Skip to content
PodSubs
ShowsPeopleGuestsJobsNewsEvents
More
BlogPodCred
Menu
ShowsPeopleGuestsJobsNewsEventsBlogPodCredNetworksSign in
Join nowSign in
  • Shows
  • People
  • Jobs
  • News
  • More
    • Guests
    • Events
    • Blog
    • PodCred
    • Networks
    Join nowSign in

Games At Work dot Biz

Games At Work dot Biz artwork

e550 — Moontella

Apr 13, 202624 minAudio
  • Apple Podcasts
ShareXLinkedInFacebookWhatsAppEmail

e550 with Michael, Andy and Michael - celebrating #MoonJoy with the #ArtemisII #crew, #Nutella in space, #AI, an #isomorphic reboot of #Wolf3D and a whole lot more!

Audio and video play straight from the show's own hosting. PodSubs stores and serves no media.

Transcript

From the show. Tap a time to play from there.

[Music] This is GamesItWork.biz, your weekly podcast about gaming, technology, and play. Your hosts are Michael Martine, Andy Piper, and Michael Rowe. The thoughts and opinions on this podcast are those of the hosts and guests alone, and are not the opinions of any organization which they have been, are, or may be, affiliated with. [upbeat music]

This is episode 550, Moon Tallah. [upbeat music] Good morning, good evening. Good afternoon, Michael Rowe here on Friday, which is always the time I get to have fun and talk to my friends, Michael and Andy about tech, gaming, innovation, business, all that fun stuff. Aye, aye, aye, aye, aye. And so let me toss it over to Michael, how are you? Aye, aye, aye, great. I'm doing fantastic Andy. How are you today?

AI. I'm good. I'm just teasing you both. We seem to have been overtaken by this whole AI wave of stuff. One of those things that is difficult to avoid. We've got some games and other things to talk about as well. And some space. Yeah, yeah, yeah, yeah, yeah. One of the great things is I can edit and take all of that out, but I won't. Well, we won't know until later. So, let's dive right in, right? So, I think thinking first of all about the things that are in the news all the time these days, we have Artemis 2,

Hey, I which at the time of recording, the folks in the capsule are still zooming their way. three and a half hours, four hours away. Well, no, no, no, no, even more, they're supposed to land at 807 time. Oh, yeah, I always forget. No, it's 14 right now. Yep, you're right. I am right. Sorry, math. It's hard. Of course you are. So looking forward to having that happen, I've been keeping track on the dashboard that we talked about last week. This has been lots of fun. So shout out there once again for really cool stuff. Of course you are.

And this week we had a couple of follow on things that I thought were cool too. First up was a way to put the pictures that the team took on your phone iPad or Mac. And I did the shortcuts. And so I've replaced my pictures on my phone. With the shots of the moon, I have to say it's just really, really cool. And so thank you for that. There's, and if one of our listeners has found a version of this for either Linux or Windows, that would be great to tag our show with because it's got to be out there.

Yes, so this is go a gallery of 14 images, right? So Michael, is it automatically changing in them for you or you just get a static one? Well, well, no, no, no, I did the suggested a further automation that every time I take it off the charger. I take my phone off the charger. It'll rotate of randomly to another picture. Oh, I see. Yes, there's a shortcut called Luna Wall, which does that. I'm just reading through the story here. Yeah. Okay. That's pretty neat. Yeah, so lunar lunar wall will get you the pictures. And then you can do another shortcut that says upon Uh disconnecting from a power supply run. [laughs]

And this and so the run this is to switch the picture. Yeah, that would be an automation in shortcuts because it triggers based off of a system event. Right. So yeah. So cool stuff. And apparently what I heard today when I dialed up on the YouTube live stream is that some of the, actually not some, but the good stuff is actually still coming down with the capsule. So they haven't done all the pictures and transmissions. There's, there's, Kind of like how we move our files around for our edit.

but wait, there's more. So can't wait for that. Michael, you saw something floating around in the capsule there, didn't you? that apparently lots of other people saw too. Yeah, evidently the entire world did and the world went nuts over it. I'm a child of a certain time, and I grew up in a certain place in one of our great breakfast components was on a porch and you'd put Nutella on it and they had a giant flying floating jar of Nutella on the spaceship.

Now, it wasn't a Costco-sized giant-sized jar. It was sort of the standard jar, wasn't it? No, it was the US normal size. Yeah, yeah, not, yeah. I don't know, you know, I was much smaller when I was growing up in Europe and it may or may not have been the same size, Yeah. but it always looked big to me. Yeah, the Costco-sized Nutella is like a bucket. Bring me a bucket. So that would be the new American size, a biggie-fry version. Probably so.

