Daniel Whitenack: I think there’s still open research question ones here that are worth exploring, and then there’s workflow and practical implications, I think. On the first side, as was mentioned already, and we were discussing this reward model, as far as I can tell, it’s not totally determined what the architecture of this reward model should look like, how big it should be in relation to the model that you’re fine-tuning, how much human feedback should you use, how does the amount of human feedback that you get influence the harmfulness or the utility of the output, and that sort of thing… So I think there’s a lot to explore around that dynamic between the reward model and the language model.

In addition – I mean, language models are still being developed, right? So ChatGPT used the GPT-3.5 language model as this original policy, right? And they actually used the fine-tuned version of that, using supervised methods and human chat conversations. So they started with a fine-tuned version of GPT-3.5. So obviously, we’re going to have a GPT-4, GPT-5, we’re going to have other language models from other providers, right? From other research groups - big science, or Google, or Microsoft, or whoever’s developing these other language models, we’re going to have updated versions of those.

So I think we can see a research direction with this where people are trying different pre-trained models as their original policy, where people are trying different reward models, or they’re mixing them up in interesting ways, where they’re maybe using slightly modified versions of the PPO algorithm, or other reinforcement learning algorithms to do the updates… So there’s a research direction where I think we’ll just see a lot of exploration with this kind of template as the structure that they’re exploring.

The second piece, which is maybe more interesting to some of our audience, is what are the implications of this in terms of people’s workflow…

Johnny Boursiquot: So if we do get a sponsor - which, we are looking for sponsors, obviously - the money won’t be towards lining the hosts’ sprockets. It’s gonna be for logistics, getting the shows editing, and all that good stuff.

Adam Stacoviak: [32:07] Awesome. Well, that’s a big part of producing podcasts, we feel - quality content. There’s so many facets to the idea of quality content. It’s not only the content you’re creating and making it engaging and informative, but also taking a positive stance towards good mics, good post-processing, good EQ, good mixing, and the awesome website we’re launching on now… All the work behind all these things - there’s so many moving parts… I never thought I would ever be part of producing a podcast. Jerod, my mind is blown by how far we have taken our desire for quality content. [unintelligible 00:32:48.24] But it’s fun, though.

So let’s wrap up with the plan for season two. Unless there’s anything else y’all wanna share about season one, let’s move on to some plans on season two. Anything before we open that up yet?

Johnny Boursiquot: Exactly, exactly. Things of that nature. I’m lik, okay, so it sounds like the structs package is inching towards sort of more efficient, or some different way of lining up and doing some struct packing, I would imagine…

Jeff Smith: No, I’d be glad to. As you mentioned those three things, it is the perfect storm, especially for topics today… And it was a journey to get here. And I can give you a little bit about my journey and how we ended up with AI and mental health.

I would say in my background - I’m a corporate guy gone good. A classically-trained entrepreneur. I built six companies, three nonprofits… It’s where I find the joy. Identify a problem in the world, create a unique solution, wrap a company around it and build it to scale.

And this one’s called Chrp, named from the story of the canary in the coal mine. When that bird stops chirping, you get the heck out. It can send signals that we cannot. Methane, carbon monoxide… And so it becomes an early indicator for health and wellness. And we’ve created a platform harnessing that, using music.

And so to go even further back on myself and how I even got in this business - for years, I was the go-to guy for most ad agencies in New York to do all their social impact branding, corporate social responsibility… I became an expert on weaving purpose into the brand narrative, and bringing people alive at work, and through the products, and the brands, and these global campaigns… And along the way found a significant disconnect, I would say, between the leadership – the leadership that cares. They care about purpose, they care about their employees, they’re throwing millions at perks, rewards, telehealth, but their employees aren’t feeling it. They’re not feeling seen, they’re not feeling heard. There’s work-life enmeshment. They’re depressed. They’re anxious. They’re looking for jobs. And so a few of us decided “Hey, let’s take that on.”

We created a small company to come up with solutions to address workplace flourishing, and thought “That’d be kind of cool. Let’s bring people alive in the workplace.” And we started there, and said “Hey, what’s this disconnect between leadership and employees? What’s the problem? If there’s intent and there’s resource, but not results, where’s the breakdown?” And we’ve found that it was an information problem. We’ve found it was the corporate survey, the quarterly polls. How are you feeling? Nobody answers it, they lie in their responses, fear of reprisal, and they’re just making blind bets.

