EP07: Open Source AI vs ChatGPT, What VCs Like in 2024, What Makes a Great CTO?
James: Welcome to the Galileo Venture Podcast of 2024. I'm James Alexander, joined by my co host Hugh. This is your go to pod dishing out the latest unfiltered news and views on early stage tech served up by two seed VCs who don't have time to sugarcoat things, whether you're a founder or an investor, we've got the operator insights that just might make you say aha, or at least ha interesting. On today's menu, we cover our new exclusive program for ambitious software engineers, which we have quietly launched, but we haven't really launched, but we're going to give a preview today.
We're going to talk about open source AI to rival chat GPT. In particular, a new product launch that just literally happened this week as we're recording. And we're going to talk about what VCs are interested in 2024. How does it sound to you here? Hugh: Very exciting.
Bonus question did you actually get ChatGPT to write that introduction? James: It's funny you say that. I may have gotten ChatGPT to slightly edit the introduction. Hugh: see? It's there's that interesting turn of phrase that you think to yourself I think that's generative. James: I also just don't think it's like in my usual style and you know me quite well, because I wouldn't come up with a witty script like that.
Hugh: It would certainly take you too long. James: That's right. In fact, the more and more I use ChatGPT, the more and more I realize how it can be really helpful. In summarizing things, which is still the best thing it can do. But ultimately though, you end up with this weird persona sometimes if you like adopt too much of his text, where this it's this it's like averaging out all the humans into a particular way of talking.
Hugh: Ultimately is. Effectively, how it's designed, right? James: yeah which is great, which is just means we're going to go for more homogenous culture. That's exactly what we want in this day and age. Hugh: This was the that is the interesting pieces is what does that mean for, writing online and writing on the internet? And, certainly given I obviously spend a lot of time in the marketing space. The big question is, if everyone's using generative AI, even trained on your own data, ultimately, how does a unique voice come through James: And I think that's a very good question that needs to be answered. In fact, if some people may remember when ChatGPT first launched, Sam Altman himself was saying he would love a situation where people had their own tuned LLM that wasn't just some tune using global weights.
And obviously we're not at a place where we can do that yet, but it is an intriguing idea to think about. That we all might go around with our own personally chewed LLM. The Hugh LLM.
What would the Hugh LLM be very good at doing? Hugh: very much? James: Not very much. But first, a couple of weeks ago, a journalist reported that the Australian government planned to lift the investor threshold test. Of course, this set off a wildfire of VCs and angel investors tweeting and posting on LinkedIn about how this is going to destroy our industry. But, quote from the AFR Weekend., AFR Weekend understands that
the government plans to lift the thresholds required to qualify as a sophisticated investor from about 2. 5 million in assets. A level that now admits 16 percent of Australians through no final decision has been made on the exact level. sO essentially what that means is very few Australians could then invest in venture capital funds and certain private asset classes and angel invest. However, last week, Capital Brief, the new upstart tech media publication, did a little bit of digging and they discovered that, in fact, there is no current proposal from Treasury to government, that they'll increase the threshold percentage.
. Now, the interesting thing is, the threshold for the sophisticated investor test, which is a net asset test or an income test, has not been raised in Australia, has not been changed for the last 20 years. Interestingly in the UK, they've just changed it and they're introducing it as of Jan. And of course, similarly to here, there's a huge petition going around the UK right now to basically put some caveats in that. My question to you is Hugh, does the sophisticated investor test still make sense in this day and age? And should we raise the threshold? Hugh: Yeah. Um, It's a challenging question. I think it makes sense.
Like we've got to go back and I think sometimes the setup sector forgets that the role of these regulatory regimes. is ultimately to protect everyday consumers, right? And I'm going to include a mom and dad with a lot of money can be still a everyday consumer. And certainly in terms of just because you're sophisticated in one aspect doesn't necessarily mean you're in others.
And, we know that the reason that these things that these regulatory regimes exist. Is to do that, right? And the question I think for policymakers is really what is the right policy lever? If you imagine a balance between on one side, rampant fraud and on the other side, no innovation, where is the right place to actually put, the dial? Where should we how? How much sort of fraud or risk of fraud, do we accept? Versus on the other side, if we accept too much fraud, we'll have sorry if we accept too little fraud, right? We'll have no innovation. And I think that's where, there's been this challenge.
And I think something that the startup sector forgets to or probably inadequately is able to talk about is that piece of well, how much fraud is okay. And as we talk about having the opportunity for more people to invest in, early stage startups or private, the type of things that, you know, requiring these kinds of sophisticated investor rules we need to go well at the same time. Yeah.
We're going to accept more fraud. And that is going to have a flow on effect. And what should we do about that? And who should be exposed to that risk? And who shouldn't? So does it make sense? I think absolutely. I think there's some degree of it.
It does. And it, it should make sense whether there should be changes to eligibility criteria for exemptions for people that have done in the U. S.
For example, you can do a certain course, for example, and so I guess qualify. Okay. Bye. And I think, in some ways, the Matt Levine joke would be to say that you could run a course that just says, I will lose all of my money, and if I do lose all my money, I will not complain about it, sign on the dotted line, course complete. But maybe there's an option for an exemption like that or something like that. But I think we as a sector need to also recognize that the government has a valid reason to have these rules and I think raising them to make it impossible is going to have a flaw in effect to obviously our sector, but also other sectors, there are small hedge funds and things like that, but we have this, but there are other exemptions anyway, like the two, the 20 and 12 rule that allow non sophisticated or retail investors non wholesale investors, maybe to already invest.
