 times.
31:13Like you're losing fidelity there.
31:16And it's already a non-deterministic system, probabilistic by thing.
31:20So, the the theory behind how Ralph came to be is like, "Okay, there is a dumb zone.
31:25And what I want to do is deterministically allocate everything it needs.
31:30Cuz if it's if it's not allocated, then it's essentially the the search space of what it can do is not constrained.
31:37But also leaving a bit of a headroom.
31:39Leaving a bit of headroom.
31:40So, I also I get meat sweats when I go above 100k, even with these million context windows.
31:46And this is really important to think about.
31:48Um a lot of people they think that you want to use LLMs um at a company.
31:53And it's like, "I've got this data."
31:54It's like, "Sweet.
31:56Okay, you're going to have to use a loop to batch this data."
32:00What you need to think about the context windows is uh essentially remember the 720k floppy disk.
32:07You know what I mean?
32:07You've only got about a eighth of that floppy disk of usable memory that you can actually use for an LLM.
32:13So, you actually have to batch it.
32:15You can only allocate roughly around about Star Wars if you get Star Wars Episode 1 movie script and you tokenize it, you can actually just hold two of those movie scripts in memory
32:26before the context window is called.
32:27That's around about 150 kilobytes of data on a text-based movie script.
32:32So, be even very careful about this.
32:35Something I've done it a long time and it's very silly, is uh I run a model bare um without any skills or any markdown.
32:44Actually, I get rid of all my skills and all my markdown and everything when the new model is released because the the the models actually have tastes and preferences.
32:54For example, GPT-55 when it first came out, if you screamed at it in uppercase, it became weak and timid.
33:01But if you use Anthropic, it wants you to yell at Go read the model cards, folks.
33:05Like for the integrators, like there is unique tastes for it.
33:10So, keeping it on keeping it on the rails is actually it's engineering.
33:17It's really engineering.
33:21Thank you.
33:23Around 10 days ago, Jeff coined the term convergence engineering.
33:28He said it's where your loop slops together.
33:30Your loop uh it's where your loop slop uh comes together as a discrete like system under task until it converges.
33:36Jacks, what is wrong with how we are using loops today?
33:39How do we ensure looping slop together doesn't just produce more slop?
33:45Uh we got to read the code.
33:47I will actually highlight uh an experiment that Jeff did earlier this year.
33:51Uh I believe what was it called?
33:53Loom?
33:54Where we had Ralph loops trying to build a software platform for the future.
33:59And I think you built you built AWS and you built GitHub and then you realized, "Okay, how do we how do we get the model feedback on things that it's not good at yet?"
34:07Like UI testing and things like this.
34:09Well, okay, the way you create a loop for is this UI good?
34:13Is you give the model something like PostHog, where it's like, "Okay, we can deploy multiple different experiments.
34:18We can see which ones the users use."
34:19And then rather than looking at screenshots and PNGs, the model can look at data and see, "Okay, uh this one is performing better.
34:27That must be the right color for the button."
34:29And so now you've even removed like the human visual taste from the equation.
34:33And all of this sounded really cool and in the point of like how do we ensure looping doesn't bring slop together?
34:39I I don't think you can.
34:41And this is like a perfect example of the hype outrunning the discipline in the sense of Jeff, what's what's going on with Loom now?
34:49It's still there.
34:50It's on GitHub.
34:51Um but are you still working on it?
34:54Uh it's been 6 months because I've been looking into engineering ways of verification.
35:00Right.
35:00What was the thing you said to me?
35:01You said Loom's not going to work until we get better programming languages or we get better much better models.
35:08And that is a textbook for me of the hype is outrunning the discipline.
35:12We're really excited about all this stuff and by the way like everyone should do what Jeff did.
35:16Like Loom is awesome.
35:17Like go experiment, try to push the frontier cuz that's how you learn where it is and what's possible.
35:22Otherwise you just keep using your old skills with every new model and you assume it has the same limitations.
35:27But it's also a a a key point of like I don't know what I'm trying to say.
35:33Like that this it doesn't work yet.
35:35That thing doesn't work yet.
35:36It will work someday and like there's inevitability but again it's what works today versus what is hype.
35:43So I don't know if that fully answers your question but the answer is like the way to not loop slop together and make more slop is to like read the thing that's coming out the other end and make sure it's not slop.