The data they’re getting back is at fault, mainly because people hate surveys. They want to fill them out. And so we said, “Let’s start there. Let’s come up with a better diagnostic tool, so people can feel seen and heard and where they’re truly at. And how do we do that?” And we set out to improve the survey; the survey tool, different modalities, different lengths, maybe even the happy faces you see in the bathrooms at airports… You know, just make it super-simple. And none of those are really tracking after a few months, and I was training for a Spartan race. One of those crazy things that we do to keep ourselves alive… And I found my mood changing with my music. I switched to, I think it was Motley Crue, Kickstart Your Heart, and just found my energy level shifting, and started thinking, as I was running, “What’s happening here? Am I being affected by the music? Am I making certain choices in my music selection that’s a reflection of this?” And that was where I had the a-ha moment, to say “Hey, is music the signal that we’ve been looking for? Is that a reflection of how I feel?” And so we dug into that, looked at music science, listening behaviors, research AI, and found a direct link, as music is a mirror for your mood.

In the simplest form, Chris, if you’re driving in the car and you’re listening to the radio, and you change, change, change until you find a song you like, that’s just your mirror neurons lining up your emotion with that song. it’s how you’re feeling or how you want to feel. It’s very hard to listen to music you’re not feeling. It’s that grind… And so we thought, “Okay, if we can bottle this up, we’ve got a rocket ship. If not, we’ll sell the algorithm and move on to the next task.”

[08:07] And so fast-forward, raised a bunch of capital, surrounded myself with brilliant people, technologists, HR leaders, music… So a buddy of mine, Suman Debroy - we’ve built some amazing things together. As a doctor in machine learning, he jumped in to help figure out the models… HR leaders from enterprise companies, managing hundreds of thousands of employees, speaking into “What does that experience need to look like inside of the company?” The music industry, songwriters, musicians, former execs from the big music streaming companies saying “Hey, this is the data that’s available to you, or even the intent from the musicians.” And that was phenomenal, to understand what were they feeling when they wrote a song? What do they want to put out in the world?

And so that was fascinating. And then even the attorneys, legal counsel. We’ve got the former privacy chief of Homeland Security to really look at “What are the privacy blockers? How do we hold integrity in this conversation?” Because music is so personal. And so we brought them together and said “Hey, let’s solve this problem. Music is our answer.” And like any, I guess now a new tech company, you’re testing it across an alpha group, looking at everything: adoption, the science… You end up with a black box on the table, it works beautifully. And so that was like end of last year. Now you’re shifting to product-market fit. Who is this best built for, right? A healthcare company at 2,500, a sports team, automotive company? So that’s where I’d say the rubber hits the road, and where we’re at today.

And just incredible leaders - you mentioned Greg Enos and others - just looking at this, saying “Hey, here’s a direction. Let’s really look at that how we can apply it.”

Jerod Santo: It’s time to make computing personal again

Benj Edwards deftly describes how “surveillance capitalism and DRM turned home tech from friend to foe” by asking a litany of rhetorical questions about the past:

• What percentage of your income had to go towards annual software subscriptions on a 20th century Windows PC?
• Which part of this TV set kept track of everything you watched and then secretly sold the data to advertisers?
• Which part of Windows 95 fed you ads without your consent and kept track of everything you did remotely so Microsoft could keep stats on it?
• Which part of Amazon.com in 2000 tried to get you to buy millions of no-name counterfeit and dangerous goods propped up by stealth advertising and fake reviews?
• Which part of Google in the 1990s and early 2000s blanketed its results with deceptive ads or made you add “Reddit” to every search to get good results that weren’t overwhelmed by SEO-seeking filler content?

When you say it like that… he continues:

Every generation looks back and says, “Things used to be better,” whether they are accurate or not.

But I’m not suggesting we live in the past. It is possible to learn from history and integrate the best of today’s technology with fair business practices that are more sustainable and healthy for everyone in the long run.

Benj has a few ideas on what we can do individually to push this ideal forward, but he believes it will take collective action to make meaningful changes:

Whether through purposeful reform or the eventual collapse of digital strip mining, I believe the personal computer will eventually rise again—–along with our chance to reclaim control of our digital lives.

Nick Fitzgerald: That’s it, right? So what compilers do is within the memory they’ll lay out where structures are, the same way they would lay it out in native memory. But if I’m a Rust program and I wanna talk to a C program, C doesn’t really understand Rust data structures. Rust can kind of talk at the C level, but it’s a little bit painful… And now we’re talking about interacting with JavaScript, but [unintelligible 00:30:22.15] but JavaScript running inside SpiderMonkey, a JavaScript engine that’s on WASM… So the structures are just totally foreign, and communicating is really difficult. But this is where another WebAssembly proposal comes in called Interface Types. Interface types are basically a type grammar, sort of similar to Web IDL, where Web IDL defines “Here’s records, here’s different string types etc.” Interface types defines a similar type grammar, and then the idea is you’ll be able to communicate like this.