There's obviously equity crowdfunding legislation, we do have some Existing exemptions for these rules. And maybe it's an opportunity to go actually, let's talk about crafting a better regime, regulatory regime that makes sense that tries to draw this balance of fraud versus innovation. We know that just letting it go, too far into letting everyone do everything just really dials you into that fraud. And we saw that when we saw that with cryptocurrency, right? Yes, for all of the interesting projects in crypto, there were also a whole series of rug pulls and also stuff that some might say, Oh that's a startup. Others would say that's a scam.
James: cRypto is an interesting example of a totally unregulated new asset coming into play, right? Which went after investors. But ultimately, you could argue that the negative impacts for Australian mum and dad investors was not as bad as one might have predicted. Hugh: Yeah. But I mean, could you, could you also say that's because of the relatively quick regulation that was put in place, like we had fairly fast like sort of regulation in terms of regulating on ramps, the bank's got quite you can't put, you can't buy cryptocurrency with a big four bank debit card. And the reason you can't do that is because they saw too much fraud. And for commercial reasons, beyond purely the regulatory side, for commercial reasons, the banks pulled the pin on that.
And that obviously, I imagine, has substantially dampened people's interest and capacity to invest in cryptocurrency. But is that the reason that we saw that degree of protection? But, there were absolutely all the articles about, so and I don't know, small business owner James: yeah, Hugh: runs a cafe, blah blah, lost all James: There's always going to be some examples of people losing all their money by doing a silly investment. However, I think I take a more extreme view than you and I'll explain. So I, my, my view is I agree with you and I think this is sound that we probably need to rejig the entire set of exemptions and rules for what unsophisticated investors can invest in. I don't believe an income level makes sense.
I don't believe a net asset test makes sense. I think it's just a raw instrument. I think a test Hugh: What makes sense then? I think that the hard part is, those are both easily quantifiable. James: Correct.
I so let me explain. So I think a test is nice, but I don't know how impactful that would be. It's basically like a disclaimer, right? It's like you have to do this to understand that this is high risk. my view is I think we should go to the very heart heavy, unregulated end of the spectrum, but ultimately, I think we should actually swing wildly that way as a small country with some safeguards in place. And put some, maybe you put some threshold limits on, percentage of revenue of percentage of earnings as to how much you can put into this and still get tax benefits, which is what they've done with Essex slightly, so for easy, if you're not a sophisticated, so they've already got a version of this. I think we should go wildly into that way.
Knowing two things. One, Australia's love to bet. You just have to look at the Melbourne Cup. And two, we seem to be okay with this when it comes to gambling.
Hugh: Yeah, look, you're not going to hear a single argument from me saying that, why do we allow people to go to the James: And quite, quite frankly, where I go morally when we actually intervene is when we are losing more money on startups than we are losing on pokies in New South Wales on a yearly basis. Hugh: yeah, I think there's a, James: And that's, Hugh: while I'm not, while I'm, James: Billions and Hugh: oh yeah, but while I'm far from the person who would say should be defending the pokies of all things, I think the difference for me in some ways is that element where. In the context of the pokies for some people, that is a form of entertainment, James: Correct. Hugh: right? And, does that mean we should also, overly restrict cinemas and, like other ways that people just burn time? And is it a good way for pensioners to burn time and spend all their pension? Absolutely not.
Do I support the gambling sector broadly? Not really. Absolutely not. But I think investing in startups is not exactly entirely akin to that. James: no, of course not.
And Hugh: but I do accept the point of saying hey, people take risk in buying Powerball tickets and, all that sort of stuff. And why is that any different? James: I think my question is more of a moral judgment as to this and why we could perhaps actually handle a lot more risk. I, and the other thing we see, as a fund, we see the demand, like we see a lot of unsophisticated investors that would love to be able to invest a small amount of their savings into a high risk asset like us. But they just simply can't. There's really, it's really complicated for them to do that.
And there's a philosophy here which I find really intriguing without spending too much time on this, it's around why can't middle income earners access assets that rich people can access to invest in. Hugh: yeah, I guess I, I think, it is an interesting point, right? And I think part of that is, though, that the arguments of that are that, generally speaking we talk about gen sorry, Generally speaking, It's not necessarily that we're restricting people from being able to do that. But we're saying, Hey, you need to have a certain threshold before you access these high risk high return investments. And that isn't necessarily saying all the rich people can get rich because they have access to these and all the poor people can't. Because I think the other piece is the rich people also lose money on a lot of these things.
And yeah, occasionally do make money on them, but they have enough money to be able to have a portfolio approach to the problem. And actually Manage that risk return appropriately in a way that you know what? What we wouldn't want to do is to go. Oh let's make all these asset classes. It's not just receipts of other assets as well.
But let's not make let's just make all these asset classes much more openly available and then see. Taxi drivers yellowing their entire savings into V. C. James: Hint to Hugh: I think and again, look, I don't think that people in the setup sector are advocating for that.
But I think the challenge here is that if you set the policy constraints wrong, that is the outcome you're going to see. And that's the outcome that the regulators precisely don't want. James: Yeah, Hugh: They're not wanting to try and stop the startup executive on 200k putting five grand into a startup. They're wanting to stop the, person who's on, I don't know, 60, putting their entire life savings into it in the hope that it will return, a house deposit for them in 10 James: Sure yeah.