35:54Yeah, that makes sense.
35:57No, it's it's it's The labs haven't cracked it.
36:01So what makes you think you're going to crack it?
36:03Yes.
36:04Right?
36:06And this is right now it's we're all trying to figure out how to make this all work.
36:11For sure.
36:13Skeptics say that loops fail quietly.
36:15They either spiral forever on your dime or the agent declares victory early on on a half-finished job.
36:22Exacts are already starting to question token spend.
36:25Greg, when does a loop pay for itself And how often is this actually the case?
36:29I don't think they fail quietly for us.
36:32[laughter]
36:33I think they fail very, very, um, loudly, especially when you're looking at your bills.
36:40Um, but there are cases like, as I said, I do loops, or I do engineering, I would say more so, and there are cases where I think doing loops is very valuable, or doing,
36:53um, or making a explicit decision that you're going to pay pay for the cost is very valuable.
36:59So, the concrete example here is we do security scanning after on our PRs in local, and after our PRs even land, because they will always find some things that are real,
37:10um, that we have overlooked, and they sort of beat humans on the on the code review.
37:15And it's expensive.
37:16It costs us, I think, like five bucks a PR, or something like that, to run all the checks that we want.
37:21But, that's where we made an explicit decision that it's worth it.
37:24Um, there are also cases where, like, if you look at the very, very well-specified systems, such as, um, all the experiments with, uh, Next.js rewrite, or a bun rewrite in rust, or, uh, running a browser,
37:38cases where you have years and years of test suites and specifications built in around those problems, where you can really, really where well verify, um, the outcomes, then we then
37:50looping and getting to those, um, those results seems to work.
37:55Bun in rust seems to work pretty well, from all I can tell.
37:59Um, so, there are definitely cases, and then there are cases of, um, usage that I think that I do pretty often, where you can imagine, for instance, building prototypes.
38:09I do prototypes of products that we should be doing uh doing at Sentry pretty often.
38:14Those are going to be throwaways, so I'm going to just slash gold on them, and forget about them.
38:18And if we like them, then I'm going I'm going reading the code, and I'm going to be mortified, uh, and we're going to go to the square one and start specking out what we actually meant,
38:28and and sort of go towards that solution, but it's going to be much much more involving of of human in a loop.
38:33Um so so broadly, I think they have place, uh but as Dex said, they the hype is what I have a problem with.
38:40The hype is outrunning the discipline, as he said.
38:43And I also agree very strongly with the point that you should just try things.
38:48You should just experiment yourself, try to see what actually works for you, where where the the cookie crumbles, um and you know, spend less time on Twitter, I think, is healthy nowadays.
39:01Yeah, good points.
39:03A good cost-conscious loop has to track state to know what it's already tried.
39:08That memory lives somewhere on disk and get increasingly in a shared memory store that many agents read and write, especially as you go single-player to multiplayer with agents.
39:18You end up with this access control problem that can't tell which agent wrote which memory and who can read it.
39:23That might be fine for a POC, but it's definitely not for production.
39:27The tension lies in this.
39:29Shared memory is what loops use to learn from each other and converge faster, but scoping it per agent to solve the access control problem isolates them and then kills that shared learning.
39:40Ian, how do we solve the shared memory store access control problem so loops can converge faster?
39:47Great question.
39:49Wow.
39:49Wow.
39:51Only if someone was working on a product that could help.
39:53Only if someone was thinking about it.
39:54I mean, actually, I've This is This is an unsolved problem, first and foremost.
39:58Like, let's be really honest that our our access control systems weren't designed for this this world, where machines were acting and reasoning on on behalf of uh on behalf of us.
40:07But broadly speaking, I think like some of the beginnings of the substrate are starting to emerge, and if we ignore Cloud IAM and all the other stuff for a hot second,
40:15like markdown is pretty great.
40:16And so really, if we were to say a memory is markdown for the purpose of the conversation, the really question is like what things and how do I share these markdown files and use that as a memory
40:26and then how do I attribute sort of like access control around those things.
40:30And if we were to use that model, I actually think we have like the basis for for most of it today.
40:35It's just unwieldy to think about.
40:38Um a good example would be I did this tweet recently.
40:41Um I was playing with Notion and we use Notion a lot at GitHub.
40:45Uh but I really just wanted all my Notion things to be able to be as markdown files.