Interface Types isn’t quite an IDL, in that the plan is eventually you’ll be able to kind of have dynamic adapter functions that allow you – say you’re receiving a string over this Interfaces Types boundary that lets modules communicate… And Rust’s strings are represented as UTF-8; so it’ll wanna just lower that into a flat UTF-8 array. But JavaScript strings are much more complicated. In SpiderMonkey I think there are – I forget if it’s 7, 9 or 13, but it was one of those odd numbers like that. Different representations of a string. Maybe the coolest one is called a rope. A rope is basically a tree of little strings. The idea is if you concatenate two strings and your string representation is just a buffer, you kind of have to potentially grow a new buffer that’s the size of both of them, and then you have to copy the strings in, and it’s really expensive. That’s [unintelligible 00:31:57.27] operation.

[32:01] But if you have a rope, what you can do is you can just say – you know, it’s kind of a tree, so you have a node that’s just “I am the concatenation of this one string and this other string, and creating that is order one.” So it’s very cool, but it’s very complicated. But interface types kind of will eventually allow you to define your own ways to kind of lower the platonic ideal of a string down into a rope, or something like that. Kind of like arbitrary computation for translating these types on either side.

So that’s kind of like the furthest vision. But right now we are defining just what’s called a canonical ABI, which fixes the representation; you have to use a string buffer, or something like that. There’s one representation for each type.

So with just a canonical ABI it is kind of just like an IDL, but this it’s open to that next step once we ship the first phase… So this is gonna allow all of these modules to talk to each other. And each of these modules - what’s really key about interface types is that they’re kind of [unintelligible 00:33:04.14] So if you think about npm modules, when you use an npm module, it gets all the same permissions and capabilities that your application has. And this is a problem; we’ve seen these supply chain attacks, where some generic markdown library or something - I don’t think it’s actually happened with a markdown library, but… You know, I just do something very innocent, and then actually I’m reading your SSH keys from disk, and I’m sending them off to some server, or I’m mining Bitcoins, or whatever… And so it’s not great. We talked a bit earlier in the podcast about capabilities and how a WebAssembly module can’t do anything unless you explicitly give it something to do. So Interface Types kind of preserves that ability between different WebAssembly modules. It says “Just because I can read to the disk and I’m talking with you and I’m using your markdown library doesn’t mean you can talk to the disk. All you can do is take this markdown [unintelligible 00:34:06.00]

Jerod Santo: Very cool. Well, if you are not like Suz and lucky enough to get in on the beta yet, I know there’s lots of people still awaiting… Well, while you wait, you can hear from Kyle Daigle all about it. We had him on the Changelog a few backs; I’ll link that up - “GitHub Actions is the next big thing.” We go deep into that, and he even answers my question about Bitcoin mining… So they have a response to the leeches out there who want to have the free compute power. So listen to that in the meantime, and hopefully we’ll all get in that beta soon. Or hopefully it’ll come out of beta and we can all start using it.

I’m excited for the community-shared workflows to start flowing towards my repo, so I don’t have to write very much code, and I can get all the benefit. That’s just kind of the lazy, selfish guy that I am.

Alright, thanks, Suz. Chris, you’re up. How about some shout-outs from you?

Tom Occhino: Yeah, when Jordan and I introduced it at that conference, I remember it vividly. One of the things we wanted to focus on was the sort of syntax, and how this thing that was very familiar to us we thought would be also familiar to the rest of the world. It’s like “Look, your HTML, the way that you want to think about your components - you don’t have to use these nested JSON structures, and all this stuff. Look, it just [unintelligible 00:07:46.01] But the sort of conversation ended up becoming about this mixing of technologies. And at the time, it was like “No, no, no. The whole industry is moving towards a separation of concerns, and you guys are mashing them together.” And we’re like “Wait, wait, wait… This is one concern. It’s my component. It’s my UI.”

[00:08:05.22] And so I think the biggest thing we did wrong there was we didn’t actually start with the problems that we were trying to solve, we just started with our solution. So luckily, Pete Hunt, I think later the same year at JSConf EU, did a much better job of teeing up the problem. Here’s what we were experiencing - as our applications got more complicated, as our models got super-big, people didn’t want to make changes to them… We wanted a loose coupling between components, so that those components could be used in lots of places, but high cohesion between them. So how do you do that? And it’s with the seam between components, which is the props boundary, and all these other things. So he did a much better job of teeing up the problem. And then I think it was more well accepted. People were willing to kind of give it five minutes after that point. But I’m glad that there wasn’t like a stampeding herd of adoption. This is amazing.