Yeah, food for thought. Moving on, we will move on to some global news in particular, I thought it would be really interesting to chat about the open source AI battle, which is heating up. Hugging Face just announced they're launching free chat assistance, very similar to OpenAI's ChatGPT.
And quote, there's a new free product offering users of Hugging Chat, the startup's open source alternative to OpenAI's ChatGPT, to easily create their customized AI chatbots with specific capabilities, similar both in functionality and intention to OpenAI's custom GPT builder. For those of you that don't know, OpenAI released a marketplace or app store of sorts recently. Which allows you to integrate various things. One of them I used today for our podcast editing, as Hugh very quickly picked up. Now, the main advantages of Hugging Face's GPT model versus OpenAI's GPT model is that you can choose your own bottle.
So that means you could use, any bottle, not just OpenAI. It's completely free, as opposed to 25 bucks a month. Which is about the average for the other paid models and it's publicly shareable.
I think this is really interesting because as we've already seen, this is the direction the market's been heading and our investment in relevance AI is already in this same direction, right? You could create your own assisted. Choose your own model some of its free usage and then it's publicly shareable already So for those you haven't played around, you know plug for relevance to go check it out because it's very similar vibe but obviously code no code Anyway, there's still a lot of feature parity issues. It's not the same as chat GPT etc And you know what? I actually quite skeptical about the chat GPT marketplace right now We'll see what happens. I haven't found it as useful as people are saying it is, but let's see the play out of the coming months. I find this interesting for two points.
Number first point is that open source does seem to be a serious contender to closed companies like OpenAI, ironically named. Now, in the 90s, from memory serves me correctly, and from what I've said. Hugh: From memory says, even though you were a child at this point in time.
James: James, the VC child, Hugh: From what I understand of history, James: my understanding of history and from the people that were around then is that really in the nineties, right? You, VCs were not sure about open source and they weren't funding open source companies. days, that's not true. Just look at Minstrel, right? The open source contender, it's released the language model. It's going very well.
It's actually quite powerful. And from what I've been told, although I haven't tried it yet, it's extremely powerful and very impressive given that it's a very small model relative to GPT 4. But open source seems to be a serious contender. Do you think that's true? And do you think the hugging phase to the announcement is interesting from your perspective as an active operator here? Hugh: So I think the interesting part for me is more that the question of how much we're really going to see people multi homing on models. And I think the opportunity delivered by someone like Hugging Face is obviously the, is making it easy to be able to use models of, best in case models for best, for whatever purposes you're looking to do. And that's a really interesting opportunity.
And I think, we do have a series of open source models as well. There's the Lama model developed by meta for instance those type of models and no doubt there'll be more over time whether they're driven by the community, driven by, large companies as that one obviously is I think there'll be more as we go on. And there's a question of whether or not I guess the primacy, of the model of closed providers versus open. You But the other element is really You know, what's going to be that orchestration layer? And historically in the past, that orchestration layer has increasingly gone into being, like an open source type model. And you look at databases is probably a good example where, my example in your case, James would be in the nineties. Most many of not many, not all, but many of those databases were closed source.
And, you were getting your S. A. P. Or your Oracle database and things like that. And that was your core database. Okay.
Thanks. Whereas now, Amazon, for example, did a massive migration into, their Aurora Postgres equivalent or Postgres compatible database of a lot of their internal services to get off what was presumably many billions of dollars, probably of Oracle database spend and that pushed forward some really good opportunities in that open source community because of the investment in, by a commercial operation and lots of stuff. And they obviously then also been able to monetize that into, an AWS service. So I think it is now a much more accepted business model, and that makes life a lot easier.
And I think that enables a lot of these opportunities. And I think, as you've put out I think there have been a series of examples where VCs have successfully been able to invest in the asset in, open source as a business model and get the necessary returns. And that's been able to obviously fund the development of these kinds of things in a way that hasn't been possible in the past. James: absolutely, side note. I'm actually very surprised. There's no Open source Australian based initiative in a foundational AI model.
I know you don't agree with me, but I think given the amount of AI researchers here, I would have thought something would have popped up, but maybe they're heavily involved and no one thinks of it as, one geo in the same sense that we think of the other models is. I have a follow up question though. So I broadly agree with you. It's hard to understand where the suitability of each of these sub model sizes is going to land. Or do we just have the big generalized model where you give it some Instruction to be like this, and then it's just as good as a hyper specific model.
I think that's probably unlikely in the sort of paradigm of technology, right? Because you have these like compute cases on, let's say, a hardware device or an IOT device where you can't simply be running and waiting for the model to process. And as we know, there is a very strong lag time in GPT 4. In fact, I think Sam Altman just tweeted about how they've just improved it over the new year because it's been lagging a lot.
Various reasons, theories as to why that is the case. My favorite one being that it's just lazy in the new year because apparently, heh. People are lazy in the new year according to the corpus of texts they have. Heh but I think I think I agree with you to the point where I think it's a serious contender. Who knows where this is going to land, but I love the fact that there are more and more optionality for consumers to use tools like this. And I really wonder how AI regulation may curtail some of these efforts if it becomes as extreme as some people think it should become with LLMs.
