Doug Binks

Gamedev, making Avoyd Voxel Editor & Game with and open source tech with @juulcat.

https://www.enkisoftware.com/about#dougbinks

Runtime Compiled C++, enkiTS, GLFW, ex-physicist, ex-AAA. He/Him.

Public Key

npub1th46rew0fz9nwjq7l2esvp7xd0zkmmz8kg9xgrp03kwsm9lx8a0qj5wmja

NIP-05 Address

[email protected]

Profile Code

nprofile1qqs9m6apuh853zehfq004vcxqlrxh3tda3rmyznypshcm8gdjlnr7hspz3mhxue69uhhyetvv9ujuerpd46hxtnfduqs6amnwvaz7tmwdaejumr0dsvhgqej

Publishing to

relay.damus.io

nos.lol

Author Public Key

npub1th46rew0fz9nwjq7l2esvp7xd0zkmmz8kg9xgrp03kwsm9lx8a0qj5wmja

Show more details

Last Notes

2026-04-28 09:35:45 CEST

- reply

@nprofile…wgz8 I still have a way to go before that happens, should civilisation make it that far.

2026-04-18 10:23:43 CEST

- reply

@nprofile…0edt 6V is ~ 2x Li-ion, so you could use old li-ion batteries from a laptop pack (removed and charge with something like https://www.xtar.cc/product/xtar-vc4s-charger-112.html) or buying raw 18650 batteries. You can get holders for them with wires too. I make my own large USB power packs and small 240V inverter packs this way.

2026-04-13 10:01:27 CEST

This take on LLMs that "Torvalds and the maintainers are acknowledging reality: developers are going to use AI tools to code faster, and trying to ban them is like trying to ban a specific brand of keyboard." is a straw man fallacy.
A more similar comparison would be "trying to ban them is like trying to ban copy pasting code from other projects" (which is indeed banned).
https://www.tomshardware.com/software/linux/linux-lays-down-the-law-on-ai-generated-code-yes-to-copilot-no-to-ai-slop-and-humans-take-the-fall-for-mistakes-after-months-of-fierce-debate-torvalds-and-maintainers-come-to-an-agreement
#LLM #Linux #TomsHardware

2026-04-07 17:33:49 CEST

- reply

@nprofile…jlf3 I mean it worked last time, right... right?

2026-04-02 13:55:37 CEST

Apparently the Artemis II mission is in danger of being in deep shit... after their toilet system registered a fault. According to NASA "Mission control teams are continuing to assess the data and are actively working with the crew to troubleshoot the issue".

2026-04-02 11:48:05 CEST

- reply

@nprofile…ds0l @nprofile…hmra I read Jip and Janneke when I was learning Dutch, borrowed from a friend who's kid no longer needed them before moving on to other more complex books. This was back before the internets but you might be able to get them online as they're quite old so may be out of copyright. Be aware that knowing German can throw a bit of a spanner in the works for Dutch as much as it can be helpful.

2026-04-01 09:02:33 CEST

I am on the third day of decrypting my backup drives from Bitlocker (to use a more xplatf alternative). The soft purr of the mechanical drive reminds me of the old days.

2026-03-30 18:06:10 CEST

I was writing some C++ code, involving a new struct type to go into an array. Naming was obvious for what it was. Tested the feature for that compilation unit and deployed.
Only it turns out that I had also used a struct with that name in another compilation unit. The structs created with that are now broken, and that feature crashes when used.
So I renamed the struct, tested and deployed the fix.
There are ways around this one definition rule issue which I normally use, but hadn't here.

2026-03-30 08:57:07 CEST

The authorities placed the daylight saving time [1] change on Sunday so you get a day of feeling "this is fine" before being plunged into abyss that is Monday running an hour early.
[1] Years of daylight saving and yet I still have no daylight in the bank.

2026-03-30 08:50:21 CEST

- reply

@nprofile…jj29 Just avoid any that look like Yul Brynner.

2026-03-29 15:41:58 CEST

- reply

@nprofile…x9sp When I worked with US folk this was a great period where we're an hour closer for meetings. But in general I find the hour change sucks, though as I work for myself I suppose I could mostly ignore it.

2026-03-29 11:01:44 CEST

- reply

@nprofile…x9sp European daylight saving time change.

2026-03-29 10:48:44 CEST

- reply

@nprofile…j05z At the end of the day everything is (well there's an on-going debate about gravity)!