At no point in React’s history have we ever said “Behold, this is the answer! We’ve finally done it!” It’s always been about sort of like “Here’s the problems we’re facing. Here’s one solution that worked pretty well for us. If you have the same problems, let’s chat. If not, no worries. There’s lots of good solutions out there.” But I’m glad that Pete came through and restored faith in the sort of engineering team that Jordan and I were attempting to represent. Because the team was great, and the tech was novel. Something interesting and important.

Gerhard Lazu: Okay, that sounds really fascinating. I will have a follow-up question, but I promise to trade. So this is what I’m thinking… Whenever you get a Kubernetes, you think it’s the same, and it mostly is, but not always. There’s differences between IaaSes. In some cases, you would use, for example, a persistent volume claim. But that persistent volume claim will be so different between IaaSes that maybe it is the wrong thing to pick. Initially, in our case, we couldn’t pick SSDs for our persistent volume claims, which made anything that was using them really slow, depending on reads and writes. Networking - that keeps coming up a lot. Networks are so different between IaaSes, and it becomes very obvious when you have the same Kubernetes version, but the behavior is so different. Everything is the same. So while it behaves the same most of the time, there are certain differences which are really difficult to work with in failure scenarios or in bursty scenarios, or stuff like that… So then what are you doing? You’re just feeling good about what you’re using, until stuff just breaks, and you say “Ah, I wish I had chosen something else. What am I doing…?”

But it’s not so black and white. It’s not like you can use it or not use it. There’s good things and there’s bad things in it. So what I’m thinking is, as you mentioned - is it just like the hammer, and we just go with the hammer and nothing else? What if we use Kubernetes as something else? What if, for example, we use a platform? Because if you have a CDN in front of your app - and this is like a monolithic app - could you be deploying to Kubernetes and to a platform? You know, sometimes the platform works, and it’s good enough. But then how do you manage certain things which in a platform are more difficult to configure, like for example - I keep coming back to mixing runtime concerns and configuration concerns for your infrastructure… But there is something to be said about having everything in one place, and being able to see it, and understanding how all the pieces fit together. Cognitively, there’s less overload.

So could you have this world where you mix and match different runtimes, and everything just works, and it’s more homogenous? It doesn’t really matter where it runs, and almost like the first one to serve wins? So if Kubernetes is fast, that’s the first request that wins, and that’s it. If you have another origin in your CDN, which is much faster, then that’s the one that wins. And that’s the end of it. But I think that sounds very complicated to get it working. How do you do application updates? How do you manage upgrades? Now you have 2-3 things to upgrade, maybe, if they’re not managed, and then you still have the database problem. You have to move it somewhere, and then you have to connect to that… So I’m not sure whether that’s better, but I’d be curious to find out what does that look like, and work like.

Nick Nisi: Yeah. And in development, it’s really – that’s one of the benefits of it, is that it’s really set up for being able to incrementally add and change things… Or at least it seems to have evolved that way, with things like Webpack, and being able to take any kind of file and make them all work together. You can build a build process out of that that will get your code into production with any format, and mixing and matching everything, so that you can slowly migrate over towards one format and refactor your code like that.

Then the same thing goes with introducing types. If you’re going to do that with TypeScript, you can incrementally start with one file and type that, and then make sure that it works, and then slowly add more types to more files as you go… But it’s all about getting the infrastructure in there to allow you to make those incremental changes, and then rigorously test along the way. That’s not something you have to think about much with newer projects, I think.

Rod Vagg: Well, for a start, how good a problem is that to have, that you have people lining up to contribute to your project? That’s a great problem to have, so get over yourself. But the best answer to that I think is if you look in practice at how this actually works. When you have a project that has those kinds of bottlenecks, particularly in the administrative areas, because the maintainers have their head in code or other areas where they think is more important, actually bringing on new people and giving them that sense of ownership and the sense of responsibility can actually take away a lot of those burdens. So in practice, when you - this happens almost universally - take on somebody new to a project and you say, “ You own this thing; this is yours as much as it’s mine and were are gonna work on this together, and you have permission to go and do all the things that a maintainer does as long as you do it in a collaborative way”, one of the first things that most people do is go through the issues list and the pull requests lists and start doing some pruning or some maintaining of those things.

That’s an obvious place where a lot of people can help in open source, and they know they can help because they’re doing it already often in their own projects. They can go and actually help with that interface with the community.

So very often I’ll have a project that I might be the only person doing it, and I say – it might be something trivial, but I have this contribution policy where if you show up with something non-trivial, then I’ll give you ownership of it; somebody might show up and I give them ownership… Almost immediately they’ll go through the issues list and start closing things or responding to things, just cleaning it up because they have this thing about what the project looks and feels like, so giving them that ownership is actually great for offloading some of that administrative burden.

