Turin Turambar

@SebAaltonen @Jonathan_Blow I am a software developer with 15 years of experience, and I can confirm that indeed, most of my work is glue code between a web browser, and a database. I use a IDE of course and debuggers (Visual studio), but about profilers...

@Jonathan_Blow Most of the code nowadays is glue code between web frameworks and web libraries, databases and HTML UI. The code that me and you wrote isn't considered normal code today. Most people never use IDEs, debuggers and profilers. I don't mind that AI writes this glue code nowadays...

This is such a *completely* different reality from where I live, at this point it's just difficult to say anything meaningful about it at all. x.com/sama/status/20…

🌙 March Update Preview is here! Travelers, a new dream awaits. Read the full update preview on Steam: store.steampowered.com/news/app/24447… #ShapeOfDreams #Roguelite #IndieGame

@MatanHalevy @burkov Even if they cannot train in specific puzzles, there is the risk that the finetuning only serves to make the AI better at this general type of visual puzzles and nothing more, it doesn't indicate an uplift of real logical spatial reasoning, for example

It's interesting because ARC-AGI-2 is specifically designed to resist training contamination. The tasks are novel visual patterns with rules that have to be inferred from scratch. Hard to finetune for without improving the underlying induction capability. The more interesting objection is whether abstract induction benchmarks predict anything you care about. That's a fair question. But "marginal elsewhere" is doing real work here: Gemini 3.1 leads ARC-AGI-2 by 8+ points over Opus 4.6. That's not marginal.

This is why these benchmark improvements are all crap. Gemini 3.1 has marginal improvements compared to the competition on all benchmarks but one: Useless-Puzzles-2. This means only one thing: they have finetuned 3.1 Pro specifically to improve on this single benchmark without modifying the underlying architecture. Tomorrow, someone releases another useless benchmark, and Gemini will suck on it until the benchmark becomes hyped and they add the training data for this benchmark to the finetuning dataset. So boring.

@ChrisAvellone I agree, I remember most people loving it, once the game got around two months of updates to fix the most broken stuff. It wasn't five years of wait!

Uh, what? It was buggy as hell (which was on us) but this doesn’t track.

@ElMapaInversor @jluiscava La economía en K tampoco ayuda...

Es el síntoma clínico de una economía 'zombi', @jluiscava . Tener PIB positivo con creación de empleo fallando confirma que el crecimiento es puramente nominal (gasto fiscal), no productivo. Estamos entrando oficialmente en la casilla de Estanflación. La Fed pierde su coartada del 'empleo fuerte' para mantener tipos altos. El pivote forzoso está más cerca.

Recuperación sin empleos. Los empleos creados en los sectores no agrícolas de los EE.UU. aumentaron en diciembre por debajo de las expectativas.

@brandonleblanc @WithinRafael ok, when why I have github copilot in Visual studio with errors saying it cannot find odr.exe?

@WithinRafael Where did we say this was on a slow roll out for WIP? My understanding is that this hasn't begun rolling out yet at all. I pinged some folks on my side as well just to be sure.

Anyone on Insider track with odr.exe on their machine? Microsoft claims it's on a slow rollout, so presumably > 0 users have this?

@pagb666 what's the chat/interview, btw? you have the url?

@TurinTurambar Yup... no say about that, but they had spent already so much on the game they had to go on.

Breaking: In a recent chat/interview, Romero has revealed that his current videogame project has "survived" the MS cancellation. Details: - 110 ppl working on the project - Project leads are the same - Went over a complete redesign, but the core elements are still there >>>

@vkrajacic Hey, do you plan to add the ability to add more icons to the top bar of File Pilot? I'd like to add the typical delete / cut / paste (and even the copy to / move to) to the right of the bookmark icon. Good job so far!

@MasterPhW @WindowsLatest @vkrajacic The price is clearly 40 bucks, not 200. Why do you need VIP support?

@WindowsLatest @vkrajacic File Pilot looks interesting, but 200 bucks is way to much for a Explorer alternative imho. But will try it as long as it in its beta test phase. Thanks for the recommendation.

Tested: Windows 11’s ‘faster’ File Explorer (preloaded) is still slower than Windows 10. On a 4GB RAM test machine, File Explorer used ~32.4MB of RAM without preloading. After turning on preloading, Windows quietly adds ~35MB more. So Explorer now sits at ~67.4MB in memory before you even open it. 35MB isn’t a big deal in 2025. Your browser eats that for breakfast. That's not the point of my testing. Why is the preloaded File Explorer still slower than Windows 10? I made a video to compare the two versions.

