I love that y'all are keeping this tradition going for her. Cooking for that many people is a whole lot of work
After my grandma passed away 5 months ago, my grandpa’s two brothers and one cousin’s family have been coming to our home to celebrate every festival together.
We’ll continue this tradition until grandma’s first death anniversary.
That’s why we have to cook for almost 40–50 people during every festival!
So, we cook outside in an open space using a traditional chulha (wood-fired stove).
I'd throw some game code at it first and see if it catches the bugs I miss
My DeepSeek-V4.1-Flash-1017 quant is NOW LIVE for TP=2 DGX Sparks.
Same 353 questions for all three packs:
new / Mia 2.9 / my old q28-v2
MMLU-Pro STEM: 93 / 91 / 90 (of 100)
LiveCodeBench: 14 / 11 / 11 (of 17)
MBPP+: 50 / 49 / 50 (of 60)
HumanEval+: 34 / 34 / 33 (of 40)
SciCode: 6 / 6 / 5 (of 10)
MATH-500: 36 / 36 / 36 (of 40)
MGSM: 40 / 40 / 40
BFCL tool calls: 39 / 39 / 39 (of 40)
Multi-step tool chains: 6/6, 6/6, 4/6
These are small samples, so I read it as "at least as good, in a smaller pack". It's not a blowout.
The gaps I care about are LiveCodeBench and the tool chains, which my old quant kept dropping. Calibrated it for coding, science and agent work.
Weights/Recipe are in the comments. Recipe has been updated with my current config but is mostly unchanged.
@MiaAI_lab I still don't know how you get them so good. I can't wait to use your Engine Aware Quant (EAQ) system when it's ready.
2.8 BPW, EXL3, 203.8 GB, running on two DGX Sparks with @ashxhart TensorFold.
- 1M context
- image input
- 2.1M-token cache total with NVMe offload.
As usual, if you have any issues, please let me know or submit a PR.
What will you throw at it first?
I get impatient when the tops of the biscuits start browning and the middle still needs a few minutes
That's pretty cool, having AI agents on your wrist and approving stuff with a button
⌚ Vibe Watch — AI Agent Control at Your Fingertips 🤖
Made by @GOROman with the #M5Stack StopWatch, Vibe Watch puts multiple AI coding agents right on your wrist. ✨
👀 Check the status of six agents at a glance.
✅ Approve or reject actions with physical buttons, sound, and haptic feedback.
🎙 Toggle Plan mode or use push-to-talk for your next prompt.
🏆 Best AI Application award winner - M5Stack Innovation Contest 2026
👉 Explore the project on Hackster.io: bit.ly/4i22WcA
#M5Stack #VibeWatch #AIAgents #VibeCoding #WearableTech #ESP32 #M5StackInnovationContest
Frhanck Da Tahnk retweeted
Trying some soup brands, seeing as most now contain folic acid.
I bought a carton of Covent Garden vegetable soup & I have just tried it.
I wish I had looked at the "Trustpilot" reviews before I bought it (1.6 BAD!)
Remember the movie "RoboCop" & the scene where they feed Robo some gloopy, baby food?
Well, that was the soup I have just had.
Tasteless, gloopy baby food!
Absolutely will NOT be buying again!
Even "Bear" my Newfie, just sniffed it & walked off & he usually loves soup.
There was something suspicious about the ingredient label too, as it just listed the vegetables it contained. NO thickeners & NO preservatives. Almost all commercial soups contain both.
Reading some of the reviews, some people even said their soup tasted of chemicals!
Anyway. It looks like I will have to make my own soup from now on, too.
These coding agents are getting scary good, I need to keep up with this stuff
Cheking out REA (Reverse Engineer Anything) - the #1 Github repository right now.
It enables coding agents to look at a finished app or software application, figure out exactly how it works without looking at source code, and help you replicate it. github.com/morluto/rea
I don't need AI doing every little thing for me. I still like making things with my own hands.
The AI maximalists say that we need AI to do all the jobs so that we can be freed to do more meaningful things like make art and read to our children. But then they also want AI to make art and read to our children. So what exactly will be left for humans to do? The answer, though they won’t admit it, is that we get to be passive consumers sitting around slack jawed staring at screens until we die. That’s the only form of happiness these people can conceive of. It’s all any of us should want, as far as they’re concerned.
I love cooking at home, but these grocery prices are getting ridiculous 😬
MY FRIEND: “Why don’t you cook more?”
ME: “I do.”
FRIEND: “Then groceries shouldn’t be that expensive.”
ME: “Chicken was $17.”
FRIEND: “Okay.”
ME: “Eggs were $6.”
FRIEND: “Okay.”
ME: “Olive oil was $14.”
FRIEND: “Okay.”
ME: “And I haven’t even made dinner yet.”
Apparently cooking at home is still cheap if you remember prices from 2018.
Man I hate that AI has people feeling this way. I'd still be excited to see what you make
I'd probably end up reading about stuff I never searched for. AI summaries are interesting
Timely Saturday morning coffee read: AI summaries increase both the topical diversity of information consumed as well as the amount of information consumed for news content.
Researchers at UT Dallas, Boston University, and elsewhere conducted a field study of roughly 38,000 unique users at the Washington Post, where 71% were exposed to the normal search experience and 29% were provided with AI summaries in their search results. The authors found that the treatment group saw an effective topic breadth — the distribution of topics to which it was exposed — 21% larger than the control. In addition, the treatment group conducted 29% more searches and opened 5% more articles.
One methodological quirk about the study is that mere exposure to the AI summary contributed to measured topic consumption, so an AI summary that draws from five articles spanning five different topics would have greater topical breadth / entropy than a single opened article focused on one topic. But the results make intuitive sense: an AI summary can draw from a more diverse pool of information, resulting in a less concentrated topic distribution.
Notably: the broader exposure with AI summaries also increased the information readers encountered in common, with measured consumption of widely shared topics rising 69% and readers’ topic distributions becoming more similar to one another as measured through within-group pairwise cosine similarity across topic consumption vectors. So the treatment group saw greater topical diversity at the individual level with more of a shared topical distribution at the group level.
The engagement results mirror what other researchers have found in studying Google’s AI Overviews: while overall time spent and queries may increase, click-through rates decline.
Article linked in comments.
That whole spread looks good 😋 I'd be asking for seconds before anybody even got to the serving spoon.
Good morning, X family! ☀️
On today’s episode of Cooking with Christine, dinner is served! 🍽️
Sean asked me for a serving spoon.
I helped. 😂
Have a beautiful and blessed day! ❤️
Sorry about the language but this is for @LFleezy23
#CookingWithChristine
ETH leads the pack on that +217% run or it doesn't happen at all. Lock in.
my advice to my onchain brethren is to not forget that the cex alts also can give 5-10x returns, you all are gambling on lowcap shitters all day you might as well also sling some of your port on perps
& cex alts typically have more size buyers behind them when they do have a real fundamental narrative
uptrends can last several months during bull, some started about two weeks ago, best of luck
Replying to @danielsethlewis
@danielsethlewis #BhaskarMenon #ArnaudMassenet #HaritTalwar should be held accountable for what they did to $BETR






