[44:22] In terms of getting a flood of people contributing, as I said, I think that’s a great problem for open source project to have, and that was not something to be afraid of with Node. That was something we wanted to embrace, because that meant that Node was healthy and if we couldn’t unblock that, then it was going to be unhealthy.

But there is a fear there of the outsider coming in with different ideas to what you already have. As somebody who starts an open source project particularly, you have these ideas in your head about what this thing could be, where it’s going, and very often you’ll hold to those ideas pretty tightly. It’s a rare open source project that comes to life and has the idea that it’s finished and it’s ready to go. Mostly, these things come to life and we have these ideas about where it could go; it could evolve into this particular space and solve this particular problem set, and it would be this magical thing.

If you are unable to let go of that internalized vision, or if you are unable to actually communicate that in a way that people are actually able to get on board with, then it can be really scary, because you’re saying, “These other people that don’t share my inner monologue about this project, these other people are gonna come in and they’re gonna bring their own ideas about where this thing can go.”

That was one of the things I think we faced with Node, and it’s probably gonna be a challenge for a lot of existing projects. They have this idea, that this project is on a path towards this point in time and certain people, particularly people that are just interested in technical problem solving, the people who are thinking more architecturally and organizationally have this idea about what the steps to get there looks like and where it needs to go. Bringing in people from the outside that don’t have that is really scary, because it means that the project is probably not gonna go in that same place, unless you can do some pretty heavy communication and convincing work.

I think that’s the problem we faced with Node and I can see that same problem with some other projects that are considering, but not quite at the stage of accepting this kind of model.

Quincy Larson: Yeah. So you could lay off one of every five developers, maybe. One of every six, or something. I’m not sure exactly how the math shakes out. But yeah. And that’s basing it on the productivity increase… But we haven’t laid anybody off. We don’t believe that – and we’re using this stuff extensively, internally. There’s just been so much work to be done, we just shift people around. So I can talk about how we’re shifting the team around… But let me talk just a little bit about people who are on the job market. If you’re listening to this, if you’re on the job market, trying to get a job as a developer - the jobs are going to come back. Hang in there. There’s a very slow correction… It takes forever for these hiring cycles to happen. It takes forever for the Gartner hype cycle to ride the wave from the peak of inflated expectations to the trout of disillusionment, to the plateau of productivity. If you’re familiar with the Gartner hype cycle, it’s this phenomenon that pretty much every technology has gone through. And LLMs are going through that. And when people say AI, they’re talking about LLMs. That’s been the major breakthrough.

So I think that if you’re in a situation where you are trying to learn to code and you’re hoping to get a developer job, my advice would be don’t quit your day job. The same thing I’ve always been saying. Like, if you’re working at a Starbucks, or if you’ve got a job at some accounting consultancy… I worked as an accountant for a while; I did temp work, essentially, moving from company to company… Doing that sort of stuff, it sucked. Keep doing that work. Whatever’s paying the bills, whatever’s keeping food on the table, and keep paying down your debt, and doing all that stuff. But just plan long-term. Expect it to take a couple years.

If you look at the pre-2000 bubble, if you had basic HTML, CSS skills, if you knew how to run a web server, FTP some files or something like that, you could get a job as a web designer. And then in 2012-ish, I would say, the tools became so good that a lot of people could get jobs as WordPress developers, or doing basic Ruby on Rails work… That’s what my first job was, doing Rails dev on a small team, just maintaining a Rails codebase… And over time, it’s gotten a little bit harder, but the jobs were increasing, so we didn’t really think too much about it. It’s just “Oh, now we’ve got to learn React. Now we’ve got to learn about a whole lot of security considerations. We’ve got to think about accessibility…” There’s always been this layering of additional things you need to learn, and that’s not going to go away. It’s just going to get harder and harder in terms of the actual skills that you need to know to work as a software engineer.

But at the same time, the number of openings started to fall… So now there’s this bigger gap between what needs to be learned, and the rewards or the likeliness of being able to find a job. And I do think that if you just continue applying for jobs patiently as you continue to build your skills, your network, your reputation, then you will eventually get a developer job, but it may take a little bit longer now than it took in 2022. That is kind of my thinking on that.

[00:33:58.21] So my advice to people who are on the job market is just be patient and keep learning, and don’t give up. Because if you give up, you’re never going to get a developer job. But if you keep at it, you will eventually be able to build out, through your community, through people you know, through different social groups you’re a part of… Hopefully, you’re going out and putting yourself out there and trying to meet people, through building projects and publishing stuff on Twitter, Reddit, wherever you share your stuff… Eventually, you’re going to have a decent portfolio. Maybe you’re going to build a big, impressive app that ties a bunch of stuff together, and is impressive from an engineering standpoint, and you’re going to be able to get a job. It’s just going to be a grind.