40:49And because it would just made it easier for the agent to work with it instead of going through MCP and I was a bit of a CLI maxi and was looking through that.
40:57So I think what we're missing really is uh and MZ it you know Dex and I debated this last time about MCP but like we're really the challenge really is how do I present a world to an agent so that I can understand it and then how do I attribute what can access at any one point in
41:10time and how to make that wieldy for anyone to do it.
41:12And I don't I don't think we've actually cracked it but certainly there's some beginnings of patterns that make a lot of sense.
41:20All right, now that we discussed the anatomy of the loop, we're going to debate the future of the loop.
41:25Like are we essentially well positioned now for software factories?
41:28Has loop engineering gotten so good that like we're ready for the full software factory?
41:32And we'll start with if loops are only good for verifiable tasks, that means fully autonomous software factories must be able to verify everything they do.
41:42Greg, is this realistic?
41:43What other parts of good engineering work such as deciding what to build, whether the abstraction is right, and what tradeoffs are acceptable?
41:50Uh is this realistic?
41:52I think if if the compute is free, that would be pretty pretty good beginning but um
41:57[laughter]
41:58uh I think I think we're getting to the point where um or let me put it this way.
42:05The decisions that you are making as a human in in the loop are the decisions of like design, architecture, the the the important ones that um that I would say I wouldn't trust the the agent to do for me.
42:19Uh and I don't see the future where that becomes reality yet.
42:23I think the reason for that is when you're when you're looking at large organizations and I think any engineer who has had like years of experience will tell you it's not always about what you should be what you should build, but also about what you shouldn't build.
42:36What are the actual right trade-offs, where the complexity is that you want to um that you want to invest in versus where you should be um investing in maximal simplicity.
42:46Um in my experience, agents love complexity.
42:49They will keep adding to the stack um unbounded.
42:53And so, I think we are we are shifting the the the post.
42:58We are getting to the point where as we are adding more validation, more uh semantic verification, they are able to do much much more.
43:07Um and I'm not uh not neglecting that.
43:10But I do want to be in the loop for the actual architectural decisions, and I do not see them uh taking that over any anytime soon.
43:19Thank you.
43:21Shubha Vice Head of Engineering told everyone at Crossover's Compile Conference that your job is just to write loops.
43:26Steinberger tweeted, "Here's your monthly reminder that you shouldn't be prompting coding agents anymore.
43:31You should be designing loops that prompt your agents."
43:34The view count on that tweet was 8 million people.
43:38Jeff, you've said in your original blog post, "Loops need senior expertise and it sometimes it tops out at 90% of the way there."
43:47So, is just write the loops advice, is that safe to give to 3,000 people or 8 million people?
43:53Or only the people that have really learned to tune a loop in the first place?
43:59Yeah, this is an interesting thing.
44:00Um it's really hard to tailor and teach like like what you should do and what you should not do to a broad that broad of an audience.
44:10So, all I can do is write to when I was originally writing, I was writing to my peers.
44:16I had people I looked up to.
44:19For some reason, I had it clicked in my head that everything has changed.
44:22The game has changed, but I was really concerned that some of the some of the best people in functional programming and like and like real peers.
44:30I remember like um Fly IO, Thomas Petek.
44:33He he he that made him click in his head.
44:36He's got a blog post saying as such.
44:38So, I started writing for my peers and the people I looked up on.
44:43To disseminate down to like the the entire world, it's hard.
44:49Um but in the same the same sense, you go to YouTube.
44:53You got people who are creating things who've never created things before.
44:56They go to sleep and they wake up and they got a brand new Discord bot.
44:59It's magic.
45:01I want people to remember that it is actually a pattern for allocating basically it was a pattern as an orchestration pattern.
45:13It was condensed down to a while true loop using cat cuz cat is the simplest teaching primitive.
45:19To that If you want to teach something, you got to make it really simple.
45:22Make it really simple.
45:23So, cat prompt I you engineer the prompt what it's going to be.
45:27You use the file system as state.
45:28You recycle the context window.
45:30Run the loop.
45:31It's a little bit maybe, but it got it was just bang for buck.
45:34It just works.
45:36But, the entire attempt there is that there should be some sort of like PID controller on top or some sort of factory or some sort of some sort of some sort of determinator
45:47as such saying whether a loop should continue or not.
45:52Um it's really hard.