2026-03-29 10:36:32 CEST

People keep talking about quantum computers being the next big thing and I'd like to remind everyone that all semiconductor devices work on quantum principles.

2026-03-29 10:33:53 CEST

- reply

@nprofile…j05z I think that's how we market it: "True pixel GPU" :)

2026-03-29 10:30:10 CEST

- reply

@nprofile…j05z Sure but without certain features for efficient tensor operations and low bit float support a given GPU could be made which was great for gaming (both in terms of cost and performance) but bad for LLMs. It's a bit harder on CPUs but removing AVX support might be a good step.

2026-03-29 10:21:04 CEST

Dreamt I was part of a team founded to make CPUs and GPUs which can't run LLMs (missing features) so gamers can afford hardware. Got into an argument about ramping up a production test first as I had some experience with the process of making integrated circuits and it was pretty tricky. Of course in my dream the production process was μm scale, which is manageable in a fairly small lab and the sort of thing I'd had experience with. Making modern chips requires multi-billion euro facilities.

2026-03-29 10:14:25 CEST

Happy frack up your circadian rhythm day again for those that celebrate.

2026-03-28 11:35:34 CET

- reply

@nprofile…nkpn Both cooling and cosmic radiation are already issues similar satellites such as starlink have to deal with, so I presume these are manageable.

2026-03-28 11:20:07 CET

- reply

In other words, the GPU LLMs will likely be obsolete before the solar panels start generating more energy than it took to launch them. So you might as well power the same datacentre with the fuel you were going to use to launch it.

2026-03-28 11:11:31 CET

I did a very rough calculation of how much energy it might take to launch a datacentre to low earth orbit, vs the power draw of such a datacentre.
For a ~1 tonne GPU LLM rack satellite consuming ~100kW the energy of the fuel (i.e. not including oxidiser, launcher and ground energy costs) needed to launch this into space alone is comparable to that required to run the datacentre for 3-10+ years even on a future Starship.
So my assumption is that this isn't a good idea.

2026-03-26 16:15:22 CET

- reply

@nprofile…fnex I think you might be right, but the use of the term "data" makes me think it's potentially more than that. The linked document clarifies nothing sadlly.

2026-03-26 15:22:10 CET

- reply

@nprofile…fnex It states "Allow GitHub to use my data for AI model training", which seems pretty clear.
There is another setting for "Repository access: Choose which repositories Copilot coding agent should be enabled in." which is here:
https://github.com/settings/copilot/coding_agent

2026-03-26 14:17:08 CET

- reply

@nprofile…h0jy Personally I 'd rather neither train on the code I write, but I am indeed more perturbed by those who want to sell access to it back.

2026-03-26 09:57:05 CET

Github has a setting "Allow GitHub to use my data for AI model training" which defaults to Enabled. You might want to turn it off, thought it's probably too late and likely won't stop other bots crawling your code.
https://github.com/settings/copilot/features

2026-03-25 08:16:27 CET

- reply

@nprofile…3v4s Wow I also didn't realise they'd moved to UE5 for shipping games.

2026-03-19 16:01:47 CET

I think this article about LLMs is pretty accurate from The Register, and it's about more than just the technical issues.
https://www.theregister.com/2026/03/17/ai_businesses_faking_it_reckoning_coming_codestrap/

2026-02-20 10:37:19 CET

- reply

@nprofile…3vpz @nprofile…hm63 You can simplify an awful lot of OpenGL these days, but the largest problem is that tool support for decent profiling has pretty much disappeared. A secondary issue is that I have experienced some performance issues due to shader compilation for long shaders, which is why Avoyd's path tracer runs on Vulkan.
A lot of Vulkan's verbosity issues can be worked around with some support code, and it is getting simpler to write (though I stick to an older subset for compatibility atm).

2026-02-19 12:31:12 CET

https://www.bbc.co.uk/news/live/c70kjr9wjw0t
He is a systematic failure which will re-occur in another member of the family without change. The UK monarchy should go.
When I mention that I am for the removal of the UK monarchy* I am often asked what would I replace it with. My answer is nothing. Their powers should disappear or go to parliament. This would also conveniently help divorce the state from the Church of England.
Republicanism, not to be confused with the US pro-monarchist party of the same name.

2026-02-07 19:29:44 CET

- reply

@nprofile…9wsp Your answers are likely better found by watching the video, but it's not limited to a US plug. However some homeowners might be, and they are usually limited to less power than this device can provide to each hob (10kW to one hob).