We used to do Nutella and banana sandwiches for the kids. Nope, that's good. They liked that. That was good stuff. And my favorite coffee shop has a vanilla breakfast, which is a waffle with banana and Nutella. Ooh, that's kind of cool. Yeah, the vanilla. So, I dropped a link in here too very late because driving home from the office yesterday the day before there was a piece on NPR on the radio that dealt with memorable moon movies and Bob Mandela had a three-minute or so piece on the radio and I was just waiting.

I was waiting to hear for things like moon raker, he did mention that, but I think Bob all the things that he mentioned were all these movies with moon in the title, which I think is weak sauce because there are lots of movies with, you know, things like, oh, I don't know, moon unit alpha that was a major point of the movie in the, um, oh, no, I've lost it, dang it. or or princes under the cherry moon. Well, that was listed in there too because the word moon was in the title.

I was it. I hadn't read the article yet. Yeah. it is that it's yeah it's in there just literally have my eyes are lighted on as you said it Michael yeah it's a good it's a good article yeah by the... it... Flowers killers of the flower moon. Yep. Saw that under a paper moon. I like that. I remember that it's a small child watching it. Yeah, so a lot of things with Moon in the title and I think that's what he wound up doing was just you know Yeah. Yes. There are a lot of things with moon in the title. [laughs] Focusing on that. So there should have been other movies with more of with more moons. Where the moon was part of the story. So that was that was my my little contribution there. All right. Having said that Let's let's move on to what was that next section Annie? Oh

AI AI AI AI There we go You don't need to say it again, I'll just edit it in four or five times. Okay, so so we have a couple of articles here in the web of a I think Michael this may have come from you Around Google's Gemini Now generating 3d models was that was that from you or is that from a new Andy? I don't remember which I think it might have been from me. It was, it was Michael and then I commented in our chat that Jim and I had completely failed to do things I wanted this week.

That's right. That's right. Well, we'll come to that in a second. No, I think that's one of the same. One of the great things about all the new models is they're doing some really cool new stuff and failing miserably at other stuff, which is kind of, for the last year and a half, I've seen that consistently. No matter when there's a new model, there'll be something new and cool, and then something that you just consider so basic that it fails at miserably. But yeah, I mean, the fact that they are able to do some level of simulation is...

That seems like a basic fundamental science challenge and doing fundamental science is a good thing. I do like the idea of actually generating the model and allowing you to manipulate it. It's probably just writing JavaScript to do that, which is a coding thing. I did this week, install on my local server, I think it was...

Gemma 3 or Gemma 4 model at $32 billion, so it's getting closer to being able to run locally again in a meaningful way. But yeah, I just thought this was interesting with the... Anthropic came out with new models this week that are doing some crazy stuff. in the cyber security space? Every model is going to come out new. Oh, yeah. Oh, yeah. There's some. That's really interesting stuff there. And I think we have it in the storyline for later.

We do, but we might as well go to it now because it's right there and we're just talking about it. So the point is that even though this new release from anthropic is not specially designed or trained to do security work, it is apparently adept at doing so and finding lots of problems and lots of browsers and operating systems. And well, what's so funny to me on this is I've seen multiple stories on this and it depends on what the objective of the story is one is they've released this thing and it's going to destroy all of civilization because hackers are going to use it and exploit stuff.

Yes. And the other ones, this is the best thing since Lysprid because they found all these bugs in the Linux kernel from 25 years ago. Yes? There's... It's like you've got an AI model and the logic around it that can do inference and understanding of complex code paths. And that, if you have basic understanding of security and other things, it should be able to find things. That's a given nowadays.

And I'm excited by it, and I hope... I know Anthropic has supposedly worked... working with not only some very large companies, like Apple and Mesa and Microsoft, but also some of their competitors, to say you can use this model to go search through your code base and find these hidden security bugs and clean up your security. My only concern there is, as anyone who's ever programmed before knows, every time you touch the code, you open up the opportunity to add more bugs.

So, what do you think? I'm sorry, I got distracted with every time we touch so I will let that escape my brain now. This is why we do this show virtually, right? Yeah, yeah, exactly, just starting to clap, anyway, where I found some interest here is that any time you do a scan with a new tool, obviously, you're going to find things that other tools didn't, right? And this is a lovely use case for something like an AI model anyway because it can do things that humans can't. And if it's identifying, you know, vulnerabilities in the kernel from 20 years ago that hadn't been exploited yet, well, guess what? They're now that zero day is now today, right? Because they hadn't been found.

These are minus 25 years, day, and we just didn't know about it. That's more than likely it. Minus 25 year old day, yeah, exactly, but they hadn't been exploit or maybe they have and we just don't know. Yeah. So, so you have to wonder. With all of the investments that are made on all these variety of tools that are doing, you know, one thing versus another twist lock and zap and all the other kinds of stuff and companies that have focused here. And now all of a sudden you have a preview of a new LLM that's finding things that all of those open source and corporate tools hadn't hadn't found yet that that isn't that is an eyebrow raising thing for me.