45:54Um is it safe or unsafe?
45:57Safe is an interesting word.
45:59Um no one should be using coding tools on your local laptop.
46:04And this is not because of AI, but this is because of NPM supply chain attacks.
46:08This has been true.
46:09I tried to try cracking this problem for 7 years.
46:12Um and like it's now on the attention again that we that like AI could be unsafe running unsafe commands, but like your software development practices day-to-day at workplace are already unsafe, folks.
46:26So, um fix that.
46:29Then these techniques start to become safe.
46:34Thank you.
46:36Dax, you tweeted that most engineers are seeing a two to three x speed up from coding agents and that's realistic.
46:42And if you try for that 100 x speed up, you're going to get lost in the meta meta problem of optimization.
46:47And you may never get to that life-changing 10 x speed up that is possible by staying pragmatic.
46:52If only two to three x is what's possible, how do we ever get to a fully autonomous software factory?
46:58Yeah, and I think this is it's it's a good question.
47:01I think there's like there's there's there's a mix here, too, right?
47:04We're talking about what works versus what is hype.
47:06And I think it is definitely worth I want to highlight like you should try again.
47:12Like you should try to push the frontier and do the things that might not work today, but you should not assume that all your work is changed just because you saw something work on Twitter, basically.
47:23Is like don't throw away all the things we've learned.
47:26Don't go out of your way to cast aside this, you know, decades-long career of software engineering that we as a we as a community have built up and put together.
47:37Um and it it gets back to like what I see as the biggest anti-pattern for how people set about designing and creating their software factory, which is they say,
47:49"Okay, cool.
47:49I'm going to go away for 3 months and I've read a bunch of blog posts and I'm going to go make my software factory."
47:55And it's going to be the software factory.
47:56It's going to be the future of how we ship everything.
47:59Uh and then you come back 3 months later and you never touched the problem.
48:03You never put it in anybody's hands.
48:05It's just like any product.
48:06You're building a software factory, you're building a product for your teammates, you're building a product for 5, 10, 500, 5,000 engineers.
48:12Uh and the the right approach is to start small and iterate and figure out what works.
48:18Try things, they might not work, they might work, but the way we learn AI and how to use it effectively is through building up intuition, which is why you should try a bunch of stuff that probably won't work, but you should acknowledge that.
48:29You should not try to push like through that through that frontier.
48:33Um and so my advice is kind of like instead of trying to automate everything end-to-end, build these small incremental loops throughout your system and you will wake up one day and you will be moving two to three times faster
48:44while still being able to read the code, while still owning the architecture.
48:47And so you don't have to throw away everything we've learned and everything we know and all your intuitions just to get to this place.
48:54And so I would I would caution people like go figure out how to make move 2x faster or 3x faster because you're going to like blow everything up by trying to go 100x faster.
49:04Um and uh yeah, I mean, if can you imagine if every software engineer in the world was two to three x faster and had a like near human 99%
49:14level human level of quality, that would change every single enterprise in the world, every single startup in the world.
49:20It would change the entire math.
49:22And we're trying to like go to Just don't don't go too far.
49:26Shoot Shoot for what you can do.
49:27Build things up iteratively.
49:29You'll learn a lot and you'll be ready for when the next models come out and you can go 5x, 10x faster.
49:36Thank you.
49:38When it comes to verification, it's not just about verifying the work, it's also about verifying who did the work.
49:44Employers have made it clear that humans are responsible for the code they ship whether an agent wrote it or not.
49:50Yet only one person or agent can sign a commit today.
49:54Ian, are we ready for software factories to be writing and reviewing all code if we can't determine who did the work and who the work was done on behalf of?
50:02I think Yeah, actually I was I was playing around with this problem like two weeks ago.
50:06So, um first and foremost, we have a problem.
50:08Git only allows one signer on a commit, so we got to fix that.
50:12Um two is you know, there's some things around SOC 2 and compliance, right?
50:17But, I think like more importantly, the way to think about agents is is kind of how we think about service ownership in large organization.
50:23It's like at some point, a human has to be attributable for an agent's actions.
50:28Like, there's no world where it's going to be like an agent is its own entity and its own attributable thing and somehow has liability.
50:32Like, the only people who have liability are people that can have consequences, right?
50:36And that always has to be grounded in it being a human.
50:38And so, if a human designs a loop and that loop presents bad software, guess who's attributable for that liability?