Jimmy Miller: But the basic idea for that one, just to be clear, is salespeople – obviously, there’s the real accounting. I know there were some comments on Hacker News of like “That just sounds like fraud.” There’s like the real accounting system, and then there was the rewards system, where they’d get bonuses and stuff… This was the rewards system. And yeah, they were trying to get bonuses every month. And if they made a big sale at the end of the month, and they’d already gotten their max bonus, they would just move that to the next month, so they could get their bonus for next month. So they had a cap on commissions, and things like that, and this was moving it around…

Jerod Santo: The great npm garbage patch

The Phylum Research Team:

Like the island of discarded plastic twice the size of Texas floating in the North Pacific Ocean, npm has accrued an astonishing amount of spam packages over the past six months…

Our 95% confidence interval for the estimate of Tea protocol spam in new packages over the past six months jumped to between 68.66% and 74.67%, or somewhere between 613,000 and 667,000 packages. In other words, among all new packages published to npm in the past six months, about five out of every seven packages are Tea spam.

I first covered the unintended consequence of the Tea Protocol’s crypto rewards back in February (issue #83). It appears the damage is even worse than previously discovered! What a mess…

Joseph Jacks: Yeah, because there’s quite a few clauses in the LLaMA license that they can prosecute people on. But they just relaxed the license terms. I would not be surprised if they just made the whole thing like full Apache, or full some traditional, conventional open source license in the next couple of releases. I would not be surprised. The reason is - and Zuck literally just said this on Bloomberg yesterday. I believe Emily Chang interviewed him. He said they have no need to monetize these models. They have a very strong advertising business, and they’re not selling, they’re not licensing software… This is not a core business-differentiating product, and they can give it away completely. And as a result of giving it away, they reap the same rewards as they did open sourcing PyTorch, and open sourcing Open Compute, which is indirectly saving them - actually, pretty directly - many billions of dollars on building out their compute infrastructure for serving applications that support billions of people, which is what they do.

And so it’s actually quite a beautiful, very aligned company to be embracing and investing so much in open source and open platforms. And you kind of think, like, “Why isn’t Google doing the same thing?” They also print money with advertising. Why isn’t X doing the same thing? I mean, Elon - maybe he doesn’t quite as much print money with advertising; he’s lost a lot of advertisers. But you know, the point still stands. If your business model is not in diametric opposition to your innovation model, you should be innovating like bananas. Like crazy. And your business model should be reasonably sound, and protected, and you have different aspects that compound the value of an advertising business, which is collecting and generating more user data, which is what they do… And unfortunately, that’s the trade. We use these free tools for free, we don’t have to pay for them, and we happen to part ways with our very precious personal data. That kind of sucks. I’m actually very anti that. I deleted my Facebook account many years ago. But it is a very aligned kind of company in terms of acknowledging “Look, we can invent and create crazy powerful technology, give it away, no problem, because our business model doesn’t need us to charge money for that technology, even if it’s costing us billions or tens of billions of dollars to produce”, because we get rewards and benefits elsewhere, that show up elsewhere on the P&L… Which is like compute efficiency, improvement, talent acquisition, whatever.

So I think Meta is a really great example of – I think, frankly, a pretty incredible steward of this open source kind of philosophy in the last few years. I mean, it’s been pretty surprising, sort of rebranding almost from the previous years of lots of drama and blow-ups that Mark Zuckerberg has had to deal with.

Birk Jernström: So that’s certainly a helpful tactic, but you don’t have to deploy it. And I think this goes back, again, to – it’s just a key principle of ours… Like, this is not issue bounties for open source. I’m very much against them. We’ve designed issue funding and rewards with maintainers in mind, and putting them in the driver’s seat. So some maintainers, they don’t want a funding goal, because they don’t want to necessarily say that “By X amount I promise to do it.” They want to leave it open-ended, and so we give them that complete control and flexibility.

Jessica Sachs: So the one FAQ on the Kickstarter that makes me curious of the structure of the company is “What if I’m already a Font Awesome pro subscriber?” That FAQ. And it says “If you’re already a Font Awesome Pro subscriber, and if your rewards include a year of Font Awesome Pro, your subscription can be extended by a year.” Does that mean that the whole model of like every company that’s currently using Font Awesome is going to be affected by this awesome – oh God, I used “awesome” as the adjective…

Robert Ross: You know, make a rewards system for our most prolific incident commanders.