2026-02-07 19:27:27 CET

- reply

@nprofile…5x8e @nprofile…yk06 Indeed!

2026-02-07 11:17:38 CET

Interesting induction stove design. Somewhat similar principle to the Control Freak but with an added battery to increase its power capabilities over a standard US plug, and a better temperature sensor
Stoves use very little energy over the day, but require a large amount of power for short bursts thus need special wiring. This stove uses a battery which is trickle charged
The temperature sensor uses thermal balancing
Expensive though
CW for video: cooking meat at end
https://www.youtube.com/watch?v=Sue70X7ZeS0

2026-01-29 18:42:17 CET

- reply

@nprofile…vdqu Yeah glsl is a bit primitive in that regard.

2026-01-29 18:36:16 CET

Half a day spent GPU Vulkan compute shader debugging with Renderdoc. I managed to narrow it down to a few parts of the code, only to realise that although I calculated a variable I forgot to then use it in initialising a struct.
One Fricking Single Line Fricking Fix for 1/2 a day!
Oh, Renderdoc is ace by the way.

2026-01-27 12:48:39 CET

- reply

@nprofile…p578 I can't find the article/blog post but I saw a 2D plot showing social vs economic values which placed Reform/UKIP voters as closer to the Conservatives than Labour voters so this might make sense.
Sadly Labour trying to appeal to Reform voters will likely loose them votes and gain none.

2026-01-27 12:18:32 CET

- reply

Naturally trying to time this is a pain because the driver caches compiled shaders in such a way that I have to frack things around a bit to get it to recompile fully.

2026-01-27 12:15:28 CET

It seems that the 'reform' in Reform UK means reformed out of Conservative party members, much like reconstituted meat but less wholesome.

2026-01-27 12:02:52 CET

Vulkan question: when creating multiple pipelines with vkCreateComputePipelines which are all specializations of the same shader code with the same pipeline layout do drivers optimise this better than I could by just submitting each creation to a task on a separate thread?
I'm testing this right now on the hardware & platforms I have, but was wondering if others had wider experience with this.

2026-01-25 17:57:25 CET

- reply

@nprofile…hw6s Indeed! I got a call from the bank which started with them asking me to prove who I was by providing information from my account, I responded by asking them to prove who they were and they were surprised, didn't have any way to respond so I just told them I'd call their support directly. I can't believe they make this kind of call given how easy it would be to fake.

2026-01-25 17:03:31 CET

- reply

@nprofile…x7h4 I've not seen that here in Europe.

2026-01-25 16:07:57 CET

I really wish banks and governments would somehow distinguish in their emails between "there is an important document in your inbox, sign in to read" which is just a regular update you likely already know (monthly statement with balance information etc.), and one which could have severe consequences if you don't read it.
I get that privacy and security principles prevent them from sending the actual information, but one would have thought some meta data could be added.

2026-01-21 16:41:02 CET

- reply

@nprofile…40t9 In my case the frustum code change was fairly small, as I cull against projected planes. There are a few rendering tricks for orthographic cube rendering which you could use to improve the occlusion culling performance, but I don't have the links to hand.

2026-01-21 16:35:35 CET

- reply

@nprofile…40t9 You probably have to do culling for the shadow camera as well, that's what I do in Avoyd (though currently I only frustum cull).

2026-01-21 11:03:13 CET

- reply

@nprofile…40t9 These single reflections are pretty cheap once you go down the ray tracing path. It's things like diffuse bounces, multiple reflections and volumetric scattering which can kill performance. That and traversing a deep data structure (which is somewhat needed to fit entire large maps in VRAM).

2026-01-21 10:59:39 CET

- reply

@nprofile…40t9 No, though I think I can get close with LOD and denoising at some point.

2026-01-21 10:56:49 CET

- reply

@nprofile…40t9 Yes (see alt text), it's a path traced render from Avoyd. There are more in our gallery: https://www.avoyd.com/media-gallery.html and there's a collection of people's Minecraft renders on https://www.planetminecraft.com/collection/198498/avoyd-renders/ (site has lots of trackers etc. so I'd use a private browser).