Which is going to be an issue. My second question for you is How long do we think OpenAI's first mover, Moat, will last? Now if I was to ask anyone at OpenAI, and I have asked them this They go, we have more money, we're further ahead Every time an open source model gets released, we've already got something that's way better You know, what would your response be to that Knowing that a lot of founders and early stage investors are listening to this? Hugh: Yeah, I think it's an interesting challenge. I think the big challenge is that I and through, let's be honest, through chat GPT, not through really a lot else through chat GPT. I've now has been very successful in I guess building that consumer brand. And I think if you look at a lot of the like I'm a member of various business communities, let's call them.
But if you look at the conversations that a lot of those people are having, it's all about chat GPT. And they know who OpenAI, they have no idea who Anthropic is. They have no idea any of the other, major players in that context. They might know Bard, because obviously, again, Microsoft, has a lot of money to deploy in advertising and the media reports a lot on it. But they all know OpenAI. They all know that OpenAI is ChatGPT.
And I think That advantage will be quite substantial as long as they can just keep up. And I think comfortably with the amount of both capital and, very smart people and everything else they have within the team, I can imagine they're not going to have too hard a time just keeping up. And so I think that first, that moat will last for quite a decent amount of time. But I think a lot of that is due to that brand piece more so than necessarily any degree of quality. The two come obviously a little bit hand in hand.
But I think the way that people think OpenAI immediately when they think about AI will be a big advantage for quite a period of time. James: I agree. I think that's just spot on. I think the brand piece is very much I'm not sure they overhype their internal tech capability a lot. I'm not sure how much I attribute their success to Oh, look at our amazing engineers versus, Oh, we have a lot of compute, if that makes sense. And so my view would be to this, I think there's two things that will happen for the scale to change.
One is compute. OpenAI clearly have the most GPUs of all of them. And as long as they keep that lead and as long as this architecture remains this inefficient in regard to how LLMs work and requiring GPUs to actually process each token or each query. I think they're going to have that advantage. So if suddenly next year we have a hyper efficient open source model that changes that, I think that's going to go away. And I think they're going to have a lot of issues.
The second thing that I think, that's my first view. So it's like compute argument, I'll say. Now, I think in the history of tech, typically to early technologies, compute stays there for quite a while. So I think that's back to your, I think they're going to have that moat a little bit.
And then the second part will be functional LLMs. So really specific LLMs where it makes way more sense to use an open source one and train it on private data. For example, I'm a pharmaceutical company. I have private. Human trial results. It doesn't make sense for me to use a big open source model now.
And I think those functional use cases, we're still working out what they are, but there's going to be a set of them. And it just suddenly becomes this thing where there's going to be these other players there. That's my prediction. Hugh: we'll see.
so moving on to the next section. James, you were recently featured by the Australian in a, in an article talking about various VCs predictions for 2024, the classic what's going to happen this year piece that everyone publishes seemingly towards the start of every year. wHat were your thoughts? Like what did you share? James: It's a funny little article because we ended up getting included right at the end. Hugh: It's all right. Sometimes nice guys do finish last. James: Correct.
Yes. And I think, we are probably the smallest fund next door to a square peg and blackbird which manage a little bit more money, Hey, what do they know? So look, I think my view was that there's a bit of a sea change. So actually what I said in the interview was that from 2020 until 2022 was the era of SAS. I don't know if you agree with that or not.
Hugh: Blockchain people might disagree with you on that one. James: Absolutely. But I don't think they can conclusively say it was the era of blockchain anymore. Because as we can see, none of that really came to fruition as fast as they thought it would Hugh: you could say it was the era of blockchain. It's just that era is over.
James: Maybe that's the case. But from a mainstream VC perspective, right? So I exclude the specialty blockchain funds that all popped up really quickly. That decade of vast enterprise and wealth was created in majority of software, online software companies serving enterprise, you could argue, and. I think as a result of that now with what we're going into, I think we're in a totally new era. And so I, one of the comments I made on there was that, software is not as compelling to VCs. And of course in Australia, that's like sacrilegious.
You can't say enterprise software is not compelling because and I got some flack for this, just, to keep it by. anD I think I still hold this true. And what I mean by that is I think VCs getting interested in a lot more other assets than simply enterprise SaaS.
And I think we're seeing return profiles that are very similar and people might forget. But when VCs first started as an industry, it was not enterprise software that they were investing in. It was things like offshore oil rigs. That was literally one of the biggest returners for some of the first venture funds alongside microchips, right? Which was the first wave.
So it had nothing to do with software. And I think we're going into that sector now, whether that reality now for VCs. And I think the reason for that is that AI everywhere, so being able to Use a generalized model in any sector to do things unlocks a lot of software like opportunities. I made three points in a follow up post, which is around which I'll link in the show notes if people are interested, but one, I think there's a lot of industries where previously enterprise software couldn't enter, but now can, and that therefore the market is a lot bigger now for software enabled by, I'll call it enabled by AI, , two, convergence. There seems to be quite a few different things happening, which is converging in particular with the chips.
And mobile chips and GPU chips, that, people forget that GPUs have really enabled the current LLM revolution we're seeing. And that was because of games. That's got nothing to do with AR, and so like that is a very big mega trend that's been happening in the background, maybe it doesn't get as much mentioning. And then the third thing, and this is a bit more again, this is a little bit more of my crazy, pontificating here, which is software is becoming at least basic software is becoming more like a commodity, and as a result of that.
It's becoming easier to make, software, small, simple apps and things like that. We see that in our pitches with our startups when they come to us with MVPs, which are made with low code tools, which is impressive to see the traction, but obviously doesn't, you know, you still need a strong team and strong IP and things like that. So I think that's what I meant. I don't know. Do you have any views, Hugh? The Oz did interview you.
