nicolas

Launching a Vision Pro app soon. I think it’s quite special and will make owning one as a dev very stimulating…

@MatthewBerman out preforming GEPA interesting

All software will be self-evolving soon. Meta-Harness is AutoResearch but for the agent harness itself. Imagine the meta-harness self-evolves as well.

PSA: If you've been running out of Claude session quotas on Max tier, you're not alone. Read this. Some insane Redditor reverse engineered the Claude binaries with MITM to find 2 bugs that could have caused cache-invalidation. Tokens that aren't cached are 10x-20x more expensive and are killing your quota. If you're using your API keys with Claude this is even worse. This is also likely why this isn't uniform, while over 500 folks replied to me and said "me too", many (including me) didn't see this issue. There are 2 issues that are compounded here (per Redditor, I haven't independently confirmed this) : 1s bug he found is a string replacement bug in bun that invalidates cache. Apparently this has to do with the custom @bunjavascript binary that ships with standalone Claude CLI. The workaround there is to use Claude with `npx @anthropic-ai/claude-code` 2nd bug is worse, he claims that --resume always breaks cache. And there doesn't seem to be a workaround there, except pinning to a very old version (that will miss on tons of features) This bug is also documented on Github and confirmed by other folks. I won't entertain the conspiracy theories there that Anthropic "chooses" to ignore these bugs because it gets them more $$$, they are actively benefiting from everyone hitting as much cached tokens as possible, so this is absolutely a great find and it does align with my thoughts earlier. The very sudden spike in reporting for this, the non-uniform nature (some folks are completely fine, some folks are hitting quotas after saying "hey") definitely points to a bug. cc @trq212 @bcherny @_catwu for visibility in case this helps all of us.

@claudeai Sir I’m feeling the AGI

Computer use is now in Claude Code. Claude can open your apps, click through your UI, and test what it built, right from the CLI. Now in research preview on Pro and Max plans.

@thdxr 😆

"why isn't everyone more impressed with me"

📢GaussianGPT: autoregressive 3D Gaussian scene generation. We introduce a GPT-style model that directly generates 3D Gaussian scenes, token by token, in a series of small, discrete decision steps. Generation, completion, and large-scale outpainting in a single pipeline. Unlike diffusion-based approaches, GaussianGPT explicitly models the scene distribution at every step, allowing for quite flexible scene synthesis. 🌐 nicolasvonluetzow.github.io/GaussianGPT/ ▶️ youtu.be/zVnMHkFzHDg Great work by @nicolasvluetzow, @barbara_roessle, @katha_schmid

@DullahTheking2 Im okay with this.

It’s time to be honest. 🚶🏾‍♂️

@teortaxesTex It’s a cult forsure

Apparently you get bullied and rejected from Anthropic for being insufficiently woke on risks of open source AI. Experience with Anth HR interview. 1point3acres, translated from Chinese. People were asking sometimes why I dislike Dario Amodei. I hope this clarifies matters.

@marijajankdqcx Great so how am I supposed to read the text

craaazy

MSA breaks the 100M token barrier Memory Sparse Attention achieves unprecedented 100M token context lengths with near-linear complexity. The architecture maintains 94% accuracy at 1M tokens while outperforming RAG systems and frontier models, using end-to-end sparse attention with document-wise RoPE.

@ns123abc The originator 🙏

>be me >start the 🚨 / BREAKING + “>” poasting style in tpot >it works >everyone in tech copies it >now every slop account and bot equivalent does it >my own poasting style feels slop to me >mfw i can't even use my own format without looking like a repugnant bot like everyone else so many things I've created quietly became the default template in culture. inevitable

@bcherny Epic

A weird part of working at Anthropic: getting a few of these each day

@theapplecycle @appltrack God the name is atrocious

The long awaited Apple HomePad is expected to launch in 6 months! 👀

@DanTAhn I have. I love it!

@nos_ult If you haven't seen Foundations on Apple TV, that is a great fucking sci fi series

Project Hail Mary was mid. Interstellar still remains a master class in sci-fi and space movies.

This is huge news for owl city

@FoxNews Parents fault.

CLOSE CALL: A skier 'miraculously' avoids colliding with a child while performing a jump stunt on a trail in Durango, Colorado.

You all are overthinking your side projects. >This guy made digital balls physically bounce off real objects/sticky notes using just a webcam, a projector, and JavaScript. Go build something fun. (via ig/bongyunng)

Not too much to ask of VisionOS 27 right? 😆

Last year, models miserably failed on USAMO 2025. This year, GPT-5.4 scores an amazing 95%, essentially saturating the benchmark. Yes, LLMs still make many mistakes, but overall, one can be nothing but amazed at what they are achieving and how steep progress in AI4Math is.

@QingpingHe1 @kevinrose I mean you aren’t wrong 😆

@nos_ult @kevinrose Ah the weekly "bug fixes" release 😮‍💨

ok, theory, a frontier model creator lands a real breakthrough, they won’t ship it to the public first - they’ll aim it inward. their own products get supercharged, & suddenly you see a flurry of releases in rapid succession that feel almost unfair. everyone will wonder how they’re moving that fast…until the model finally gets announced.