2026-01-21 10:01:51 CET

Made a render to show someone Fresnel effects, turned out to be quite a nice shot. Minecraft map: Continent of Gleamora by Dannypan.
#Avoyd #Voxel #VoxelEditor #Minecraft
https://cdn.masto.host/mastodongamedevplace/media_attachments/files/115/932/264/802/487/222/original/873db60ec8682900.jpg

2026-01-18 11:55:26 CET

- reply

@nprofile…40t9 This was done as part of moving to multidraw, but even with older GPUs without multidraw I saw big performance gains from this. The fewer state changes you can make the better.

2026-01-18 11:50:23 CET

- reply

@nprofile…40t9 With OpenGL I saw a large performance increase by putting all uniforms for all drawcalls in one buffer and binding the buffer range per draw rather than changing buffer each draw. I have now forgotten how to do this with older DX versions.

2026-01-18 11:38:14 CET

- reply

@nprofile…wgz8 @nprofile…40t9 Yeah I just noticed that. I'll ping him and ask if he's moved it if I can't find it.

2026-01-18 11:37:36 CET

- reply

@nprofile…40t9 Yeah doing nothing is always faster than doing something!
Reducing state changes is also a huge win - put all your vertex/index buffers in one buffer, all your textures in another (or texture array), all your uniforms/constants in another (if possible on API) etc.

2026-01-18 11:33:43 CET

- reply

@nprofile…40t9 There are ways to work around all of that, but all of them take a lot of time to write and have their own issues. There are bigger wins than vertex/triangle reduction to try first.

2026-01-18 11:30:17 CET

- reply

@nprofile…wgz8 @nprofile…40t9 I was lucky enough to work a bit with Stan when we were both at Intel (the Larrabee days).

2026-01-18 11:03:10 CET

- reply

@nprofile…40t9 I don't use greedy meshing in Avoyd (my voxels are not always pure block voxels) but I do use triangle reduction using a variant on Stan Melax's BunnyLOD Polygon Reduction https://github.com/dougbinks/BunnyLOD
Reducing the vertex/triangle count for rendering is about a 1.2x performance bonus for Minecraft like scenes. There are also memory reduction benefits and due to the way I allocate this can improve multidraw and lead to > 2x performance benefits.

2026-01-17 20:36:21 CET

- reply

@nprofile…40t9 Personally I think the first one works better as a thumbnail/preview.

2026-01-17 11:21:05 CET

- reply

@nprofile…40t9 My reasoning is that gltf is a fairly simple and straightforward format with an excellent ecosystem of open source software. It has a reasonable extension system, and you can implement the parts you need in a fairly small amount of code.

2026-01-17 11:16:21 CET

- reply

@nprofile…40t9 gltf is in my view the best option for most 3D models these days. A lot of folk are also using USD but that's more complex.

2026-01-16 10:50:52 CET

- reply

@nprofile…zdn5 Given that I need to redo our undo/redo system I would love to know if you find any interesting articles/posts.

2026-01-08 09:57:48 CET

- reply

@nprofile…mmts @nprofile…x9sp The mesh shader version would make sense if the voxel data is stored on the GPU and the mesh shader emits visible faces towards the player.

2025-12-30 12:54:58 CET

- reply

The above post implies a filter: those whom I am helping are by definition lacking the domain knowledge I have on that particular subject and I am not saying I am *far* better at internet search than them in all domains.

2025-12-30 12:50:21 CET

Some people say you have to learn to use AI/LLMs or you will be left behind in your career by those who can.
I have noticed with my experience in helping people on the internet is that I am *far* better than them at using internet search to find a correct answer. This is not due to my ability to use the search tool, it is because my domain experience helps me to both create search phrases and quickly interpret results.
Domain experience beats tool experience. AI/LLM use works against that.

2025-12-29 16:04:09 CET

- reply

@nprofile…cpuy If your Spirit Medium has high absorption or scattering you might not.

2025-12-17 11:49:00 CET

- reply

@nprofile…cpuy As someone with dyslexia who has worked with both I am always reading GLFW and GFWL inverted.

2025-12-12 16:13:22 CET

- reply

@nprofile…h2az @nprofile…enqw The simplest cross platform way for inter process communication I used to use was file transfer and locking on a shared drive. I used this when doing research to farm out processing to a bunch of servers in the department. For large data processing with no real latency requirement it worked really well.

2025-12-12 16:07:00 CET

- reply

@nprofile…cpuy Well just let us know if you need anything. There's a lot of Minecraft users who use Avoyd for renders, so it can also be another way of advertising the game to a similar community.

