@0xVRAM

needs more VRAM

Joined June 2013
  • Tweets191
  • Following409
  • Followers417
25Media
Pinned Tweet
0xVRAM
@0xVRAMAug 7
It’s dual spark time. Let’s see how long this takes to set up… #GX10 #DGX @NVIDIAAI
11
3
2
94
11,389
Latest @deepseek_ai 4.1 Flash distilled this annoying behaviour from Claude lol. Its at ~100k context,
2
34
Why do I have to practically verbally hit @grok with a stick for it to research stuff online using a web search tool, please fix this in Grok 4.8 😭
1
1
158
Cheers @calcsam 👋 Looking forward to reading it!
1
2
92
Why is it that every time I open X my feed is filled with Japanese people building the stuff of dreams. Some insane rigs being built in Japan..
PIYO
@C_of_CreativityOct 15
とりあえず完成した。RTX PRO 6000 Blackwell Max-Q x4スーパーワークステーション。まずはこれでテストしてあとで外付けでWorkStation Editionを繋げ5GPUにしたいと思っている。
1
53
After using it for a few weeks, in my experience: @Grok 4.7 is more capable than 4.6, but it writes a lot like Opus 5. Immensely frustrating, especially as I really liked the clear and concise writing of Grok 4.6 and 4.5.
1
1
68
>New benchmark drops >A GLM model does exceptionally well for it’s size @Zai_org is just built different
Surge AI
@HelloSurgeAIOct 8
Introducing sudo L7: Can Coding Agents Act Like Staff Engineers? Coding agents are getting very good at solving well-defined problems. But staff engineers do much more than write correct code. They understand unfamiliar systems, make architectural decisions, spot risks the ticket never mentioned, and know when passing the tests isn't enough. sudo L7 measures that judgment. Our new software engineering benchmark contains 60 expert-authored tasks, primarily in private production codebases from real companies, complete with legacy code, technical debt, and messy dependencies. We evaluate eight dimensions of engineering behavior, from functional correctness and engineering craft to practical judgment, thought partnership, and knowing when to push back. The best-performing agents score under ~45%. Coding is only part of the job. See you next cycle. Explore sudo L7: surgehq.ai/blog/sudo-l7
1
39
Using @MistralAI Large 4 for EU legal stuff is actually genuinely good, I'm impressed.
2
16
Anyone know if the v7 kernel issues for DGX Spark are fixed and if it is safe to upgrade?
2
96
Got curious after seeing the Micro Centers in the US running out of DGX Sparks, I can find and buy at least 120 shown in stock in just Dutch stores. Really strange how Europe seems to have so much more Sparks available than the US. They are around 7.5k USD incl. VAT though.
1
2
113
Ok Opus 5.5 makes me want to get a Claude sub again, but I just *know* that in 2-3 weeks it's going to be quanted to uselessness again.
2
80
Replying to @X
@X I'd really like to be able to use Grok to sort my X bookmarks into subfolders by category, but apparently it doesn't have access 😭
1
77
Ok apparently @bot can do this just fine! Would love to be able to use this in the X web app though.
122
Gave my buddy access to the sparks, his greed sickens me 😭
2
4
104
I like @posthog but its AI is genuinely unusable at times, it seems completely oblivious to how the product actually works, making suggestions that aren't present in the UI.. Guess moving fast has its downsides 😬
1
106
Man I’m seeing RTX 5090s listed at 12.000 EUR (14k USD) by first-party retailers in the Netherlands, this is getting insane
1
1
87
OK I kinda like @Muse Spark 1.3. The Meta team cooked I'm afraid
71
People called these NVMe-on-GPU designs by ASUS stupid btw, seems to me they are actually perfect for the new n-gram models. Maybe someone will be crazy enough to make something like this again? (If there even is manufacturing capacity)
1
74
So grok just spat out it's reasoning summarizer prompt in my grok build lmao
3
62
DeepSeek V4.1 Flash might fix on 2x Spark at NVFP4/EXL3 if you offload the new engram / n-gram lookup table to disc? Don’t know at what context though.. curious to see what people like @MiaAI_lab come up with
1
108
Paseo is great for getting a nice UI and easy remote control for Pi + other harnesses btw, really enjoy using it. Open source too!
0xVRAM
@0xVRAMAug 31
GLM 5.3 Flash on Medium Thinking with Pi is genuinely so good. Only suboptimal part is that it's noticeably slower than other options running on 2x DGX Spark. Great work @Zai_org!
113
GLM 5.3 Flash on Medium Thinking with Pi is genuinely so good. Only suboptimal part is that it's noticeably slower than other options running on 2x DGX Spark. Great work @Zai_org!
1
1
2
196
Load more