What would you say would your prediction for 2034 be? Hugh: wHat would I say? Look, I think I think it's an interesting question. And certainly I agree that a lot of the certainly the current generation of, I'll say the current generation of venture capitalists in Australia have made a lot of the time, probably their own money, but also The money from their investments in things like sass and marketplaces and that might be different to historically. I would maybe disagree a little bit in saying that things like bio are probably more regularly the place that VCs were making money than necessarily, oil rigs. But. Putting that piece aside. Look, it's interesting to really see for a lot of, for a lot of VCs, particularly some of the newer cohort ourselves included, this is coming into the first time we're seeing a big downturn in the sector.
And what does that mean for our investment strategy, our thesis, our ability to raise funds as a fund? Obviously, that's an important part of that. But also, what does that mean for the deal flow? What are the businesses where historically we would say they produce the type of returns we're looking for? That maybe now in a different, different economy, different market those are no longer viable. And I think that's an interesting, thing we've got to really think about. It's James: Yeah, I agree. And these see how these play out over the coming year, right? It's hard to say, Hugh: very hard to say, I think, at James: Moving on.
I wanted to switch away from news to something that we've just launched this year. In fact, at the beginning of this year, we embarked on a very exciting journey by quietly launching a new educational initiative, specifically designed for ambitious software engineers. Essentially, we discovered, we realized after working with a lot of companies over the last decade between us, that tech companies, that there's a huge demand for world class engineering managers as well as a need for skilled engineers. And as our companies grow, you have engineers that move more into engineering leadership and lead engineer roles. And as you invest in these companies, but it's become a big challenge about hiring and training up these types of engineers that want to step into a more managerial style role. or leadership role.
If you like recognizing that gap, we actually joined forces with one of our engineering coaches. Sam Thoreau Good is a seasoned veteran with a strong background of Google over 12 years working there nearly 15 years, I believe, and an experience as a CTO in startups as well. Our conversation with Sam sheds light on what participants can expect from the new program. And just to keep in mind for anyone that's listening, this is a small first cohort we're running. We're limiting it to 10 engineers.
So it's a really intimate environment. And it's an in person experience in Sydney. For now, we will take it to Melbourne and some of the other cities as we progress and learn spaces are limited. So if you're interested, go to Galileo.
ventures forward slash Polaris. Otherwise, We're going to switch across and on to the interview. Here's Sam. Sam, welcome. And why do we need Polaris? Sam Thor: we need Polaris because I think there's two reasons. One is, you know, there's a kind of classic big tech engineer, and you can be really good at building a lot of things, but then there's leadership properties and attributes that you really want to grow into.
But I honestly think often people don't get the chance to, and taking a step like this can help people, like, unlock those abilities and features of yourself that makes you a better leader, regardless of your role. And I think this is, this comes back to the idea of leadership, wherever it is, whether it's joining a startup, whether it's leading within a company. There's this whole adage about like leadership can come from anywhere and people think that's bullshit, right? They're like, Oh, I'm just a little minion in a company.
How do I affect change? And the reality is even as a junior, you can affect change through doing things that show your leadership. And we can talk about those, the way they work and how you can get ahead, not just for getting a headset, but like for making change and how to, you know, impact your company in a successful way. James: What type of engineers should consider Polaris? And what can they expect in the four weeks? Sam Thor: Polaris is really pitched to people who don't specifically want to be founders. So I'm very happy if you wanted to be a founder eventually in your career, but the course is really for leadership in an existing environment.
And I think this really suits engineers who have worked for a couple of years, not. You know, not 10 or 20 years, but not necessarily a new grad either. Someone who's got the chops. We're going to be highly technical. So, you know, we don't want to be playing catch up on how to build React websites or whatever, like things like that.
But the goal really is to help bootstrap that. Kind of early to mid stage engineer into how to be more effective via leadership. Whether you stay at a big company, whether you're at a small company, whether you want to join a startup as a CTO and really take someone's idea and run with it and make it better in a way that they didn't know was possible through your leadership.
, James: let's drill down into that a little bit more. What exactly in your view is the role of, a lead engineer, a CTO and a startup? And how do you think that's a bit different to how an engineers might think of a role as just strictly engineering? Sam Thor: I think one of the things that I've learned over the years is actually Even though I'm not trying to get you to be a founder, I think actually doing engineering work is far less important than people think, right? I think that a big part of leadership is actually showing prototypes, demos, examples leading a team via like representative work, rather than necessarily getting into the code base and like just working. 12 hour days, right? Some people think, that's leadership, especially in a startup, right? When you're trying to build something really rapidly, I think it's far more effective to be more concise, effective, useful with your time and do things that show leadership to whoever needs it, right? If you have a small team, can you show them leadership by saying okay, this is the way we build things. I'm going to set you up for success.
Is it? leadership for your founders who you work for, you're saying I don't think that this feature you want to build makes any sense because it's actually going to take you 10 times as long, but if we cut these little bits and pieces, you know, we can be far more effective and precise, right? These are all examples of leadership in that kind of early stage environment. And honestly, they apply to big companies too, right? You can be a cog in a giant company and look at the problem and say, why are we doing this, right? This actually doesn't make sense. And to be, to have the ability to stand up and say, this isn't right. And I'm going to show you why And, you know, part of that is taking your great engineering skills and using them in the correct way rather than just blindly bashing out code like a monkey for hours and hours. James: Yeah, absolutely. Yeah, I've definitely seen some of that.