Karan Malhotra: I think that definitely just like straight up instruction tuning is great. There’s other ways to tune, like the eval instruct method. I would advise people to try to create new instruction methodologies, that allow us to make even better formatted data. People don’t spend enough time trying to create new instruct formats. And we’ve definitely been swamped with not doing that as well. So I think towards the general community, it’s a really easy place to get started; you don’t need to really know how to code, so much as think about how a human might more effectively phrase something, or format something, and kind of remix from there. I think that’s probably the easiest place to start.

Then there’s model merging. Model merging is great. You can just like take two models and Frankenstein them together to question-mark results. You’ve got to just try and see what happens, and feel it out. Then from there, I would say there’s stuff like DPO, there’s RLHF… DPO kind of rewards things; that can let you enable rejections, or create censorship, or put some kind of general concept or attitude towards a model. We’ve found that to be pretty effective with the latest News Hermes Mistral DPO. It seems like people really like it and prefer it over just SFT. So that’s another thing that I’d heavily recommend.

From there, we get a little more complex. We have some reward model stuff we’re working on that I won’t speak to just yet, outside of saying we’re working on it, that we think is going to be like pretty big for reasoning boosts. Of course, there’s techniques like chain of thought, and tree of thought, for like multi-step prompting. Creating datasets even out of that for any of these purposes I’ve already mentioned is going to be really effective.

Now, to stuff that maybe not everybody can – actually, a lot of people would already be able to do this. There’s something that we like to call over at Nous activations hacking, where you’re kind of messing with the way that a model – I’m trying to think about how to say this in like the most layman’s terms… You’re trying to mess with how a model generally vibes about something. [laughs] So rather than just doing a system prompt or something like that, you can actually change the model vectors to kind of be like more political about something, less political about something, more terse, or more specific… And it has far more effect and control over a model than a system prompt. It’s basically like a system prompt that tells it to embody certain characteristics, but it’s not something you can really jailbreak or get around, as far as my testing has shown. Certainly not as easily as a system prompt. We have no problem jailbreaking even the most censored closed models today. It can be done by anybody, with the right words. But this activation stuff, it really creates a bit more of a robustness and fidelity to the concepts that you’re trying to tell it to embody.

There’s a few more I’m trying to think of that would be useful for people… One thing is soft prompting. It’s not really around anymore. It used to be pretty big during the GPT-J, like pre-LLaMA days, and the Cobalt AI guys really pioneered the use of it in the open source community. But a soft prompt basically takes like a massive prompt and it compresses it down to like way less tokens. So you can give your model like a huge prompt, a huge system prompt, or a huge amount of information, and use like way less tokens. So soft prompting is cool. It’s not going to be too difficult to update it for like LLaMA, Mistral, today’s architectures. Nobody’s really done it that I’ve seen. So to the community, if you guys do that, please share. [laughs] That’s actually much easier than the activation stuff, I think.

And then finally, probably the hardest unsolved is sampling methods. Today we use like Top-K, Top-P, [unintelligible 00:39:52.10] sampling etc, whatever. There’s better ways to pick tokens, for sure. There’s better ways to judge the value of tokens, for sure. Everyone has been too concerned with higher levels to get that low, and do whatever the magic math is that I can’t do, that would enable some steering, and some – even beyond steering, like alternative sampling paradigms. And I think that would probably bring the biggest change and transformation to literally all models, regardless of the tune, regardless of the architecture etc. if it gets pulled off. So I’m really looking forward to something like that happening in the space.

Adam Stacoviak: [01:15:49.00] The most innovative content happening out there, in my opinion, the proving ground is happening in virality, as well as consumability on TikTok. Like, there is a lot of remix… I just think it’s a proving ground for a lot of great things. It’s not the only ground, but I think there is a lot of people on that platform, and it’s rewarding to those who can find ways to present certain things in a way that people want to consume them. I think the algorithm rewards you well as well. I think a hub and spoke, multifaceted approach is the right way; maybe long-form or clips on YouTube, maybe very personal things on TikTok… I mean, a newsletter is still very great… I think you’ll see some dividends be paid to you, if you – even in this case here, you’re exposing your idea to a larger audience. Like, do this again, but on a whole different podcast.

Jerod Santo: Sorry, digital rewards just don’t get us.

