Ryan Fogle

It is here let’s go! @nvidia @NVIDIAAI #LocalAI #NVIDIA #AI #DGXSpark

@mr_r0b0t It sure doesn’t feel like it with this X community! 🤣

You are still early. Extremely early! Updated July 2026: Each dot = ~3.3 million people 2,500 dots = 8.3 billion humans Gray: Never used generative AI → 5.9 billion (71%) Green: Free chatbot users → 2.3 billion (28%) Yellow: Pays for AI → ~80 million (1%) Red: Uses coding agents → ~12 million (0.14%) Most of humanity still has no idea what’s coming. ~2% of U.S. households pay for AI subscriptions. We’re not even close to mainstream yet. Source: gradually.ai (July 2026)

@TechMDAI @alecqfong @NVIDIAAI Crazy setup

Dreams money can buy - 2x DGX Stations Most advanced local rig on here. Displaying CX8 connection 2x DGX Stations - GLM 5.2 nvfp4 (120 tok/sec) by @alecqfong @NVIDIAAI

@takuz0_ Sheesh looks great

Testing 8x DGX Spark nodes in an open world setup :)

@MiaAI_lab Setting this up right now. I really don’t understand the logistics of something like this or how it is even done. I’m gonna have to look into it.

Rejoice single DGX Spark owners! 💫 DeepSeek v4 Flash for single DGX Spark BEATS the 2x DGX Spark version on agentic workflows! 🤯🤯 Haven't done full coding tests but for agentic workflows it BEATS the FP8 version running on vLLM running on dual DGX Sparks. I fully expect it to be great at coding too! However, it's much slower, as expected: Aggregate tok/s (1-2-4-6-8-12) 26.7 tok/s → 32.9 → 46.5 → 54.1 → 58.5 → 58.5 tok/s Per-stream progression: 26.7 → 16.5 → 12.1 → 9.5 → 7.7 → 5.2 tok/s That means up to 59 tok/s across 12 concurrent sessions. This is the best model to run on a single DGX Spark! All credits go to @bleysg! I have published a simple start/stop recipe: github.com/MiaAI-Lab/Deep… I still recommend running DeepSeek v4 Flash on 2x DGX Sparks because the speed difference about 3x. Additional details below 👇

@Jimmy69191607 @XFreeze Oh cool gotcha thanks!

@fogle_ryan @XFreeze After one year.

SpaceXAI just launched a limited-time discount on Grok credit top-ups If you hit your usage limits, you can now buy extra credits at up to 45% off through August 14: • $5: 10% off • $20: 20% off • $50: 30% off • $100: 40% off • Larger custom top-ups: up to 45% off A $500 credit purchase now costs just $275 A seriously good deal for heavy Grok users who do not want to wait for their limits to reset

@u1tra_instinct @ciprianveg Crazy set up! I am only at 1 spark.

i'm in awe.... 16 DGX-Spark/GB10's cluster is being created here.. First one i have seen. you're not following @ciprianveg and watch what he is building frontier level local consumer inference monster, you might be missing out on the future when 1, 2, 3T parameter is locked out for you

@ciprianveg Crazy so much power!

Preparing 16xGB10 cluster for Kimi K3, Deepseek V4 Pro, and future GLM 5.5 and Minimax M4..

@wicksplay Nice Good Luck!

DeepSeek V4 Flash 0731 @ 118.2 tok/s First trial. Just getting started. Going to see what I can push this to without sacrificing quality

@wicksplay That is a crazy stat 5 years from now will look very different

The world is going to change and people dont even know it. They arent even ready for it. They have no clue what we are even working with ~2% of US households play for AI subscriptions Read that again.

@CardilloSamuel That’s crazy i’ve seen those before, but never really knew exactly how people did it congrats though 500k!

for those of you who might not know or have forgotten, i’ve got an ai MALE (no ofm bullshit) influencer (identify will remain secret) on instagram since beginning of the year and it reached over half a million followers. 🫢

Grok Imagine 1.5 has been so cool to play with. Prompt: Someone is on roller blades zooming while skating on a black hole like it is a solid surface getting closer to the center @grok @SpaceXAI Anyone else testing the new update? Drop your best ones below 👇 #GrokImagine

@SpaceX @Starlink Crazy the size of these rockets! I wonder how long that flame from the booster is?

Falcon 9 launches 24 @Starlink satellites from California

@tonysimons_ Nice!

Holy shit. My first-ever 𝕏 Creator Revenue Sharing payment just landed: $705.65 I started posting seriously on January 6 with no audience, no playbook, and absolutely no idea where any of this would lead. Seven months later, something I built one post, one reply, and one conversation at a time just paid me real money. I’m proud as hell. But mostly, I’m grateful. To every person who followed, replied, reposted, bookmarked, encouraged me, challenged me, or gave one of my strange little internet experiments a chance: Thank you. You made this possible. I’m also making a promise right now: I’ll always be transparent about what I earn on 𝕏. I’ll share the wins, the slow months, the numbers, and the lessons. I want other small creators to see what’s genuinely possible without the smoke, mirrors, or rented Lamborghinis. First payout: $705.65. I still can’t believe I get to type that. We’re just getting started. 🥹

Benchmarked @OxSero's DeepSeek V4 REAP on my Single DGX Spark: it looks like it is a bit better than the Laguna S 2.1 and Fable Fusion 27B models I had been using. huggingface.co/0xSero/DeepSee…

Connect Superwhisper to Grok Build with your SuperGrok or X Premium subscription

@mattbroadstreet @composio @NousResearch Ohh ok interesting

@fogle_ryan @composio @NousResearch Those are my two favorite, you can also use Hermes for your orchestrator and planner and have it delegate to work through pi as the actual implementation harness. Which means you still have a conductor in Pi who carries out the orchestration. Hermes becomes management & QA

We ran Kimi K3 through 3 more agent harnesses (Pi Agent, OpenCode, and Codex) bringing the comparison to 6 harnesses across 26 agentic tasks. 2 things stood out: Codex ranked last on success despite mid-pack speed and cost, while Claude Code cost about 4x more than Hermes. 🧵🧵

@0xSero About to try this out

REAP out, this is temporary for DGX Spark enjoyers Will work on exl3 which will retain all experts and perform better In the meantime, enjoy! huggingface.co/0xSero/DeepSee…

@NousResearch Awesome!

The first official Hermes Desktop plugin is Kanban, now desktop native by popular request and significantly upgraded. A plugin can add its own page, sidebar row, hotkeys, status bar actions, and backend endpoints. You can write your own plugin or import one using the SDK.