50:44It's going to be the human, right?
50:45Like, the society doesn't function if we don't have liability.
50:48Uh it just doesn't work.
50:49There has be consequences to damage, there's be consequences to bad decisions, and what those consequences are are obviously a gradient based on the on on the damage that's done.
50:58And if we don't have that, nothing works.
51:00And so, I'd actually broadly say there is no world where a human is never responsible.
51:06There's always a world where a human has a level of responsibility.
51:08And the question is whether it's a human or corporation, which is a group of humans, um the question is is how does that change the way that our systems work today?
51:16So, today with Git, I can sign a commit that says I did it and it attributes to my public key, and that's cool for last generation's way that we thought about software.
51:26Now, in the future, when I tell an agent to go do something on my behalf, it generates a loop, it generates a bunch of code, it goes to production.
51:33I have to be attributable to the initial fact that I signed that thing, the that intention to do that.
51:39And we we simply do not have the substrate required um to do it, although I do think that's going to change pretty quickly.
51:46Um this isn't like something that's like crazy difficult for us to break, but we have to rethink the way that we think about attribution across the supply chain and the SDLC.
51:55And this is not a new problem, right?
51:56Like supply chain security has always been an issue.
51:57It's actually the biggest challenge one of the biggest challenges we have with agents, and we've always wanted to have more deterministic uh pathway and sign signature chains across the supply chain, and this is just
52:08how do you do that for first-party code versus uh third-party code?
52:12Well, Ian, um, to to just to shoot straight, um, our profession is a bit of a clown show.
52:17We actually don't have liability at a personal level.
52:21Like we call ourselves engineers, we're not really engineers.
52:24Some of this stuff is going to get really complicated, folks, and maybe we need to revisit these topics.
52:31And now that we've had our main debate section, to wrap it up, each member on each team will get 2 minutes to uh wrap it up and describe their final thoughts before we decide who the winner was.
52:44Uh, Greg, what would you like to
52:45Yeah, absolutely.
52:47Um, it's it's interesting seeing how many how many points we actually agree with each other, uh, but that's how how this goes.
52:53I I do think that the biggest point that I'm trying to I have been trying to drive here is um where we are whether we are where we are trying to be and then what you're actually
53:03hearing from the ever-present hype uh hype loop.
53:08And that's where I want to sort of double down and tone this down.
53:12I basically want to say try things.
53:15Think for yourself.
53:16Don't lean into the bubble.
53:18Uh, don't lean into the hype because you will find what works for you and you will find where the system breaks the best by doing it yourself.
53:24That's always has been how um how humans learn the best is by practice, not by, you know, watching YouTube.
53:31Um, and ultimately, that's that's what I'm trying to do, and I'm slightly skeptical uh when it comes to, you know, allowing the full loop to run because I see the qualitative results not being up to snuff for my
53:46my requirements.
53:47But, I am optimistic in general.
53:49I am optimistic because I've seen how much more I can do and what are the types of problems I can address today that I couldn't address even a year ago,
53:56let alone let alone earlier than that.
53:59And and it's going to get better, but I'm also not worried about my software engineering career.
54:05I don't think we're going away.
54:07I think we are still going to be important piece of this whole software factory or whatever the next um sort of hype bubble is going to become.
54:18Ian, would you like to go next?
54:19I would love to.
54:20So, the train's left the station.
54:23This stuff works, um and there is a real productivity increment, right?
54:28Like, more stuff is getting done.
54:30It doesn't mean all the stuff getting done is good, but more stuff is getting done and a good percentage of that stuff is actually good, right?
54:35I think we can agree that.
54:36And second is now we have competitive dynamics, where it's no longer possible for a company to sit back like, "Hey, we're going to sit this loop stuff out, right?
54:45I'm going to sit this coding agent stuff out."
54:46Like, that that's over.
54:47We're we're past that.
54:48Like, that train leaves the station.
54:50As soon as that train leaves the station, everybody in the world starts saying, "Holy I got to keep my stuff together.
54:55We're going to stay up with the Joneses."
54:56And they have to because we live in a very competitive capitalist society, and that's also, you know, ducking politics, but it's it is what we are.
55:02And at the end of the day, the question is is not really um will it happen, it's when it happens and to what degree over time.