@gmWojak57 @Hesamation That's cool. But it isn't "thousands of people are playing on Steam." It's more accurate to say that 'thousands of people have played it on Steam.' Current Players 10 today's peak 25 all-time peak 370

@Hesamation For anyone wanting to check it out: store.steampowered.com/app/3401490/Re…

bro packaged math and programming into a game and 1000s of people are playing it on steam. this looks so cool 🤯

@rixor14 @ivy_sly_ "what's the reward for playing a video game" is pretty cringe. It's a fucking video game, whatever 'reward' you get is going to be make believe. Unless you are a pro esport player I guess.

@ivy_sly_ Being against achievements is cringe as hell ngl. Especially in an arcade style game with high scores. What's the reward for it? Are there in game leaderboards?

for the record: I develop games for people who play games to have fun

@mesetatron @ArturoEcho @IceSolst And then you have people inventing new render techniques in a engine en.wikipedia.org/wiki/Shadow_vo… , and having to optimize them

take MMOs for example, they have very different architectural requirements to SaaS, and very different latency requirements for requests for example, it's not unusual that an online game has a custom scaling controller for kubernetes, because game servers are heavily stateful, and you can't just shut them down and replace them with another without losing state online games likely have extensive use of high-throughput message queues (like kafka) because they need to handle a large amount of event data from players doing things in the game world, any one of which might trigger experience gain or level up they likely have several pubsub busses for things like chats and notifications the games themselves have a lot of interesting DSA problems. moving around the game world requires pathfinding, likely A*, possibly nav meshes for optimization smarter NPCs might need complex decision-making frameworks, like decision-trees, or GOAP some game logic require graph traversal solvers for determining win-state, and may need extensive alpha-beta pruning to run in reasonable time, particularly turn-based games and card/board games the list goes on. I switched from Enterprise SaaS to game dev, and can say that yes, game dev codebases are a whole lot more complex and diverse, and that doesn't include the fact that with games you also have to consider sound, voice acting, art, character art, graphical programming, writing, game mechanics, metagame progression, and all the tooling to support these, and so much more beyond code itself happy to answer questions

Games have the most complex codebases I’ve ever looked at. Not even touching on game design, all the clever tricks that come with performance and network optimization, and the fact you have the most entitled client base of any industry. Coding throughput was never the bottleneck.

@mesetatron @ArturoEcho @IceSolst This is kinda surface level, even. I think some old console games did specific visual tricks using the fact the tvs electron beam moved in a specific way. Or games who take care of ordering the data in the cd so the key part was in the outside of the cd where the speed was faster

@CodeRed_dev There are around 14.000 games produced this year. Only on Steam, mind you, the number would be much bigger if we count with mobile games, for example. If anything, there are more games than the video game market can absorb, so lots end up not making money.

Gamedev has an "I'm unemployed" vibe.

@pagb666 @ID_AA_Carmack en.wikipedia.org/wiki/Id_Tech

Good evening @ID_AA_Carmack . Quick question, is the Quake engine IdTech1 or does that term refer to the Doom engine? I always acknowledged that Quake is 1, but I'm seeing people calling it 2 and Quake2's 2.5. Thanks!

@SodStm Not a good marketing tweet. The game has 8 characters, not 4. People who don't have the game don't know you refer to only 4 characters show in the short video.

Four characters. Four Ults. One choice. 💥 Watch the full clip and cast your vote for your fav in the comments. 🔥 #PlayShapeofDreams

@pagb666 @QuadDamnPodcast @Halo Yep. And the part of "Some say @Halo had the speed..." what speed, lol! I literally watched gameplay just now in case I was remembering, but no, it's pretty slow.

@QuadDamnPodcast @Halo Nope. Two words: Regenerating Health

💥Is Halo a boomer shooter? Some say @Halo had the speed, the chaos, and is now old enough.....others say two-weapon limits and slow movement disqualify it. Let the arena purists and lore lovers fight it out?

@icanvardar Atlantis...

hardest test for americans

@pagb666 It's even more unacceptable because BL4 mostly looks like BL3, just a bit better. With the art style they use, they don't have photorealistic raytraced lightning, or super highpoly models or 4k textures, or etc etc etc.

Shut up already Randy, you're only making it worse. A 5080 for barely achieving 60fps at 1080p on Med-High settings is far from acceptable.