Sam Thor: Don't get me wrong, sometimes you've got to work hard and that means just getting stuff done. But I think leadership is less about, is far less about that than people think. James: Is there an example that jumps to mind in your head that you can talk about? Sam Thor: Many years ago, a long time ago, I worked on a little product called Google Keep. Which is now on every Android phone in the world, right? It has billions of users, But we were working with a reasonably difficult to work with team. And we found that the stuff we were working on got taken away from us, right? We were building a prototype and people were like, oh no, we're going to just take that.
You know, the Americans swooped in and said, this is us, this is ours now. I remember building a couple of prototype Android apps showing up a few really key things that I wanted to keep in the app. And one of those things, it's really mundane. If you've got a key right now on your phone, You can take a note, and when you take a short note, it's got big fonts. When you take a long note, it's got a tiny font. This is a really tiny feature.
Maybe it would have landed anyway. But I remember being so frustrated that they didn't understand what I was talking about, that I went and prototyped that feature, and they made it in, right? And that was my little leadership. I was a junior dev at the time.
But by showing them what I meant, they actually got it, and it stuck around. James: Tell us about your engineering career. Sam Thor: Sorry. I mean, I, you know, I was a bit of a hacker at high school and things like that. I mean, I want to be clear, I don't think that's a requirement to be an engineer, but that's certainly my background.
I dabbled in PHP and MySQL websites back when, before it was cool. It seemed to, trends come around again, it's cool again. I did a bunch of travel, worked in random engineering companies around the world, eventually coming back to Google, kind of after uni is my first real job.
I did that for 10 15 years. So at Google I did so many different things. I worked on, a thing called Google Wave, historically, people might remember that. I worked on Google Maps, I worked on Chrome, I worked on some backend infrastructure stuff.
I worked on what became Google Cloud in terms of App Engine and things like that. I worked in developer relations for many years, which honestly is where I sort of landed at the end and that was quite fun. And part of that was you know, we had different moments of different leadership opportunities and different chances. I led a part of Google Maps for a time being, uh, for a little time. We, Got told spontaneously that, oh yeah, Apple's going to launch Apple Maps very soon. We need to hurry up and ship Google Maps on the iPhone.
And so I was part of that team that did that in like three months flat. It was quite an amazing experience. We led the API side in Sydney, so if you want to drag Google Maps into your iPhone application, we did that. So that was part of, you know, we were part of the team broadly, but that was the bit I focused on. I led that team for a while. And in all that, I kind of learned a lot of stuff, and I gained a lot of experience about the way to be effective in kind of a big company.
Eventually I decided, like a lot of people, I got the itch, the startup itch. And so I went off and did startupping for a couple of years. I went to a Climatec startup, which is really my personal interest.
For me, I, you know, I couldn't see myself working on anything else. And still today I'm looking in that kind of space. And part of that came from, I wanted to see the other side of the world as well.
So I looked at things like, you know, we use Amazon Stack, and I immersed myself in that for a couple of years. So, doing that sort of stuff for a couple years was really enlightening. And the company's doing really well. I decided to take time off again, had some more kids and, you know, I have a different kind of lifestyle now doing consulting and things like Polaris. I mean, part of the course is as much you know, negative career stories as it is positive career stories.
What didn't work and how could I have done that better? And I really want to share that with the kind of next generation of engineers of how to affect change through leadership in whatever size company you happen to be in, in an effective way, right? You know, you want to show prototypes, you want to fail fast, you want to be able to talk and demonstrate your ideas effectively to people who care, right? I love the startup pitch idea. I, you know, I've looked at a bunch of startup pitch decks recently and I almost want that for engineering projects, right? Is this project going to succeed or fail? Let's find out quickly. Let's make an investment, whether that be money or time or whatever.
and make you effective. I wish we'd heard that feedback saying, it's fine to fail. You did a good job, it's a great prototype, and that's a great learning, right? We shouldn't do this thing that you're proposing or working on, and we can spend our time in a far more effective way building our MVP, getting customers, or whatever. And that's always easy in hindsight to say, of course, but that's part of my experience that I want to share with people. James: WhAt is the most interesting thing or one of the most interesting things you've worked on in your career? Sam Thor: it has to be Santa Tracker.
So I Google a, for many years I worked on a thing called Santa Tracker, which is, as it says on the tin, a website where you can go follow Santa around the world and see him delivering presents to all the good kids. In every country, I will note, we made sure that Santa visited every country, even the contested ones. Santa was a fun experience.
I worked on a team called Developer Relations, and our job was to, you know, communicate about how to make the web better. The reason we did it is because, A, it was fun, but also we kind of wanted to show what the web could do. And we were all, you know, we always delighted people with, like, little games and stuff that were like, oh, this runs on a browser, what a concept. It's a weird product because like, it maybe existed in a world where Google had so much money they could do whatever they liked, and we never made any money. But it was really fun to work on, and honestly having meetings about You know, what color should the elves, what color the clothes should be, or what game should they play, or like, even the mechanics of, you know, I was so proud of putting a solar panel on Santa's house one year, because that was my personal interest of climate tech. And then people were like, hang on, isn't the North Pole dark in winter? Like, how does that, and we were like, it's fine just move on.
We, you know, we made that message clearer of. We even got calls from kids, right? Saying, oh, we noticed this whole panel on the roof. It was really lovely to get, actually. James: So you are the official Santa tracker, man.
