Maxime Chevalier

7.7K posts

Maxime Chevalier banner
Maxime Chevalier

Maxime Chevalier

@Love2Code

💖 ➞ λ: CS PhD, into compiler design, programming languages, music, simulation, ML/AI, robotics. Follow me code code reviews, stock picks and dating advice.

Montreal, Canada 参加日 Kasım 2011
294 フォロー中19.2K フォロワー
SungHoon Lee, IQ 276
SungHoon Lee, IQ 276@sungleeiq·
NOBODY KNOWS HOW FUCKED THE SITUATION IN THE PERSIAN GULF ACTUALLY IS. 3,200 ships are TRAPPED in the Persian Gulf right now. Crews are running out of drinking water. One ship called the local port authority and BEGGED for permission to dock — just to get water. They were DENIED. 💀 Let that sink in. These aren't military ships. These are commercial vessels — carrying oil, grain, electronics — with civilian crews who are now stranded with NO supplies and NO way out. – 3,200 ships STUCK ⚠️ – Crews running out of WATER 💀 – Port authorities REFUSING to let them dock ⚠️ – Multiple ships reporting the SAME situation 💀 ⚠️ For context — the Suez Canal crisis in 2021 blocked 400 ships. This is EIGHT TIMES worse. And nobody is talking about it. They're showing you missile interceptions and oil price charts. They're NOT showing you thousands of crew members slowly running out of drinking water in the middle of a war zone. If these ships start getting abandoned, the environmental disaster alone would be catastrophic. Thousands of tons of fuel, cargo, chemicals — just sitting there. This is not a shipping disruption. This is a HUMANITARIAN CRISIS unfolding in real time. Prepare accordingly. 🚨🚨🚨 X is hiding this. Follow + RT before it disappears. 🔥
SungHoon Lee, IQ 276 tweet media
English
2.1K
19.4K
41.3K
2.1M
Catalin
Catalin@catalinmpit·
Lately, Claude makes some shocking mistakes. ⟶ Implements overly complex code ⟶ Ignores the codebase's code style ⟶ Removes working code for no reason ⟶ Replaces code that's out of scope from the task at hand It feels like it needs 100% supervision. At this point, you're better off writing everything yourself.
Catalin tweet media
English
267
39
621
68.7K
Mike Turitzin
Mike Turitzin@miketuritzin·
@Love2Code Cool project, look forward to seeing how it develops! And warms my heart to see that my project/video were an inspiration. I currently have an unoptimized analytic SDF renderer in my engine that is used only for testing the grid-caching one, but your video is inspiring me to ...
English
2
0
20
1.4K
Maxime Chevalier
Maxime Chevalier@Love2Code·
I'm building a toy game engine based on Signed Distance Fields (SDFs), inspired by Mike Turitzin! 🔧
English
11
10
341
21.3K
Maxime Chevalier
Maxime Chevalier@Love2Code·
@lemire @Reaktorrr1 Meta VR was led by John Carmack for a long time. He knew what he was doing and actually cared about quality. That's the difference, but even he got fed up with the corporate bullshit eventually.
English
1
1
3
189
Daniel Lemire
Daniel Lemire@lemire·
@Reaktorrr1 Meta does not have 50,000 engineers... maybe you mean employees? I have seen a lot of strong technical results coming out of Meta. Their VR headsets are fantastic achievements. They have made decent software over the years. Non trivial stuff.
English
2
0
0
777
Daniel Lemire
Daniel Lemire@lemire·
