Elijah Wilt

[ali @waterloo_intern]

https://t.co/aivc7B4A7Z

[Elijah Wilt @oojbuffalo]

Bravo 👏🏻👏🏻

[ali @waterloo_intern]

I spent 48 hours with the Kimi K3 modeling code.

It took:
- 650 mg of caffeine (mandatory)
- 40 cans of LaCroix (optional... world record (?))
- 8 papers
- 6 months off my lifespan

Finally grokked the entire lineage of Kimi K3 and how we got here... every single step, since 2019 GPT-2

[Elijah Wilt @oojbuffalo]

Why would we integrate our military tech with ANY other country? The means by which our country defends its sovereignty should remain sovereign… obviously.

[Thomas Massie @RepThomasMassie]

Unfortunately the Rule passed just now and no debate or vote was allowed on section 219, integration of US military technology and supply chains with Israel’s.

The NDAA will receive a recorded vote tomorrow with section 219 in it. I’ll vote no.

[Paul Graham @paulg]

Finally, sudo make me a sandwich is reality.

[Elijah Wilt @oojbuffalo]

@herdrdev Herdr is the best

[Elijah Wilt @oojbuffalo]

@OpenAIDevs @work_louder Meh. It’s a macro pad. I could make this in a couple of hours.

[Elijah Wilt @oojbuffalo]

This popped into my head this morning

Potter Puppet Pals: The Mysterious Ticking Noise https://t.co/t9NtQIq937

[Elijah Wilt @oojbuffalo]

https://t.co/P6vmJ3enRN

[Elijah Wilt @oojbuffalo]

herdr is such an upgrade.
new setup: herdr + cc + nvim + leaf

[Elijah Wilt @oojbuffalo]

https://t.co/IoSaUu8cuD @herdrdev

[Elijah Wilt @oojbuffalo]

@duborges I use cursor with the terminal pane pulled up all the way. I wish I was able to hide agent sidebar tho.

[Jia-Bin Huang @jbhuang0604]

Explaining JEPA in 10 seconds

[Elijah Wilt @oojbuffalo]

Hmmmmm

[Matt Van Horn @mvanhorn]

https://t.co/DM0CAuyprS

[Elijah Wilt @oojbuffalo]

The new hermes desktop app is sooo much better than openclaw's. I'd say openclaw is in trouble... 👀

[Elijah Wilt @oojbuffalo]

@konar_dev @borntogambles @sts_3d Yeah its @sts_3d

[NVIDIA AI @NVIDIAAI]

This #CVPR2026 paper from our research team is trending #1 on @HuggingFace 🤗

Meet LocateAnything: a vision-language detection model that rethinks bounding box prediction. For AI agents and robots, “seeing” is only useful if a model can pinpoint where something is fast enough to act.

Trained on 138M high-quality samples, LocateAnything decodes bounding boxes in parallel instead of one coordinate at a time, improving localization accuracy while dramatically increasing throughput for visual grounding and detection.

Project page: https://t.co/O7JMe8tzFM

[luthira @luthiraabeykoon]

We implemented @karpathy 's MicroGPT fully on FPGA fabric.

No GPU.
No PyTorch.
No CPU inference loop.

Just a transformer burned into hardware, generating 50,000+ tokens/sec.

The model is small, but the idea is not: inference does not have to live only in software 👇

[Elijah Wilt @oojbuffalo]

The guy is always on point. Couldn’t have said it better myself.

Andrej Karpathy: From Vibe Coding to Agentic Engineering https://t.co/H0UOqGADXI via @YouTube