Sam Thor: yeah. I mean, historically, it's moved on now. But a lot of, you know, what you see there today is what I wrote. It was a really fun project for a lot of years.
And surprisingly technical. You know, it's a complicated big website that does a lot of things. And I got to write games and push the web forward.
Which is actually really lovely. You know, we did things that people didn't think were possible. One interesting thing about this project actually is that it was surprisingly more about partnerships within a big company than you'd expect. You know, people think of it as a cute little website. But I did a lot of What's the sales term? Like, walking the pavement or whatever.
The kind of, you know, going to knock on doors and saying, Hey, who wants to buy my thing? And I'm an engineer, right? I don't necessarily thrive on that kind of interaction with random teams. But Google, being a microcosm of the world, I just went to people and said, Hey, we have this cute product. Do you want to partner with us so we can, you know, so we can have a win situation, right? So we said, Hey, like, one of my favorite ones was One year, you could search for Santa in Google search, and he showed up as a 3D thing you could put into your living room.
And that was really fun, right? It was a human, you know, AR was really big, and and so I, you know, I talked to people, and I got a designer to draw a 3D Santa, and we were asking questions about his 2D, you know, you know, how do we make him 3D? Okay, we extrude him a bit, and what does that look like? And having designer reviews of like, what does a fat 3D Santa look like? I have a photo of my daughter, you know, in the room. She's quite little and Santa's standing there and and what was really funny actually is that Santa was considered to be an animal because at the time Google search only supported searching for like, I think you can still do it. If you search for like, show me a shark, you can see an AR version of a shark in your living room. And so Santa was classified as an animal because that's the only way search actually had programmed the matching to work. So he's like the dangerous.
Santa uh, you know, one name animal that he matched in search results. James: What can engineers expect in terms of the core structure? Sam Thor: What we're going to do is have basically two major topics we're covering every night. , the format is basically. A presentation and discussion and some interactive work, right? So we want to do it in person in Sydney because we think that makes a lot of sense. Yet again, it comes back to the idea of it's a leadership course, but you can all practice the product things we want to demonstrate. And, you know, with little microcosms of what they look like in the real world.
I'm not going to get you get your laptops out and make you build a prototype in an evening, but certainly experimenting with writing and how you communicate is a really important part of that. And so we like the idea of getting together in person. We also plan to have a guest speaker Potentially on the same night or potentially some sometime during the week based on their availability So we have some really kind of nice high end speakers come in and talk about their experiences not necessarily engineering ladies, but also product managers and things like that who are not necessarily your role models, but are the people who You will interact with and they can talk about what they think as of a successful leadership. Like, you know, in any company, in a big company you work with a PM, in a small company you work with a CTO, sorry, with a founder, and we want to give examples of those people and what they would expect out of successful leadership . James: WHat type of engineers should apply for Polaris? Sam Thor: We really think the course is targeted at not people who necessarily want to be a founder, at least not right now, but someone who's been working for a couple of years is experienced enough. You know, we're not going slow. It's not a course for new grads of how to get your first job, but it's for the sort of people who, their selves as someone who can level up their experience.
So you might be doing engineering and being really good at that, but how do you take that into the next level and affect change? You give you tools to affect change in the company you're in, or to look at going to like a small company, like a startup, whether you're a CTO there or a leader what are the skills you need that might be important in a really high speed, high growth environment that maybe you're not getting experience with. purely as an engineer at a traditional company. So I think that's really the candidates we have in mind. And we're not against people going to found companies, but we're very much focused on that goal of take an idea and run with it and make it better. James: Why should engineers care about leadership? Sam Thor: Not every engineer should care about leadership.
I think that's actually a bit of a misnomer. I think, Some people are very happy just writing code and told what to do. I don't mean that in a derogatory way. I mean, I think some engineers from a purist point of view, a huge fans of saying, here's a function a, it needs to run in complexity B, how can you deliver this in the fastest possible way? And that is totally a valid thing. And as an engineer that you want to do, but I don't necessarily think that's leadership. I think for engineers, some of them should care about leadership because.
That's how you grow our industry, right? Like if everyone is a cog in the system, we never challenge ourselves. We never create new technologies. We never push the boundaries of what's possible. And the way you do that is not necessarily obvious. And so Polaris is really designed to help you give you some of the skills to grow into that world where you can affect change. You can say hang on, why are we doing it this way? Can we do it differently? Oh, you have a problem.
How do we approach that? How do we approach it in a sensible way that isn't going to break the bank? Or, you know, cost us millions of years to prototype? Those kind of, skills and abilities are something that we really want to talk about and have a discussion about and hopefully take you, uh, give you the skills to go away with and go and effect change in a way that you see that it needs to happen. James: so sam, why are you excited to run this short course for us? Sam Thor: I'm at the point in my career, I think, where I want to share this knowledge with people and have a discussion and honestly, I think I will probably learn from the audience. I will, if I don't come away growing in my own way, I'll be a bit disappointed, right? So I think I'm so excited to meet these different variety of people who want to grow their skills and everyone will bring something to that world.
For me, part of my experience comes from, I did a lot of things at Google and at Startup Life where Having the right amount of leadership really propelled my career and the product I worked on forward. And I think sharing that with people is something I'm really excited about because I see it not happen enough. And there's some negative cases, right? I mean, I want to talk about myself and say, you know, why it grew my career. But for me, this idea of being a really successful follower, and that's not a dirty word, right? This course is very much for people who want to help. an existing company or grow a company being a successful follower and often finishing that last 10 percent of someone's work or really riffing on it and making it the best it can be through leadership is a hugely important thing to help grow products, make them better, and to advance your own career.