2025-12-12 15:55:30 CET

- reply

@nprofile…cpuy If you want editor support for your structures let us know and we can add something to Avoyd to export to this json format. It won't look the same in Avoyd, but we can create a material palette based on your blocks.

2025-12-09 18:23:29 CET

- reply

@nprofile…fnpd @nprofile…j2qh Thanks I didn't spot that one!

2025-12-09 17:38:33 CET

- reply

@nprofile…nzck Funny to hear you mention Sega and Rez: @nprofile…q933 and I both love Rez (it's an influence for Avoyd), and I was lucky enough to work with Sega on some Dreamcast to PC ports. Sadly didn't get to go to Japan during that time.

2025-12-09 17:35:34 CET

- reply

@nprofile…kmm9 Corrected, sorry about that! Now you have to do a podcast :)

2025-12-09 17:32:45 CET

Recently watched a few of the Łukasz / Wookash Podcasts where folk discuss programming & their careers
So far I've enjoyed @nprofile…w52x, @nprofile…kmm9, and Alex Evans plus Anton Mikhailov's discussions. Next up is @nprofile…aved + Dennis Gustafsson, @nprofile…yspl, @nprofile…ascj, @nprofile…szgl, @nprofile…fnpd
https://www.youtube.com/@WookashPodcast/videos

2025-11-30 11:32:16 CET

- reply

@nprofile…fnpd @nprofile…t5tr This explains the Quake story - Carmack goes into a nail salon to get some pro filing and comes out with well optimized software *and* nine inch nails.

2025-11-29 21:54:00 CET

- reply

@nprofile…mksd I admit that I did a bad job in the mastodon post of describing what we're doing!

2025-11-29 21:02:30 CET

- reply

@nprofile…mksd Also on the readme at the top I have:
This is intended as a simple format for exchanging voxel scenes, models and animations etc. There will be a binary voxel data representation with some compression, but extremely highly compressed voxel data is not a primary goal.
So OpenVDB on its own does not solve this, plus it is far from simple.

2025-11-29 20:51:49 CET

- reply

@nprofile…mksd see the section on USD here which briefly mentions OpenVDB https://github.com/enkisoftware/OpenVoxelFileFormat/discussions/2
Also note that this is not about just.a format for voxels, but a lightweight format for scenes made with voxels.

2025-11-29 12:44:50 CET

- reply

@nprofile…t5tr It's only debugging if it's from the municipality of Buggingen, DE. Otherwise it's just sparkling printf's.

2025-11-26 10:39:43 CET

- reply

@nprofile…yspl I mentioned that in the start of the discussion under USD since usdVOL uses OpenVDB. I think it's a bit too complicated a format for my intended use.

2025-11-25 13:21:27 CET

Started a discussion about an Open Voxel File Format on Github. This is about storing models, scenes, animations etc. made with voxels.
Please comment there if you're interested, and boosts welcome for reach.
https://github.com/enkisoftware/OpenVoxelFileFormat/discussions/2#discussioncomment-15074076
#Voxel #VoxelGameDev #glTF #GameDev #IndieDev #OpenSource

2025-11-24 12:51:53 CET

Yet another bug carefully relocated out of my code by a combination of Runtime Compiled C++ and Dear ImGui. Being able to add debug output after observing an issue, and then fix it whilst never having to exit the program really helps.

2025-11-22 12:11:46 CET

- reply

@nprofile…aels I've made a draft PR for this as https://github.com/glfw/glfw/pull/2792 which you can follow if you want to know when/if it's merged or chime in on the design questions.

2025-11-21 19:01:43 CET

- reply

@nprofile…aels In case you just tried this I have just pushed a bugfix :)

2025-11-21 18:28:22 CET

- reply

@nprofile…aels I have a potential fix on my branch: https://github.com/dougbinks/glfw/tree/wait-events
This only has Win32 and Wayland implementations for now, but the other backends should work as before. I've not fully tested all paths for events so some may not cause a wait to end.

2025-11-21 12:12:17 CET

- reply

@nprofile…aels Did you stick with Fedora Kinoite or another distro? It would be great to have a local distro to test with.

2025-11-20 19:28:26 CET

@nprofile…nzck Another use of Dear ImGui: https://n-e-r-v-o-u-s.com/blog/?p=9861 see about half way down the page for their tool.

2025-11-20 18:34:44 CET

- reply

@nprofile…aels Damn I copied the wrong line, apologies! Glad you got it to work.

