@digvan
Account based inUnited States
About this account
- Account based in
- United States
- Connected via
- United States App Store
Account-level information from X, not a live location or the device used for a specific post.
When I’m not building software or analyzing equity markets, I’m usually behind the lens crafting visual stories.
Joined February 2008
- Tweets2.9K
- Following4.4K
- Followers2.5K
- Likes30.5K
I have a new obsession.
I was never good at visual creative work. My first art teacher once told me I'd better not focus on art. Years later, I got obsessed with creating anime and this is very fun 🤩
Here is what I got so far. It’s not perfect. There's a lot that goes into an animation: story, characters, music, physics, props, and more. This video animation took only 10 minutes after 3 months building the engine. Let me know what you think.
Khash retweeted
یعنی حتی یک نفر در سازمان ملل به کشتار دیماه اشاره نکرد؟
ویدیو کهریزک و کیسه های سیاه بیپایان رو نشون این مدعیان حقوق بشر نداد؟ پس سازمان کدوم مللید؟! فلسطین و شرکا؟!
گه بگیرن اون حقوق بشر انتخابی تون رو!
#IranMassacre
#KingRezaPahlaviForIran
Khash retweeted
Run open models like Gemma 4 completely offline in the Antigravity SDK.
Built on Google AI Edge’s LiteRT, you can now run Gemma 4 directly on your local GPU. Zero API costs, total data privacy, and no internet required.
Khash retweeted
🥳 Excited to start revealing what we've been working on in the last few months. First, we decided to reinvent Kubernetes for agentic workloads with statefulness and fast resumption. Secondly, we are building an agentic orchestrator that will be Google's open agentic orchestrator and runtime. github.com/google/ax
Very cool
"DiffusionGemma as Jev" showcases the power of non-autoregressive architectures.
While Jev demonstrates the value of rapid decision models, running DiffusionGemma in this paradigm leverages canvas diffusion to evaluate structured choices in a single parallel pass:
⚡ ️Massive Parallelism: Denoises across an open canvas in a single step instead of sequential autoregressive token generation (~0.2s on a DGX spark).
🧠 Full Bidirectional Attention: Allows every option to attend to the full context concurrently, yielding well-calibrated decision distributions.
👁️ Multimodal Grounding: Inherits Gemma 4's spatial vision capabilities for complex visual and text decisions.
Read more about this approach here:
github.com/vllm-project/vllm…
nitter.cf/mmastrac/status/210037…
nitter.cf/mmastrac/status/210062…
Khash retweeted
We’re resetting Gemini quotas on Antigravity.
TPUs are melting with 3.8 Flash usage but we want you all to keep building!
Khash retweeted
Gemini 3.7 Flash could make a decent DOOM (1992) clone from a single prompt. Gemini 3.8 Flash can make a decent GoldenEye (1997) clone.
That's a five year jump in two weeks.
Khash retweeted
If you think looping transformers are "new", I refer you to @m__dehghani et al amazing work on universal transformers, published shortly after the transformer paper itself came out. arxiv.org/abs/1807.03819
Jeff Pu expects TPU shipments to surpass NVIDIA’s by 2028…
ASIC Warp-Up
📈Stock Picks: maintains a positive view on MRVL, MTK, and Intel following recent earnings, supply chain checks and our recent notes.
👊Nvidia vs. ASIC: Nvidia's CoWoS share at TSM will remain stable in 2027; Nvidia Vera Rubin outperforms competitors in performance/TCO/supply chain.
* TPU Dynamics: On June 14 report, we lowered Broadcom's Sunfish volume for 2026/27 due to packaging re-design and set 460k CoWoS (ASIC) in 2027; MTK Humufish (which leads in N2 vs Whalefish) potential launch in 4Q27 suggesting the window for Broadcom’s CoWoS back to TPU could be only in 1H26. Expect Humufish tape-out in around Nov-2026 with a not too high passing rate.
* Google Custom Silicon: MRVL expects meaningful contributions from Google CXL and DPU (~$3bn each by 2028), with an MPU start on TSMC N2 yielding $3bn+ in CY28 revenue.
* AWS Trainium: Reit MRVL T3 of $1bn/$2bn in CY26/27 and likely >$3bn from T4 in CY28 with a much higher ASP and better NPO scale-up.
* Intel Backend Packaging: Reit Intel's FY27/FY28 back-end revenue forecast of $1.1bn/$7bn, supported by AWS T3 EMIB-T and Google Humufish/Triggerfish volumes through 2H27-2028. We expect more wins for U.S. fabless including MSFT, QCOM, MRVL, with products across ASIC, CPU and CPO. ABF substrate to greatly benefit from EMIB.
#NVDA #ASIC #MRVL #Mediatek #SMTC #INTC #ABF
Khash retweeted
No biggie. Just Gemini 3.7 flash casually beating Fable 5, Opus 5 and GPT-5.6 on @ArtificialAnlys Analyst Agent benchmark.
Khash retweeted
Golf courses use 30 times more water than data centers.
Nobody is proposing we ban golf courses.
Ro Khanna’s kids own three of them.
Gemini 3.7 Flash from @Google on ARC-AGI (Verified):
- ARC-AGI-2: 84.6%, $0.25/task
- ARC-AGI-1: 95.5%, $0.12/task
Gemini 3.7 Flash stands out for its low cost and high scores on ARC-AGI-1 and ARC-AGI-2 relative to other frontier models.
