Monday morning is the perfect time to set goals for the week ahead.
I like setting up progressive goals. Daily goals contributing to weekly goals contributing to monthly goals, etc.
Of course not all goals need to be progressive.
What are your goals for the week?
Unlock local, agentic workflows with Gemma 4 12B and Google AI Edge, directly on your laptop. Experience 100% on-device AI:
• Generate code in AI Edge Gallery (new to Mac)
• Dictate and edit text via AI Edge Eloquent (new to Mac)
• Serve Gemma 4 12B locally with LiteRT-LM
Dive in: https://t.co/gr7tOrZmc0
I've been quite happy with glm-4.7-flash-mlx-6bit for my local LLM, but I might have a new favorite. NVIDIA-Nemotron-3-Nano-30B-A3B-MLX-4bit is smaller, faster, and better at tool calling. @Zai_org the ball is back in your court. 😄
My granddad is the best person i know
At 11, he tried a cig and didn't like it and never smoked since
At 18, he was on the verge of death from sepsis when his mom's friend's husband, a long haul Aeroflot pilot, brought penicillin back from the US. When the antibiotics started working, the doc told his mom - don't cry, he might even live up to 40
At 21, he got a degree in nuclear physics, but wasn't allowed to work in the industry due to weakened health. He found himself in the Soviet space programme
At 26, he sent Sputnik to space, a few weeks before he had his first child
At 32, he made the discovery of the Earth's plasmasphere
At 60, he learnt English because the iron curtain had fallen and he could travel to the international space conferences. He needed to write and present in English
At 70, he would fight me for the dial-up internet as I wanted to chat to online friends, while he needed to send some work emails from home
At 85, he visited me in London and went to the British Museum five days in a row. One of the days we were having afternoon tea, and he exclaimed: "I'm so lucky! Had I not lived to this age, I would not have seen Amenkhotep III statue and had these wonderful scones at the Ritz"
At 90, he was the only person in my family who said I must absolutely take the opportunity to work in crypto
At 94, he still was still co-authoring scientific papers. And this hasn't stopped yet
Yesterday, he turned 95
Happy birthday granddad 😊
@MiniMax_AI Open weights != open source. Is it open source (like MIT or Apache)? If so, yea!
Open weights are fine too, but open source is even better.
@victoriawu77 No idea. I can't even decide on 'best' without some evaluation criteria, but, especially when zoomed out, I like B best (the bicycle frame is more complete and less stretched out, and the proportions of the pelican are the best for my eye). So I'm going B/B.
@LottoLabs What I'm still doing most of the time anyway, glm-4.7-flash. Oh, and sometimes Turnary-Bonsai-8B-mlx-2bit.
Options, options, options. It's a good thing.
I got sick in Hiroshima. Like really sick, fever and everything. Probably ate something bad or just caught a bug.
I went to a pharmacy looking for medicine. I was trying to explain my symptoms to the pharmacist with Google Translate, probably looking miserable.
She asked if I'd seen a doctor. I said no, I'm just a tourist, I don't know how.
She made a phone call, spoke fast Japanese, then told me to wait.
Twenty minutes later, her colleague came to cover her shift, and she walked me to a clinic. Like personally escorted me there, three blocks away.
At the clinic, she explained my situation to the receptionist. Waited with me until I was called in.
The doctor spoke decent English, examined me, said it was just a virus, and prescribed some medicine.
When I came out, the pharmacist was still there. She walked me back to a different pharmacy near the clinic, helped me get the prescription filled, explained how to take everything.
I kept thanking her, apologizing for taking her time. She said "it's okay, my coworker can handle the shop."
The whole thing took like two hours of her day.
I tried to offer her money for her help. She looked almost offended and said "no, no, I just want you to feel better."
Before she left, she wrote down her pharmacy's phone number and said "if medicine doesn't work, call here. We can help you."
The medicine worked. I felt better in two days.
I went back to her pharmacy before I left Hiroshima to thank her properly. Brought her a gift - some fancy cookies from the department store.
She accepted them but only after I insisted like five times.
She asked if I was feeling better, if I enjoyed Hiroshima despite being sick. I said yes, people like her made it memorable.
She seemed embarrassed by the compliment, just smiled and said "please enjoy the rest of your trip to Japan."
MiniMax M3 just dropped — their first natively multimodal model.
So I ran it through my form-filling test. (The model has to place each element at the right pixel position on a blank form image, not type into a field.)
Verdict: it got everything on the paper.
> Name, DOB, ID, gender, marital status, nationality, email, phone, address, postal code, all there.
> Best character spacing I've seen yet: it actually calculates the gap between each character, clean across the DOB and number boxes
> A few fields slightly misaligned, but every piece of data made it onto the form
The reasoning chain is the interesting part: it does the easy fields first, then works into the tight one-char-per-box fields, reasoning through y-coordinates, baselines, and label clearance in obsessive detail.
The cost: 40:33 and 126.7k output tokens. That's a long think — but it's MiniMax's first multimodal model, and it nailed the content.
M3 is here on Opencode
They have also updated their site
And it looks like m3 will have signifcantly less requests compared to m2.7
So a big jump in api pricing
Initial impression of MiniMax M3. It's good, and fast. Damn. $20 for the cheapest subscription, however, puts it at the same cost as OpenAI and Anthropic. Too expensive. More testing required to see if it's much better than DeepSeek and Xiaomi MiMo.
@kgonia7@cabrerajelias Xiaomi MiMo is close to the same as DeepSeek, and is a pretty solid model too. Hopefully the pressure will influence the other China developed models. 😀
What’s new with MiMo-V2.5 series inference?
We just published a blog on our full pipeline inference optimizations for MiMo-V2.5 series, including how we pushed hybrid SWA efficiency to the limit.
Read the full blog here:
https://t.co/lYBEcgaVhU
@rubenssoto_ai Yes. I use both, and they're both solid. I often ask for a detailed plan that an agent can easily implement from Gemini or GPT, and then use DeepSeek V4 or MiMo 2.5 Pro to implement. Then I have Gemini or GPT review. Solid results so far. Grok now also coming into the mix.