aaron

eng @ 🎆, prev: open-source / cv research

[aaron @thepurewal]

such a pleasure working with this group on this incredible paper!! check it out below

[Francisco Girbal Eiras @fgirbal]

Gen AI is poised to transform many fields, sparking major debates over its risks & calls for tighter regulation.
❗Over-regulation could be catastrophic to open-source Gen AI.
🚀 Our paper (https://t.co/j6XIoeJGw8) argues the benefits of open-source Gen AI outweigh its risks. 🧵

[Fireworks AI @FireworksAI_HQ]

Every company must own its intelligence.

We've raised $1.5B Series D at a $17.5B valuation.

We’ve surpassed $1B ARR and serve over 40 trillion tokens daily, with 95%+ coming from models specialized on customer data.

We’re just getting started.

More: https://t.co/97J3vTRB0A

[Dmytro Dzhulgakov @dzhulgakov]

you may have heard that glm-5.2 at 280 token/s is cool, how about 318

and we still have room to go

[aaron @thepurewal]

@bernhardsson I have been telling my friends this forever and no one believes me, especially on the new rideshare app Empower. Seems like straight line distance calculation half the time

[aaron @thepurewal]

@KimiDevs Free Kimi incoming

[aaron @thepurewal]

@Altimor @chandlerjward Curious how did Atlas compare to Fireworks?

[resham @Reshusaur]

new walk of shame:
agent still working, but the cafe closed

[aaron @thepurewal]

@hiiinternet @MenloVentures @suno @higgsfield @ChargeVC @Medal_TV can i come dj

[Simon Sarris @simonsarris]

the Pomodario technique

[aaron @thepurewal]

@maddiehfaulkner DM’d

[aaron @thepurewal]

@GarrettLord Very interested!

[aaron @thepurewal]

the singularity point is near, what are the polymarket prediction timelines at?

[Z.ai @Zai_org]

SOTA on SWE-Bench Pro (58.4): GLM-5.1 delivers significant leaps in coding and agentic performance.

[Z.ai @Zai_org]

SOTA on SWE-Bench Pro (58.4): GLM-5.1 delivers significant leaps in coding and agentic performance.

[aaron @thepurewal]

@jordan_nebula this is absolutely incredible, you deserve a brokers fee

[aaron @thepurewal]

two years ago I co-authored an ICML paper on open-source generative AI.

in 2026, the biggest thing we got right: open models did not stay far behind for long. they accelerated research, distribution, customization, and competition fast

https://t.co/7W3Z2j9iIL

[Unsloth AI @UnslothAI]

You can now train Qwen3.5 with RL in our free notebook!

You just need 8GB VRAM to RL Qwen3.5-2B locally!

Qwen3.5 will learn to solve math problems autonomously via vision GRPO.

RL Guide: https://t.co/iR9AF3BIFu
GitHub: https://t.co/aZWYAtakBP

Qwen3-4B: https://t.co/OzzCLFkSoW

[Unsloth AI @UnslothAI]

You can now train Qwen3.5 with RL in our free notebook!

You just need 8GB VRAM to RL Qwen3.5-2B locally!

Qwen3.5 will learn to solve math problems autonomously via vision GRPO.

RL Guide: https://t.co/iR9AF3BIFu
GitHub: https://t.co/aZWYAtakBP

Qwen3-4B: https://t.co/OzzCLFkSoW

[Unsloth AI @UnslothAI]

Introducing Unsloth Studio ✨
A new open-source web UI to train and run LLMs.

• Run models locally on Mac, Windows, Linux
• Train 500+ models 2x faster with 70% less VRAM
• Supports GGUF, vision, audio, embedding models
• Auto-create datasets from PDF, CSV, DOCX
• Self-healing tool calling and code execution
• Compare models side by side + export to GGUF
GitHub: https://t.co/2kXqhhvLsb
Blog and Guide: https://t.co/ENuTWal5AA

Available now on Hugging Face, NVIDIA, Docker and Colab.

[Gracia @gracia_vr]

Not long ago, we thought this might be impossible. But here we are: our Gaussian Splatting videos are now streamable, just like regular video.

No download, no app. 4DGS plays instantly in the browser on headsets, phones and laptops, with no limits on splats or length. Give it a try: https://t.co/yd7p192BTh

[aaron @thepurewal]

@bentossell One gold ⭐️ for you

[Martin Maly @mountain_mal]

i built an app that converts any space into a digital clone in minutes

as the founder of Teleport - the only iphone app that can capture high-quality 360° panoramas - i already had the perfect input when @theworldlabs released their 3d reconstruction api

📍 first test - a co-working space in chiang mai 🇹🇭

the flow:
1. capture 16 ultra-wide photos
2. stitch into a 360° equirectangular image
3. reconstruct a fully navigable 3d environment via @theworldlabs api

there’s something profound about exploring these 3d worlds. it brings you back in a way photos never could.

[aaron @thepurewal]

@allnightkev @arcmusicfest @AurisPresents