Well, three, three, four years ago I was working with a startup out of the UK that was trying to convince us to use their tool to do security code scanning and auto test generation. It was AI driven, that was the magic word back then. And the company I was working for back then decided that no, we didn't need that because we have smart people. Well, you know, yes, we

have smart people, but you're going to work on the priorities you're given. So I think it's not that developers didn't find these bugs. They had other priorities and they had other priorities period and the other the other key thing is the for for decades now we've said in the industry that finding good security developers. you know, or they were considered to be lower priority too, right, you know that what?

There's not enough of them and basically it's because to be that good takes a lot of training and skills and effort and they're very expensive. Now we've just changed the entry point to throw a couple of tens of thousand tokens at it and life is good. So what if you flip the model and say, hey, I'm going to develop code that already knows these vulnerabilities, and so therefore side step avoid them. So what code is created in theory, because the system already has this as part of the corpus, if there's a pothole in the road, don't drive over the pothole, right?

Right. So, and what I've been told, I'm sorry I had to go ahead. Yeah, but if you if you say blinded by all of the other information that you've missed the pothole in the road, even though you know about it, I think this is a danger with a lot of these things that the yes, they can the models can be trained on all of this stuff. Yeah. It does not mean that they will always pay attention to all of the information in the model. - Exactly. And therefore they will themselves miss some of these things.

- And reintroduce the same bugs that you just fixed with the model. If I'm just going to circle back to what I was referring to about yesterday with Gemini because I think it's somewhat interesting and it may be slightly related to this. So what I wanted to do, and I thought this was a good task for an LLM, was to have it do for me because it's a boring piece of repetitive work with research involved, find a list

50 number one songs on the date of my birth. So, you know, the methodology is relatively straightforward. You look back for the last 50 years, figure out the week of the year in which that date falls and then figure out what was the number one song in that week for each of those years. For which location? For which location for the UK in this case, but yes, you're right, it is important. Okay, which chart? Now, there are a couple of interesting wrinkles here.

That partway through that period of years, the actual period, the charting period changed. It used to run from a UK from a Sunday to a Saturday and then it'd be announced on the Sunday and then it changed from Friday to a Thursday, part of the way through. So that's a little hiccup, but, you know, the information is available on the internet to do it. It's tedious for you, for somebody to have to go and do all that research.

I tried a number of models and Gemini would go through the motions of doing the research, start to reply with a table, so it had done the research and then go, nope, can't do it, sorry, nope, can't, don't, no, nothing. Now, something in there was tripping up, guardrail, right, what guardrail was it? Guard Real kicked in. Could have been a brand related one, I'm not sure. It could have been, I'm making too many calls to the same URL to get the data. It could have been whatever. What I ended up doing to get some results was to actually tell it myself to manually chunk it into five batches of 10 and then at the end of it aggregate those five batches into a single list of 50. But it was not the only, I mean several of the models made a terrible mess of it.

Mystery, which is a is a French company has a really nice. Agent called La Chat, all the cat is made a terrible mess of it. But yeah, it was it was really weird and you think that with a task that you can describe as a human quite straightforwardly algorithmically that they would be capable of doing these things and they still make mistakes, even though they're told to check their work, how to do it, all of these are the things.

I think going back to the security piece which got us here. The yes you can be as clever as you like, educating and providing as much information to the model as you want. But to Michael Rosepoint, same mistakes are still going to be made because humans have made them and we've learned from humans. Yeah. The other, the other interesting thing as you were going through that Andy, I was thinking the, maybe the task you gave it was the wrong task. You could have said, write me a Python script that goes searches these things and just get the answer. Because that's a lot quicker.

Now, now, now one of them, one of the models I used did exactly that and said, "Here's a script, Hmm. go running yourself." I couldn't do that because of what was happening in my life at that particular time. I wasn't in a position to go ahead and run it. And even then, I would have needed to go and check that and make sure, you know, I had access to the right. But you're right. One of them knows that I like to write Python and therefore gave me the code in Python to do it myself. Yeah. Yeah, yeah. Oh, of course. Yeah. Yeah. go do it yourself. I love that. And you touched exactly on the point I was going to make to Andy, which is the human still needs to look at it. And that's what I've been hearing from the security experts I'm talking to. There's not a fail safe thing, right? The human in the loop still needs to be there, which if you take it at face value says, can you accelerate code development?

