Pinned Tweet
GLM 5.3-Flash on 2x DGX sparks: my own EXL3 4bpw quant on stock TensorFold, no patches.
vs my NVFP4 vLLM recipe:
- 8 streams (was 2)
- 1M context (was 327k)
- 34 tok/s single, 59 aggregate (was 28 / 29)
- 2.5 min boot (was 16)
Clone, ./run.sh, done:
github.com/sfxnz/GLM-5.3-Fla…
I don't know who's side loading apps on IOS but best to keep your devices up to date these days.
A vulnerability (worse_query) affecting iOS 26.7.1, 27.0.1, and 27.2 allows malicious apps to read local files from other apps, putting seed phrases, private keys, and login credentials at risk across the entire ecosystem. This is an iOS-level issue, not unique to Binance.
Immediate Action Required:
- Delete any apps installed from untrusted sources (such as sideloaded or manually signed IPAs).
- Update iOS and all apps to the latest official versions.
- If you suspect a breach, transfer funds to a new wallet created on a clean device and reset all active login sessions.
Never install unauthorized packages, and keep your software up to date.
125
My GLM-5.3-Flash recipe on 2x DGX spark now serves 8 agents at once, each with its own long context.
• follow-up turns start in 2.6s (was 7 min)
• tool calls stream live
• Anthropic-style /v1/messages
• optional API keys
Same quality, Same speed. 1M context still works.
github.com/sfxnz/GLM-5.3-Fla…
Per Elon Grok 4.8 will be nothing like Grok 4.7
We have dumped JAX/SGLang and rewritten the entire RL and inference stack from scratch in C/C++, which took some extra time.
Training (aka pre-training) is done. Lot of cleanup was needed on the SFT data to get that mostly right. Now, 4.8 is in RL and the evals are starting to climb. We will do a soft rollout to Bot users later this week and see how it goes.
Grok 4.8 will bear no resemblance to 4.7 (thankfully). It is not yet incredible, but, this time, the foundations are built right, so I expect it to evolve to be something special.
2
5
674
Genuinely one of the things I’m most excited for in the next 1-2 weeks.
For as long as I can remember, the Hermes agent desktop app felt clunky, bloated and unpredictable.
That’s all about to change.
🚨 BREAKING: HERMES DESKTOP IS GETTING A HUGE REDESIGN. 🔥
A whole new way to run your agents. Here's what's coming:
▸ All-new navigation: Chats, Agent settings, Capabilities, Library, Routines and Plugins
▸ Your agents grouped by where they run, so every machine gets its own section
▸ Pinned conversations right in the sidebar
▸ A dedicated home for your Bot Chats
▸ A page for every plugin you install
▸ A fresh sidebar, chat header, new-chat screen and message box
▸ Everything carries over: your skins, light or dark mode, profiles, chats and pins
▸ Love the classic look? One setting brings it back
Hermes Desktop is about to feel like a whole new app.
When it drops, you'll hear it here.
ALT The Hermes Agent wordmark above a silver monitor displaying the official Hermes Desktop redesign preview, with colorful wallpaper around the window. Below the screen, italic lettering reads Redesign Incoming.
I think the biggest opportunity to grow on X, grow your brand and really set yourself apart in general is right around the corner when everyone is busy playing GTA 6.
Do with that as you will.
3
464
I wouldn’t be surprised if this were true.
Ram Altman doesn’t want you to have the GPU at all, and by any chance you do get it, Anthropic doesn’t want you to run the only useful models.
THE CHINESE ONES.
Anthropic now plans to push for govts to require firmware level locks on Western GPUs that prevent "unapproved" (Chinese open-source) models from being run on them, in the name of "AI safety" and "national security."
AMD/Google are on board, only Jensen opposed it
1/
Readers added context they thought people might want to know
This is a satirical account parodying The Daily Mail; no such information has been revealed by credible sources.
dailymail.co.uk
1
183
The meme’s people will create with this will be unprecedented
Clipping is here 🎬
Grab your favorite moment from any video, add your take, and post it without leaving X.
Creators: every clip credits your original video and links back to it, so new viewers can find your work and keep watching.
Clippers: you now have an easy, native way to share moments from your favorite creators directly on X.
The best part? Both the original creator and the clipper can earn from views on the clip through Original Content Rewards.
Everyone wins when great moments travel. Clipping is one more step toward making X the best place to create, grow and earn.
5
400
Anthropic just dropped Fable 5.2, breaking probably the longest period we have had without a new frontier model release.
Huge jump over Fable 5.1 particularly in Robotics which is a new and interesting benchmark.
Fable 5.2 and Mythos 5.2 are a step above Opus 5.5 in almost all the capabilities we evaluated.
Testers have found that they communicate more clearly than past Fable and Mythos models. They generate outputs up to 30% faster than Fable 5.1 and Mythos 5.1.
4
302
Detailed metrics for your machines so that you're never left guessing.
Now supporting Discrete NVIDIA GPU setups
Support for AMD and Apple hardware landing in v0.1.1
2
7
664