2025-11-20 18:34:03 CET

- reply

@nprofile…aels This is non trivial in Wayland, events are dispatched to callback functions. I've been intending to add a GLFW event counter or flag so we only exit on events the user code cares about, but haven't yet had time to write this.

2025-11-20 18:08:47 CET

- reply

@nprofile…aels For a simple workaround I would switch the app to X11:
if(glfwPlatformSupported(GLFW_PLATFORM_NULL))
glfwInitHint(GLFW_PLATFORM, GLFW_PLATFORM_NULL);

2025-11-20 18:07:24 CET

- reply

@nprofile…aels We do have a similar issue reported: https://github.com/glfw/glfw/issues/1911
I have not yet been able to reproduce this.

2025-11-20 18:04:34 CET

- reply

@nprofile…aels I just tested the latest main branch on Ubuntu 24.04 on WSL2 and glfwWaitEventsTimeout(1.0) waits 1sec for me unless there has been input.
I would check the latest code and the tests/events.c program first changing the glfwWaitEvents() to:
glfwWaitEventsTimeout(1.0);
printf("%08x at %0.3f: glfwWaitEventsTimeout returned\n",
counter++, glfwGetTime());

2025-11-20 14:20:38 CET

- reply

@nprofile…aels I'm likely going to be out for most of the rest of the day but if you add some info I can look into it. To check which platform your GLFW program is using on Linux see https://www.glfw.org/docs/latest/intro_guide.html#platform

2025-11-20 14:19:03 CET

- reply

@nprofile…aels So I help maintain GLFW. Are you using Wayland or X11 by the way?
Depending on the backend the wait can return due to events which GLFW doesn't explicitly handle, for example on Windows any message sent to a GLFW window will cause the wait to return. I don't recall if the same occurs on X11 but I think this shouldn't happen on Wayland.
So this could be a bug or some messaging issue (which is a slightly different bug).

2025-11-17 17:49:33 CET

- reply

@nprofile…548d Conservative Stupid Guys (CSG)

2025-11-17 15:09:34 CET

@nprofile…nzck Kitten Space Agency (spiritual successor to Kerbal Space Program) is using Dear ImGui (see 5:57) in game for some of the UI at least: https://youtu.be/Z5_g5zXV5_Y?si=QgMoxi9nJ8K5ck8Z&t=357

2025-11-17 15:00:51 CET

- reply

@nprofile…04cx I like the IDE, particularly it's powerful debugging. However I turn off pretty much every smart developer tool, even those from before Laughable Language Manipulators. I do find basic code completion useful though (type object.M and get a list of functions) beginning with M.

2025-11-17 14:20:02 CET

- reply

@nprofile…yspl Indeed. I think what these people want is artificial intelligence without agency/self-awareness. Something obedient to replace those pesky workers with their free will.

2025-11-17 13:55:55 CET

"Visual Studio 2026 is AI-native, making it the world’s first Intelligent Developer Environment (IDE). "
Oh gods please let it stop.

2025-11-16 12:40:22 CET

- reply

@nprofile…5e4e The need for cooling does depend on a lot of factors, and admittedly many scientific sensors need to be pretty cold. However the sun will heat stuff up pretty quickly with no convection or conduction to space if you don't design well.

2025-11-16 12:36:01 CET

- reply

@nprofile…5e4e Actually one of the big issues in space is cooling. Fortunately if you can orientate your craft correctly you can point your cooler at something very cold. Lots of different approaches - at the Atmospheric and Planetary Physics labs I did my PhD at I recall we had Sterling cycle heat pumps, and passive conical black body radiators. Looking into the later was like staring into an absolute void - a high surface area structure deeply black in visible and infrared wavelengths.

2025-11-15 12:13:16 CET

- reply

@nprofile…fnpd @nprofile…vlqt Ah! So DWORD is actually short for Dos Word.

2025-11-15 11:53:48 CET

In their video "Could the AI bubble pop?" the folks at The Economist speculate that the money being thrown into the fire only makes sense if this is a bet on one of the companies building an AI which self improves and leaves all the other companies in the dust (beyond human intelligence AGI). AKA Potentially The Singularity.
Not mentioned is that no one in the industry has shown technology beyond powerful pattern matching. The rest is pure wishful thinking.
https://www.youtube.com/watch?v=T-26Zv7pVW8

2025-11-07 17:50:51 CET

- reply

@nprofile…t5tr I am imagining Sauron's Eye but with a turbine where the eye is.