Yes, can you avoid some of the road bumps and the in the potholes? Yes, but the human still has to Let's play a role. Let's move along to something a bit more fun to round off the episode because otherwise I think we could we could get ourselves into a top hit of LLM debate. just yes. Yeah, we should probably think about this at, you know, different levels, maybe in 2D or 3D, maybe kind of look at it from a different angle. Oh, different dimensions. Yeah, I like it. I like it. So what we got here was actually the story about Google Gemini's 3D model simulation and making simulations. So let's talk about

alternative approach to this, which is Android on Samsung Galaxy devices, doing some kind of clever 2D to 3D. I don't know, transformation for there. Extended reality headsets. Michael, you've probably got more of an idea about this because you have a extended reality headset and we don't. Yeah, I do. I don't have the galaxy, and one of the things, I think we talked about this right when I first got it, or shortly six months afterwards, where the apple in their headset was dynamically converting images into spatial images for you, and I felt that was emotionally really fulfilling because I could take these pictures that I remember as a little kid.

And I'd be like sitting there with my grandfather again, right? And so, but this conceptually, though, is really exciting. We tend to think spatially and in multiple dimensions and it's a way of looking at things differently and having a headset that does this beyond just images into whole applications and websites, I think is really cool.

Yeah, it does look really interesting and on a taking a slightly different spin on the different dimensions Line of conversation here something I came across just today, which I threw in at the last minute to today's show Right of show was Wolfenstein and Wolfenstein started 2d and then went 3d and then when all kinds of cool, you know big blowy-upy games but somebody has taken the original Wolfenstein 3d and

into an isometric adventure game, which is really kind of cool So I think a by some metric games going back to the 8-bit era and some of the games on the BBC Micro that That you could wander around in these kind of overhead three quarters view rooms and This is really quite clever you take the original Wolfenstein map data and drop it into the same directory as their code and it will transform it into this Yeah

the Metric version of the game. It's quite challenging to play, but it's hard because you've got, you know, four back left right, but they are from a nice and metric perspective. So yeah, the movement is a little bit weird, but I like the, I like the reimagining of the way that they've done it. I was about to say, how easy was that for you? Just had to just have to remap the keys, right? Well, and in the fact that there was a quick reaction there to make a Linux version on it I thought was pretty cool too after you'd said hey, can we please have a Linux version and there is one?

Yeah, I did. I went in and asked on the thread on the ModDB website and said hey is there any chance of a Linux version because I could see they built it using Godo, the game engine, and yeah, just before we started recording, they delivered a Linux version which is really nice of them. I shall play it some more now. amazing All right. Well, and with that, I think we need to close off because their places people have to be in stuff we all have to do. So come to us again next week on games at work.biz and we'll talk about maybe, I don't know. Hey, who knows? We'll see. Bye. Actually, I shall try to play it this weekend

See ya. [upbeat music] You've been listening to games@work.biz, the podcast about gaming technology and play. We are part of the Blueberry podcasting network. And we'd like to thank the band, Random Encounters for their song, Big Blue. You can follow us at our website at games@work.biz. [upbeat music]

Open the transcript file

Hosts and Guests

Nobody is credited on this episode yet.

Credit the people on this episode

Be the first to rate this episode

More episodes

All episodes
  1. e575 — Listen to You(Tube)

    Nov 16, 202628 minAudio

    Photo by Anastasiya Badun on Unsplash Published 16 November 2026 e575 with Michael, Michael and Andy: listening to satellites, ships, and locations, geocaching songs with Meadow, moving your cursor with MousePods, iPod...

    • Apple Podcasts

    Credit the people on this episode

  2. e574 — Ring in the New Time

    Nov 9, 202632 minAudio

    Photo by Dwayne joe on Unsplash e574 with Andy and Michael M: from the Aina smart ring to the One ring, MCP architecture, Jellyfish Skills and a whole lot more! Andy and Michael start things off with the Aina smart ring...

    • Apple Podcasts

    Credit the people on this episode

  3. e573 — Daft Punk Is Playing at My House

    Nov 2, 202639 minAudio

    Photo by Robert Arnar on Unsplash Published 2 November 2026 e573 with Michael M and Andy: stories about Ikea’s Wall of Sound, LEGO’s collab with Daft Punk, decomp and recomp retro games, persuasive games and a whole lot...

    • Apple Podcasts

    Credit the people on this episode

  4. Over the Hill Gaming

    Oct 26, 202640 minAudio

    A fun discussion of gaming with special guest Ian “epredator” Hughes - retired tech luminary.

    • Apple Podcasts

    Credit the people on this episode

  5. e571 – Rando Blocko

    Oct 12, 202631 minAudio

    Photo by Nigel Hoare on Unsplash Published 12 October 2026 e571 with Andy and Michael R: lots of bricks and blocks (LEGO, and Minecraft), podracing, and random typefaces. We kick off with a look at the just-announced...

    • Apple Podcasts

    Credit the people on this episode