55:11And I don't think there's a choice but to actually stay up-to-date, and we're all just kind of holding on to a rocket ship ride, where we don't actually know what that We don't really know the trajectory other than it feels real fast
55:22with crazy acceleration, and sometimes like we're way ahead of each other, and sometimes we're way behind, but I don't think you have a choice not to be one, figuring out what loops are, two, figuring out where you can apply them in your code base.
55:32There's going to be places where, "Hey, this is highly verifiable.
55:35This is a problem that like computers can solve.
55:38It doesn't make a lot of sense.
55:39Cough connectors is a good place of like, you know, last 10 years.
55:42The amount of companies that have made money because they're basically just connector farms, there's no longer a lot of value there, right?
55:48If you can automate the connection connector creation.
55:51So, there's places in software that you can apply loops today and get real value and there's places where you probably shouldn't and you should decide where that is and it's probably the core value of what you're creating, right?
56:01And so, that's how I think about it.
56:03Broadly speaking, trends left the station.
56:06The productivity curve is what drives society.
56:09The ROI, so how much more productive we are as a society is what drives GDP.
56:13What drives GDP is ultimately where the dollars go um and how capital gets allocated.
56:18Okay, get over to you next.
56:20I eagerly await the world where the life soft software factory is feasible.
56:26The I I would love a world where we don't have to read the code, where where we can just do everything.
56:31Um if you watch my talk on Tuesday, I I think this is actually a problem we can only solve at the model level right now.
56:38I don't think the harness can do it uh unfortunately cuz I love building harnesses and doing context engineering.
56:44Um so, my advice is uh pay attention to Jeff.
56:48Uh let me know when Loom is actually working and until then uh use loops, but not like that.
56:55[laughter]
56:57Love it.
57:00So, so much to say, folks.
57:01Um factories represents where we're heading to the future.
57:06Like it it's it's essentially like a perpetual energy motion machine, right?
57:11It's it's like it's the pipe dream.
57:14Companies are only just getting founded today and getting receiving their rounds today.
57:20Don't think you can just take this and implement it in your company cuz it's just generally not solved in market.
57:27But, I will say to add to the meta monologue, if you try to run in loops or try to build a factory to using Python, it's going to be a clown show.
57:39[laughter]
57:39If you do it in Ruby, it's going to be a clown show.
57:42Static types are a form of verification, folks.
57:46I encourage you to to come up with a couple of cutters and a couple of experiments.
57:50Try running some loops.
57:52Build an application in Ruby and then try to modify it again with these loops and you'll see the maintainability mess.
57:59And then try doing it with Haskell.
58:01I don't care if you don't know Haskell.
58:03It doesn't matter if you don't understand Haskell.
58:05The LLM understands Haskell and you can actually prompt the the agent to explain this as if you to to your son or daughter.
58:15So, I'm not sure even code needs to be readable these days, but this is frontier frontier thinking.
58:20It needs to be explainable.
58:22So, I'm playing around with different domains of verification here, but uh one thing's for sure, um type systems are in.
58:31Very much in.
58:32Rust is very good because how the ecosystem models some types.
58:37And because supply chain was mentioned, I need to say it has been 10 months now.
58:43I do I minimally use any open source software.
58:46Minimally.
58:47I just generate it to my requirements.
58:50And then when a supply chain attack happens, I'm like, "Didn't affect me."
58:55All software has security flaws, but it's about minimizing the blast radius.
59:00And the thing is, if you deal with open source project, the person's on on leave, the maintainer, what have you, you can't talk to a human.
59:08That's not AGI.
59:10You want to be venturing all your source code as much as possible so the so the agent can actually modify and in without a loop or other prompting.
59:19You need to own your supply chain.
59:21Awesome.
59:22Thank you so much for that.
59:23I personally have to close it at that, but we can now decide our winner.
59:28Um Show of hands in the audience if your mind was changed and you're now on Dex's side, raise your hand.
59:37Let's go.
59:38Okay.
59:39If your mind was changed and you're now on Ian and Jeff's side.
59:43Yeah.
59:45I kind of couldn't see with the light.
59:46What did you guys think?
59:50It was pretty good.
59:51It's pretty close, yeah.
59:52It's impossible.
59:52The lights are so bright.
59:54All right.
59:54Well, that was a good debate.
59:55Thank you so much for listening.
59:57We'll see you next time.
59:58[applause]
60:13[music]