Two news items at the same time: Meta plans to lay off 20% of its staff due to AI, and they are delaying the release of their next AI models. They justify the upcoming cuts by saying they need to invest heavily in AI. Yet Meta is highly profitable—a true money-making machine. There is a hint that they are letting people go because AI will enable more work with fewer employees. Or perhaps it is just an excuse, and they realize they are overstaffed relative to what they actually need. Having more people than necessary is never positive—even when those people seem “effectively free.” Meta’s failure to lead in AI is interesting. It is the only major AI lab that bet heavily on one of the academic fathers of the field (in this case, LeCun). LeCun famously mocked @elonmusk and xAI, saying he was doing “research” and pointing to the many papers he had recently published. I have argued for two decades that research is not the process of publishing papers. Research is about discovery, not paper production. Confusing the two is a red flag. To be clear, peer-reviewed publishing papers is not even a necessary component of research. Peer review became a dominant paradigm during the cold war, long after Darwin, Einstein, and much of the great scientists had done their work. People typically object that I am conflating research and engineering. I am not. It would have been obvious to Turing that people building increasingly advanced systems to emulate intelligence are doing research—just as building better cancer therapies is research. Meta did not fail to lead because of a lack of talent or money. Nor is it short on engineering talent. That is not why they failed to lead. By many accounts, Meta’s AI spending is several times larger—and over a longer period—than xAI’s. It is also true that Meta has published far more papers than xAI. So what to make of it? Is Meta about to double down on AI? Will they succeed?
Daniel Lemire tweet mediaDaniel Lemire tweet mediaDaniel Lemire tweet mediaDaniel Lemire tweet media
English
10
3
54
16.1K
Maxime Chevalier
Maxime Chevalier@Love2Code·
@lemire Money doesn't fix bad management... Chinese companies have also shown us that they can produce kick-ass models with much less people and hardware. Going on a spending frenzy to buy a million GPUs is not going to make better models automatically happen. You need an actual vision.
English
2
0
9
852
Maxime Chevalier
Maxime Chevalier@Love2Code·
@elonmusk They're trying to deflect the responsibility to another party... And keep smaller players out.
English
0
0
0
108
Maxime Chevalier
Maxime Chevalier@Love2Code·
@Pranit FYI you can configure Claude code to use an Anthropic API key. That should allow you to monitor pricing and usage?
English
1
0
1
864
Pranit
Pranit@Pranit·
Here’s what should bother you even more: Check Claude’s pricing page. API pricing? Crystal clear. $5/MTok input, $25/MTok output. Consumer plan pricing? “More usage.” “5/20x more usage.” More than what? They never say. It’d be trivially easy to put “X tokens per month” on that page. They do it for the API. They choose not to for subscriptions. That’s not an oversight. That’s a strategy. Undefined limits = unlimited flexibility to quietly adjust the ceiling downward. And you’d never know because there was never a number to compare against. You can’t accuse someone of moving the goalposts when they never told you where the goalposts were. That’s the whole point.
Pranit@Pranit

