FeedScope

What people are saying on X · AI ·

NVIDIA achieved 423 tokens per second per user on the 2.8‑trillion‑parameter Kimi K3 model using a new vLLM optimization

69,919 posts · 42,966 accounts · -15% vs the day before

  • NVIDIA reported a throughput of 423 tok/s/user on the 2.8 T‑parameter Kimi K3 model, a 1.68× speed‑up on the same hardware, achieved by CentML’s Benjamin using a new vLLM optimization[60]
  • DeepSeek announced its V4.1‑Flash‑1017 model, claiming faster, more efficient performance and superior reasoning, coding, and agentic capabilities[68]
  • U.S. chip stocks suffered a 16% plunge, marking their longest eight‑session losing streak since the 2008 financial crisis, driven by AI debt fears
  • Berkshire Hathaway increased its stake in Bending Spoons to over $900 million after a $330 million investment, signaling a major shift toward AI‑focused software assets[49]
  • The Grok Bot team released its own bots on the marketplace, enabling users to import and connect tools directly, highlighting a new wave of AI‑agent integrations[50]
  • Claude Opus 5.5 has accelerated game decompilation, leading to a surge of unofficial ports, mods, and game clones, prompting publishers to issue DMCA takedown notices[72]

Posts cited

  1. [49] Hüseyin Örskaya · @orskyai ·

    Berkshire Hathaway just signaled a massive shift in the software landscape with a $330 million investment in Bending Spoons, pushing their total stake to over $900 million. This isn't just a random bet; it is a calculated move to scoop up high-quality digital assets that are being devalued by the AI wave.

    Open on X
  2. [50] CyrilXBT · @cyrilxbt ·

    Holy Sh*t , this is GOLD The Grok Bot team just put its own bots in the marketplace. Import one. Connect your tools. Done. The 10 worth stealing, by job: BUILD Engineer Bot → supervises coding agents, you merge the PRs Proto Bot → turns feature ideas into draft PRs figma bro → turns Figma frames into build specs Startup QA Bot → tests your product and reports bugs with steps GROW Distro Bot → runs campaigns, asks before posting or spending Signal Prospector → finds buying signals and drafts outreach Video Editor → cuts raw footage into captioned vertical clips RUN THE BUSINESS Stalk Bot → tracks competitors' pricing and positioning Haggle Bot → finds wasted SaaS spend and drafts negotiations MAKE YOUR OWN dr eggbot → describe a job, it builds a specialist bot Bookmark this.

    Open on X
  3. [60] SemiAnalysis · @semianalysis_ ·

    BREAKING NEWS🚨: NVIDIA has just achieved 423 tok/s/user 🚀 on the 2.8T Parameter Kimi K3 model via a new upstream vLLM optimization. 1.68x Faster on the Same Hardware! This was done by NVIDIA CentML's Benjamin, who wrote NVIDIA's new decode-specialized kernels in Cute DSL via aggressive Megakernel fusion. This is a 12,000 LoC Megakernel and is the future of advanced AI-assisted kernel generation. NVIDIA is already working on DeepSeekv4.1 and other models for vLLM too. Someone should give this NVIDIA 10x engineer a raise for this great new major feature!

    Open on X
  4. [68] Polymarket · @polymarket ·

    JUST IN: China's DeepSeek releases V4.1-Flash-1017, claiming it’s faster, more efficient, better at reasoning, coding, & agentic tasks.

    Open on X
  5. [72] Pirat_Nation 🔴 · @pirat_nation ·

    AI “vibe coding” is starting to create a serious problem for the video game industry. The release of Claude Opus 5.5 has made it much easier to decompile and recompile games, leading to a wave of unofficial ports, mods and game clones. Publishers are already filing DMCA takedowns over some of this activity. Kotaku Former Take-Two AI chief Luke Dicken says developers and publishers need to take the trend seriously: “This new wave of decompile-and-port is definitely something that developers and publishers need to take seriously.” “More invasive DRM, platform-level solutions, and probably more of a push towards streaming solutions.” Godot executive director Emilio Coppola puts more responsibility on platforms such as Steam, arguing that developers have very few ways to protect themselves while bad actors can profit from selling AI-made clones: “The burden should fall onto the platforms who are enabling this sort of behavior.” There’s also a growing concern among indie developers about showing unfinished games publicly. Coppola says he has heard from developers who are now: “Worried and anxious to even share early access versions of their games”

    Open on X

Written by an AI model from public posts; it can be wrong. Check the sources.

More topics