Your curated collection of saved posts and media
https://t.co/xH08P84miO
Post your best codex billboard
https://t.co/xH08P84miO
https://t.co/YdaVXwUC2w
Post your best codex billboard
https://t.co/YdaVXwUC2w
Vote for this Codex image https://t.co/5va6syaEvH
Vote for this Codex image https://t.co/5va6syaEvH
With the new Baidu OCR model and @MistralAI OCR 4, you might wonder which one to use. Luckily, I got you covered. Find all SOTA OCR models here: https://t.co/6GAxByELwJ https://t.co/l86bjmZLRB
3B total parameters & 500M activated, yet powerful enough to transcribe 40+ pages in one pass while keeping context intact. Meet Unlimited OCR!
We matched Mythos on public zero-days with CVEs using widely available & open-source derived models & can run it air-gapped if needed. All this with a small team out of Europe Berkeley study ranks us #1 globally in 3 of 8 categories The full evidence: https://t.co/NGVaWxFneq https://t.co/SqRCyHmhzi
This one is by far my favorite Ifykyk https://t.co/Vqv158EYfZ
Post your best codex billboard
This one is by far my favorite Ifykyk https://t.co/Vqv158EYfZ
Today we’ve expanded our AI glasses portfolio with 26 more styles and colors. Meta Glasses available today, starting at $299. https://t.co/wByxkxlR3Z
I believe everyone deserves their own Jarvis. Including you. So I built one that lives in your phone and does the work. Text it, or just call it like you'd call an assistant. Founders, executives, real estate teams already run their day through it. No app. No dashboard. No login. Try it now for free: https://t.co/XGZGOLWmBv
The new company brain. Context is the soul of your company, I learned today. Since I wrote a book, “Age of Context,” years ago I see that context is what we now call our private data. That data should be stored in a way to let AI agents and humans get more and more use out of it. This is precisely what @PromptQL does and here @tanmaigo goes deep on how it now is a company brain, or, rather, a wiki that everyone updates and uses.
@iamlukethedev @Claw3DCity @blevlabs I have exclusive access. But it builds https://t.co/kiuZ7QXLzb
Model link: https://t.co/ZES9KYlIQh Original post by @onusoz , https://t.co/11m083JV2j
16x parallel Gemma-4-26B-A4B-NVFP4 runs 🤯🤯🤯 18 output tokens/s, aggregate 300 tok/s 1 DGX Spark with 128 GB unified memory Concurrency so high I had to demo it programmatically It can go up to 32 even! 🤯 But then my screen would not have been readable for you And this is not
Model link: https://t.co/ZES9KYlIQh Original post by @onusoz , https://t.co/11m083JV2j
You can now use an expanded set of models in the GitHub Copilot app. 🔑 Bring your own provider key, use local models or use models already part of your subscription. You decide which model fits each agents session. ⬇️ https://t.co/kWcbL0S2QK
USAID and other shady leftist charities used real suffering as a human shield for leftist graft! Almost 57% of USAID money to Haiti was...spent inside the DC Beltway. Just 2% made it to Haiti. https://t.co/KXLQrdOZtb
Ending USAID saved lives. Fact.
Today, we're launching Aside, the AI browser you’ve waited for. ⋅ Crafted in every detail: vertical tabs and Liquid Glass ⋅ SOTA on agentic browsing benchmarks: outperforms Claude Fable ⋅ Full privacy: Everything runs in local and encrypted ⋅ You can use Claude or ChatGPT subscription The first AI browser built to do real work for you. https://t.co/coZkFSGfgc
@jxnlco https://t.co/bIIz5vz1Vp
@jxnlco https://t.co/bIIz5vz1Vp
Most teams training RL agents optimize for tokens per second. For RL, that's the wrong number to chase. So we rebuilt our backend around trajectories per second. 💥 Meet AOM, a Megatron backend for our open-source library ART, with 12X the throughput of our old Unsloth backend. https://t.co/PmFG0anqsf
Shared prefix fixes this. Instead of 16 separate sequences, you lay the tokens out as a tree. The system prompt gets encoded once and the 16 rollout branches hang off it. The mask lets every branch see the shared prefix while staying blind to the other branches. https://t.co/pIqAzoaHAe
The heavier your prefix, the bigger the win. At a 5k prefix with 16 rollouts you get 8X the trajectory density. Push to 10k and it's 9.6X. Same token budget, far more trajectories to train on. https://t.co/ZXOZ9GKHVP
Shared prefix is the big win, but not the only one. Maxing out packed sequence length and tuning the parallelism setup got us further, up to 35 trajectories/s on 4 GPUs. We published the configs so you can use them as a starting point for your own runs. 🫡 https://t.co/S7jomILSRi
It's all open source in ART. If you're running GRPO-style RL with a heavy shared prompt, the speedup is right there. Give it a try. Blog: https://t.co/FafhwPgeVa @OpenPipeAI ART Github: https://t.co/SHx8iBxYNv
10/10 💥 https://t.co/6nNmmA0uyM
The Cross App Access (XAA) ecosystem just got a lot bigger. Today, we’re announcing 25+ new integrations with some of the industry’s most critical applications. For teams, this means improved IT visibility and robust access control into what employees’ AI agents and apps are accessing. For end users, this means fewer consent prompts and a smoother experience, so agents can keep working on their behalf. Learn more here: https://t.co/6rcLT25C77
@WayneGlows your foul language and ignorance is of no help, and you will be blocked for it, but fyi: https://t.co/kF6otU5Lc0
In today's video we walk through how to use MAI-Code-1-Flash, a small, fast, Copilot-native coding model, to ship a real feature end to end: explore the codebase, build it, run it, and test it, all from Copilot Chat! ▶️ https://t.co/ABR2UZkLFS https://t.co/okTEO2Zv5U
lesgo! https://t.co/o4vMJLtMtM