I found that for myself, that's where I've had the most success. The other reason I'm excited to do this is because I think I've seen a lot of I don't want to call it fake leadership. It's a bit of a horrible word these days. But, I've seen cases where people haven't pushed the envelope as much as I think they should. Leadership is not just saying, yes, I'm happy with the status quo. I mean, sometimes it is, right? Like, if things are working peachy, that's wonderful.
But I think you should find yourself in a position where you should be asking questions about why are we doing things a certain way? Can I make it better? Can I make the team faster? And not being boxed in. One of the great things I learned from being at Google, and Google is a special case, of course, right? Is that, Google, to some extent, owns its whole stack. If we wanted to change something we were using, we could have just, we could just go and do that. It wasn't always the right choice. Like, do I want to rewrite Chrome because we Google's building to make it successful? But I think, That kind of insight, wherever you are, and it could be in a different kind of form, can be really helpful to make yourself more productive.
And so many people are boxed into this world of like, this is the world I live in, this is how I can code things, this is how I can get ahead. And they don't see the whole vision of what's possible, right? James: What's a popular misconception about software engineering profession that you'd love to debunk? Sam Thor: Uh, I mean, I have my personal opinion, but also maybe like a more I don't know, grounded opinion. Like, I have opinions about AI and like, Copilot, which I think it's a bit of a fad, but I could be wrong, maybe I'm just an old man, I don't know. James: Why is it a fad? I'll take the bet. Sam Thor: I've seen a lot of stuff saying that AI code tends to be more corrected after the fact. So it's great for prototyping, but like, it actually has potential bugs that people aren't really aware of at the time.
And so I think people might realize that eventually the trade offs there are not worth it, right? You're better off writing the first place, having a better understanding of it. I mean, software engineering is all like this, right? Like, when I write code, or a function, or a class, like, the whole point of engineering is that I put it in a box and then I forget about it, right? This is very clear, right? That is how we build software, and if you're looking at your code constantly and looking at every single line, you're doing it wrong, right? The whole benefit of why we write software is abstraction. You want to do some functionality, you want to put it over there, you want to make the interface to it to be as simple as possible. And then it works, right? Like, that is a box, I never have to appear in it ever again, ideally. And the problem with AI is you don't know what's in that box, right? You didn't spend the time siloing yourself to build that effective abstraction to know that it's actually going to do the things you say it you set, you think it will, right? And so you end up with this kind of totally opaque thing that you don't really understand if it, like The point I'm making is that once you build it, you're supposed to forget about it. But if you didn't spend the time building it in the first place you don't really know whether it was correct or not.
James: Okay. Interesting. I will have to cycle back to that one on another time.
Maybe I'll get you in on an AI debate or something. So you mentioned a I and then there was another one. Or was that the main misconception? Sam Thor: I have a bit more of a kind of pragmatic view on the web because that's like where I spend a lot of my time and so I think of myself as like a web engineer primarily although I'm very, you know, obviously I'm quite a generalist having done a whole bunch of stuff.
I think there's kind of two parts to my answer here is I think people over index on how popular React is and I think similar things like, you know, we've seen the numbers over the years and this is from my time at Google or whatever else, is that React has a huge amount of mind space, like mind sharing people's, the way to build applications. But it's actually such still a minority of placement on the web. Like the biggest framework people use is jQuery, right? Like this hasn't changed in 20 years but I think the misconception is that people think this is the only way to build applications. And I also think people are kind of getting sick of it in a bunch of ways. Like it's got all these negative properties.
One thing I read really recently, which is quite interesting, is this view that. At a certain point, technology and frameworks have so much inertia in the way they're being used that they have no real impetus to change. Or no, there's no benefit to them to change, right? Like, if React was to suddenly say, we think this model we've done is really shit, and stop supporting it then I think you're stuck between a rock and a hard place. Like, historically, people might remember Angular, right? So Angular had a version 1, and then they had a version 2.
And now version 2 is kind of the Thing that keeps getting updated, but people are still stuck in version one. And they were stuck in a rock between a rock and a hard place because people had bought into that technology and that choice. And they were mad when Google said, oh, actually we're gonna do it totally differently. Here's a V two, which is really not the same product.
It was totally different. But what could they do? They've realized that thing they decided on that had traction wasn't that good. And so I wonder that React is in the same situation, right? It's not as good as people think, although it can make you really effective, right? I'm not diminish, diminishing, its features. But it's kind of a bunch of choices that maybe don't stack up these days. But there's such a hard thing to do to break that.
Like, do you make React 2? Do you make a whole different product? Like, what is the way forward here? James: Thanks a lot, Sam, for coming on the podcast,, and also having a chat with us about Polaris. We're really excited for the first cohort. If you would like to join the first Polaris cohort, uh, it starts on the 2nd of April and applications close on the end of March. So you've got a little bit of time to get your application in.
Please apply at the Galileo website, Galileo. ventures forward slash Polaris. And we look forward to seeing some of you there. THanks for listening for another episode. That is our first episode for 2024. We hope you enjoyed it.
We are continuing the podcast in 2024. A new episode will come out at the beginning of ever
2024-02-16 09:37
