Imad

Checkout my new article about PBR: imadr.me/pbr/

@yacineMTB Eid mubarak my man, visiting algeria anytime soon?

Eid mubarak

@the_geeko1 Thanks!

@imadr_ Nice Work!!

Checkout my new article about PBR: imadr.me/pbr/

@ludwigABAP @startupideaspod I'm not locked in here with you

"The timelines are extended gutters and the gutters are full of shit and when the drains finally scab over, all the vermin will drown. The accumulated filth of all their grift and retardation will foam up about their waists and all the lizards and entreprenoors will look up and shout 'SAVE US!'"

"OpenClaw is the new computer." — Jensen Huang This is the early PC era all over again. A few power users see it. Everyone else hasn't even started. "It's the most popular open source project in the history of humanity, and it did so in just a few weeks. It exceeded what Linux did in 30 years." A solo founder with OpenClaw can now build what used to take a 50-person team. The leverage is absurd.

@rrosztoczy @kr0der @yacineMTB Show me the software you're making

I think maybe you guys don’t actually understand your users problems? I am 2.5 years in and for the last 18 months rarelt check my code (once a week) outside of running scripts/agents. The only time I have this problem is something extremely abstract unrelated to my clients. Which is obvious and dumb?

i still have to read code line by line by the way

@ordinalkek2 @DataChaz @karpathy 🤫

@DataChaz @karpathy LOOKS DANGEROUS AF. I just looked thru AutoResearchClaw and holy shit. ACP client passes --approve-all, paper abstracts from arxiv get injected raw into LLM prompts with 0 sanitization, sandbox runs LLM code on host with full access. thats a full injection-to-execution chain

THIS is the wildest open-source project I’ve seen this month. We were all hyped about @karpathy's autoresearch project automating the experiment loop a few weeks ago. (ICYMI → github.com/karpathy/autor…) But a bunch of folks just took it ten steps further and automated the entire scientific method end-to-end. It's called AutoResearchClaw, and it's fully open-source. You pass it a single CLI command with a raw idea, and it completely takes over 🤯 The 23-stage loop they designed is insane: ✦ First, it handles the literature review. - It searches arXiv and Semantic Scholar for real papers - Cross-references them against DataCite and CrossRef. - No fake papers make it through. ✦ Second, it runs the sandbox. - It generates the code from scratch. - If the code breaks, it self-heals. - You don't have to step in. ✦ Finally, it writes the paper. - It structures 5,000+ words into Introduction, Related Work, Method, and Experiments. - Formats the math, generates the comparison charts, - Then wraps the whole thing in official ICML or ICLR LaTeX templates. You can set it to pause for human approval, or you can just pass the --auto-approve flag and walk away. What it spits out at the end: → Full academic paper draft → Conference-grade .tex files → Verified, hallucination-free citations → All experiment scripts and sandbox results This is what autonomous AI agents actually look like in 2026. Free and open-source. Link to repo in 🧵 ↓

@Desenholdb @XorDev If you can prove rigorously that a neural network estimating the ambient occlusion could never be faster than a 1.5ms GTAO (by estimating the inference time, reading and writing the buffers etc?) then there is no point

@Desenholdb @XorDev I have no idea. But I think a good starting point for making an AI-something alternative is: - having a clear unambiguous ground truth, think ray traced AO - have a state of the art algorithm like GTAO - estimate the lower bound of performance for the AI algorithm

Thoughts?

@Desenholdb @XorDev Stuff like this already exist, and it's less performant than state of the art algorithms theorangeduck.com/media/uploads/…

@XorDev I wish they were planning on specific AI blocks for small things, like an AI SSR, SSAO or a AI SSS. Because that would also enable control.