Haroon Meer: [01:34:17.27] Yeah. And again, I’m pretty convinced it’s also the path to winning. I’m pretty convinced… And again, we’ve been lucky so far, but I think the market rewards that stuff. You end up making a good product, and the market hopefully rewards it. [unintelligible 01:34:35.01] we just did the Black Hat conference, where we had a booth… And one of the things I was talking to someone about which was super-interesting is when we do a booth - like, we’ve got this really long blog post out on doing booths and why we think it’s actually good for people. Young me hates it, but booths are super-good for us. We do a booth at RSA, we do a booth at Black Hat, we get to meet all our customers, they come by and chat to us… People we’ve never seen before come and say nice things about our products, other people hear them… But at Black Hat this year something that occurred to me is we have this booth, and I’m there for the full two days… And Marco who’s our CTO is there, and Bradley, who’s one of our other founders… So literally, all of our original founders are there, plus some of our engineers. And so for two days, people are rolling up to us, talking to us, “Hey, I’ve been using you for six years. Hey, I did this”, but that’s surprisingly rare on the showroom floor. Because on the showroom floor, what lots of people have done is they’ve paid a whole bunch of young interns, a whole bunch of college students to say “Scan as many badges as you can. You scan the badge, you then get to spam all these people, trying to sell stuff”, and again, it’s horribly mixed incentives.

For us, the thing is we get to meet our customers, and we get to do demos with new people who might be interested in the product. And it’s so counterintuitive, because if you talk to – with any VC, one of the playbooks that they will tell you is the truth is not in your building. Go out and meet customers. If I told you “You’re on the showroom floor, and you’re going to meet 20,000 of your customers in two days”, why wouldn’t that advice mean that every CEO, CTO and chief product officer is the person on that floor? You’re gonna meet 20,000 of your customers or potential customers; you can talk to them about the product. But it’s just not done. Because fundamentally, what the execs are doing is they’re sitting in a suite somewhere, trying to arrange their next raise, or trying to talk to analysts, or trying to talk to the media. And again, the state of the products in our industry are a reflection of that; mostly, we build terrible products, because people just don’t care enough about them.

Cory Doctorow: So that’s chickenization. And if chickenization sounds familiar, it’s what Uber drivers have, right? It’s what DoorDash has. It’s what – there’s a company called Arise. They hire primarily black women to do work from home call center stuff, and each of these people are misclassified as contractors. They have to start an LLC. And then because they’re a company, they have to bid on the job. They work for like Carnival cruises, and Disney, and whatever - they have to pay to be trained to be the phone operator. And then they’re listened in on constantly, and if their kids cry in the background, they are struck off; they lose the job, but they still owe the money for the training. And there’s a penalty for cancellation on their side, where if they quit their jobs, they have to pay their boss to quit their jobs. So that’s a form of chickenization as well.

Chickenization is kind of working its way up the privilege gradient from the most blue collar workers now to a more middle-class group of workers. It’s your boss’s ideal, right? All the risk is pushed on to you, all the rewards go to them. Every business owner would love to have that as their arrangement, right?

[24:18] So what’s a centaur? Well, a centaur in AI research is someone who’s machine-assisted, right? I’m a chess grandmaster, you’ve got a chess playing software program. Neither of us are as good on our own as we are together. We can beat opponents together that we couldn’t beat on our own. So that’s a centaur. But a reverse center is when the machine is in charge, and you’re at support. So say you’re an Amazon warehouse worker who’s wearing haptic gloves that go “Bzz! You need to pick that up. Bzz! You need to turn your body. Bzz! You need to drop this thing in the box. Bzz! Bzz! Bzz!”

Or say you’re an Amazon driver, right? You’ve got a facial recognition camera that’s watching your eye movements and penalizing you if you look away from the road, you’ve got a clock that’s timing your drives, that’s telling you you have to turn left, even if turning left would result in your death, and then it docks you for doing it. Machine says no - that’s a reverse centaur, right? It’s when you are the disposable meat sack for the all-important algorithm.

Fred K. Schott: I would call out – I think the interesting part of React and Next sort of like merging a little bit right now is Next.js I don’t think is… That’s a strong project, that people are very happy with, and that’s starting to become – it’s gone from like “Wow, they’re doing a great job” to like “This is fairly abnormal.” I think it speaks to the fact that React is so big, and there’s not really a competitor to Next. You can think of SvelteKit and Nuxt on the Vue side, but everyone’s so framework-specific… No one really – Remix is I think the first time I’ve seen a second framework for an ecosystem come along and really challenge it… So I really wonder how much of that is like there just aren’t really other alternatives in the React ecosystem, and the fact that React is so big, it just kind of like by default is like the top of that food pyramid… But at the same time, it also speaks to just people are happy; like, they’re not really looking around for other things… Maybe because they don’t have other things in front of them, but also it’d be very different if it’s like “I use it, but I hate it.” If interest was low, there’d be a real risk there. But I think it speaks to the strength of that team continuing to innovate. There are huge risks to continuing to change things at this late stage, but there are big rewards to staying top of mind, and keeping your number one spot as the most popular – you know, you’re not the newcomer anymore.