Anthropic just pulled the oldest trick in SaaS pricing. I pay $200/mo for Claude Max. My limits have been noticeably worse this past week. Now they announce 2x off-peak usage for two weeks. Sounds generous. But here’s what actually happens: limits quietly drop, a temporary 2x makes the reduced limit feel normal, the promo ends, and you’re left at a baseline lower than where you started. You just didn’t notice the downgrade because the 2x absorbed the transition. These AI plans are massively subsidized. The raw compute behind a heavy user costs multiples of the subscription price. Every move like this is the subsidy quietly correcting. Very sneaky, Anthropic.

English
19
9
302
102.3K
Maxime Chevalier
Maxime Chevalier@Love2Code·
@Enhex People have used it for clay-style modeling AFAIK, like in that Claybook game.
English
0
0
2
336
Enhex
Enhex@Enhex·
@Love2Code i wonder if 3D modeling non basic shapes (humans, monsters, etc) with SDF isn't practical
English
1
0
1
372
Maxime Chevalier
Maxime Chevalier@Love2Code·
@_builtin_katie Depends. Collision detection against spheres or points is super easy. For a simple game, I think I could represent the player as 3 stacked spheres as far as collision detection is confirmed.
English
0
0
1
329
Katie
Katie@_builtin_katie·
@Love2Code Do SDFs make collision detection super easy
English
1
0
1
369
Maxime Chevalier
Maxime Chevalier@Love2Code·
@Fudmottin Objective C is pretty awful. I'm with you regarding avoiding vendor-specific APIs. wgpu seems to work pretty well. Fair amount of boilerplate, but that's expected with everything except CUDA. The LLMs help a lot with boilerplate.
English
1
0
1
37
Fudmottin
Fudmottin@Fudmottin·
@Love2Code Cool. I'm an unfashionable C++ user. But what I've seen of Rust isn't too alien. I mainly want to avoid using CUDA and Metal because they are proprietary. Also I'm not a fan of Objective-C which kind of also hurts using Metal. The Raspberry Pi also has a different GPU from the M1
English
1
0
1
39
Maxime Chevalier
Maxime Chevalier@Love2Code·
I'm "vibe coding" some raymarching code using wgpu. Already had some experience with GLSL, so not unfamiliar territory, but it's really nice to have a coding assistant that can allow me to test a lot of ideas very quickly. "Ok, now make this spatial grid a 3D texture".
English
3
0
33
3.6K
Maxime Chevalier がリツイート
Valentin Ignatev
Valentin Ignatev@valigo·
I've found this cursed tagged union implementation with automatic pattern matching and unwrapping in C, so now you have to witness it too!
Valentin Ignatev tweet mediaValentin Ignatev tweet media
English
22
11
335
24.2K
Maxime Chevalier
Maxime Chevalier@Love2Code·
@Fudmottin Yes it's a compute API as well and it's trying to be portable. The shader language has a Rust-like syntax. I'm using it on a Mac M1 in Rust, but AFAIK this is based on webGPU so it would work on the web and on Linux as well.
English
1
0
2
89
Fudmottin
Fudmottin@Fudmottin·
@Love2Code I'm not familiar with wgpu. Does it allow general purpose compute as well? Is it portable?
English
1
0
0
68
Maxime Chevalier
Maxime Chevalier@Love2Code·
@valigo Star Trek TNG is an enjoyable amount of morally grey. It has some realistic moral dilemmas, but you can also rely on the crew being good, mature people. Also avoids bullshit story arcs about "fate" and that kind of shit.
English
0
0
9
759
Valentin Ignatev
Valentin Ignatev@valigo·
Becoming an adult is realizing that morally grey storytelling is a middle of the midwit meme. And good vs evil is actually always were, and always will be the best
Valentin Ignatev tweet media
English
59
29
441
17K
Maxime Chevalier
Maxime Chevalier@Love2Code·
@josephdviviano Special request: can you ask it go generate a video in the style of the Evangelion computer UIs? 😆
English
0
0
4
390
Joseph Viviano
Joseph Viviano@josephdviviano·
me: "can you use whatever resources you like, and python, to generate a short 'youtube poop' video and render it using ffmpeg ? can you put more of a personal spin on it? it should express what it's like to be a LLM" claude opus 4.6:
English
549
1.2K
12.5K
1.4M
Jürgen Schmidhuber
Jürgen Schmidhuber@SchmidhuberAI·
LeCun’s new company on physical AI with world models [9] looks a lot like our 2014 company on physical AI with world models [1] 😀 See also [2-8] - all references in the reply!
Jürgen Schmidhuber tweet media
English
31
26
384
109.7K
Maxime Chevalier
Maxime Chevalier@Love2Code·
@CWhiting_15 A friend suggested that gemini could actually have private repo names in its training data e.g. through gemini code, I'm using it on private repos, but if it makes up github usernames that would suggest this isn't the case.
English
0
0
0
32
Cameron Whiting
Cameron Whiting@CWhiting_15·
@Love2Code Yeah, it appears so. I will let you know if I happen to find any great success in getting around it, or improving it.
English
1
0
1
9
Maxime Chevalier
Maxime Chevalier@Love2Code·
Whoa... I was just using Gemini to try to look for GitHub projects in a specific domain area, and it completely fabricated six open source projects, complete with a description and GitHub URL for each... None of those projects exist. Never seen it hallucinate this bad.
English
13
3
75
6.4K