@XorDev Instead of semantic g buffer it would just be albedo metallic roughness etc and the AI that will do the render will be less optimal than a normal graphics pipeline

@JangaFX Unfathomably based

Over the past year we've helped over 100 artists update their portfolios after being laid off by giving them access to our tools free of charge. We've decided to make our Layoff Assistance Program a permanent part of our business! jangafx.com/layoff-assista… #gamedev #realtimevfx

@TheGingerBill I just want my electronics to use one usb type man

I really detest the XKCD 927 comic, and I think it is actually a completely misunderstanding of the concept of standards and standardization. xkcd.com/927/

Why is every website using this slow as shit webgl library? unicorn.studio The framerate is tanking on my 4060 laptop

@sanjeetsuhag @SebAaltonen @AlexFinn Bruh for all we know the world won't exist in 10 years, who cares about the rest of your life These mac minis will probably be worthless junk in a few years

@imadr_ @SebAaltonen @AlexFinn Doesn’t matter - $200 a month for the rest of your life??

If you have your OpenClaw working 24/7 using frontier models like Opus, you're easily burning $300 a day. That's $100,000 a year. I have 3 Mac Studios and a DGX Spark running 4 high end local models (Nemotron 3, Qwen 3.5, Kimi K2.5, MiniMax2.5). They're chugging 24/7/365. I spent a third of that yearly cost to buy these computers I'll be able to use them for years for free On top of that they're completely private, secure, and personalized. Not a single prompt goes to a cloud server that can be read by an employee or used to train another model I hope this makes it painfully obvious why local is the future for AI agents. And why America needs to enter the local AI race.

@sanjeetsuhag @SebAaltonen @AlexFinn His salary is probably 100x that

@SebAaltonen @AlexFinn Do you see yourself paying that much for the rest of your life? That’s an expensive subscription.

@cmuratori @DemetriSpanos Yes!!!

Wading Through AI episode 2 is now up! In it, @DemetriSpanos and I discuss the recent Anthropic announcement about building a C compiler with "parallel Claudes", and how to properly assess AI announcements in general.

@SapphoSys non-web frameworks also allow for cross platform releases, it's been the case for decades

this article is 100% correct. there's no point in making native apps anymore because the web catched up fast and ate its lunch. plus, web frameworks allow for cross platform releases

@ssh4net @Apple This looks cool tbh

Anyone have an idea how this UI enshittification might improve UX? And when @apple lost their roots? 🤔

@FilasienoF Oh shit I've been talking to a fucking bot

Exactly. But you focus directly on the spec. You can change ideas much faster: discard everything and rewrite by altering a few sections instead of manually rewriting code. The feedback loop is quicker. It requires more skills, not fewer. That's my viewpoint, and the world is beautiful because there are many.

It mentions only two paths, but there is a third one very few are mentioning. First, the author is very insightful on how developers feel about their work. I understand the struggle, but I’ve never felt more energized, as coding is mostly drudgery. I care about hardware, data formats, and specification clarity. To me, AI is a new world to help explore strategies that previously would have taken an impossible amount of time. If you want to keep the joy of programming—that intimate feeling between craftsman and work of art—focus on specifications. Rule of the game: AI must one-shot it! If it doesn’t, your spec is wrong. So what is our contribution: we transform data in the most efficient and effective way using two key tools: hardware and math. What is in between, if specified correctly, can be done by AI. From my point of view, little has changed; I’ve just removed drudgery. Look 👇

@FilasienoF So, write a spec that is precise, correct, transform data in an efficient way using hardware and math, iterate to make it better ...or in other words: code Sorry I can't see it in another way

In a way it's true, but with fewer steps. The focus is on clarity and correctness, not code. Abstraction is faster; ideas can be experimented with much quicker. You "grow" a spec by failing and retrying in a fast feedback loop. The most powerful approach includes theorem provers, formal methods, or TLA+ in the loop. It seems AI will create a divide between those who can do formal verification and those who can't, and between those who know hardware/low-latency techniques and those who don't.

@FilasienoF This honestly looks like programming with extra steps

👇this is the way