sinuhe69

4 ms·

31.
by sinuhe69 2mo ago
I think the central question for a such memory system is whether we or the agents can find the relevant information and how to organize these data as changes continues to come in. Would we miss something in the retrieval process? How do we
32.
by sinuhe69 2mo ago
Any implication for the privacy and confidentiality of our queries? Openrouter yes sees everything. But as long as they are independent and establishing their core business on privacy and confidentiality reputation, I feel more comfortable
33.
by sinuhe69 2mo ago
Very exciting experiment! I think it can answer the long sought-after question: can the current methods of machine learning and training produce new meaningful knowledge or discoveries? It’s a good test bed because the curated curriculum is
34.
by sinuhe69 2mo ago
Not just domestic. They come from other Asian countries as well. Asia is huge with billions of people and their time zones are not so much different from each other.
35.
by sinuhe69 2mo ago
I read the same. Refreshingly honest, straightforward and many useful information included. It is a breath of fresh air.
36.
by sinuhe69 2mo ago
Now GLM 5.3! And they explicitly confirms my theory: "Scaling post-training is all we did for GLM-5.3. With GLM-5.2 we built the stack: IndexShare for efficient long-context processing, SAO for RL on long-horizon tasks, and slime for l
37.
by sinuhe69 2mo ago
The release of Gemini Flash 3.7 just 3 weeks after 3.6 confirms my theory, IMO. Only post-training refinement and reinforcement learning (RL) trajectory optimization could yield such high improvements using the same baseline pre-trained mod
38.
by sinuhe69 2mo ago
Having skimmed through the article, I assumed that they hadn't tested the design for mobile devices because it wasn't mentioned. That would be a huge mistake. In today's web design, a mobile-first approach is imperative. This
39.
by sinuhe69 2mo ago
My theory is that it all boils down to better data and longer post-training period. Cursor got curated data from the trillions reactions of real world developers in real jobs. xAI bought is and used it for its post-training and got Grok 4.5
40.
by sinuhe69 2mo ago
I don’t know about you but I find the information about prices, effective price (weighted average), providers and performance, benchmarks (down bottom) very useful. With openrouter I can even test it right away and compare with other models
41.
by sinuhe69 2mo ago
Well, one reason is that we always have to work with the quirks of each model. So, a know model is often preferred over a new/unknown one because we have to be vigilant again. (Negative) surprises are mentally exhausting in the long ru
42.
by sinuhe69 2mo ago
TBH, it sounds more like robbery to me. That is dictated by the majority doesn’t make it right.
43.
by sinuhe69 2mo ago
My read is that AGI is a long shoot and nobody expects it in the next 10 years. They will push for it as always. But the cost of AI is exploding and requires a strong hand and a pragmatic approach if you don’t want to go down with it once t
44.
by sinuhe69 2mo ago
Yeah, I think Pi is good to use now because the AI has became so good that they almost can run on their own. Previously, AI agents do so many dumb and expected things that I really didn’t want to let them run on a loose leash and manual+aut
45.
by sinuhe69 2mo ago
I wonder why nobody hasn't created a SKILL file for Bonsai yet. I thought every library today would be introduced with a SKILL file in bundle.
46.
by sinuhe69 2mo ago
When has NASA got even close to that number of dollar per kilo payload in its long history? AFAIK, never.
47.
by sinuhe69 2mo ago
It will be a very easy solution when one takes a less bureaucratic stance and be less hostile to foreign cultures: just mix the sauce with sea water and (really) slowly release it into the ocean. If Japan can release radioactive water into
48.
by sinuhe69 2mo ago
It was a good read and a heart warming one. Congratulation to your Einbürgerung! Others have talked about many aspects of the life annd integration process in Germany. But one thing I’m missing from your post is what did you do to your Germ
49.
by sinuhe69 2mo ago
Some deleted it (again?). But the comments of Liang could be found somewhere else. [1] https://aiproem.substack.com/p/must-read-deepseek-liang-wenf...
50.
by sinuhe69 2mo ago
My understanding after reading Liang’s comments during the investment meeting is that Liang firmly believes in AGI and he bets everything to reach goal. Once it reaches AGI, the game would flip totally. How he didn’t paint it out, and with
51.
by sinuhe69 3mo ago
But here it shows another problem: who controls the distribution channel, they also control which program can be found and installed (by the mass) and which not. Without a critical mass of adoption, mesh applications are largely useless.
52.
by sinuhe69 3mo ago
Best part: “Oracle petitioned the court against the requirement, arguing that these financing costs would deter future investment in the state, while emphasizing its commitment to the project, […]. Regulators, on the other hand, remained fi
53.
by sinuhe69 3mo ago
I believe that living a happy life is a long term survival strategy. Producing offsprings is the other way. If we feel miserable, we want to cut our life short or it will be due to many problems, bad health included. Long term happiness is
54.
by sinuhe69 3mo ago
“You have reached your limits on help.roblox.com” That is the first time I see something like that!
55.
by sinuhe69 3mo ago
I have a problem with the cost per task metrics of Artificial Analysis. We don’t know how they calculate it exactly. But recently, cost per task has become the most discussed topic. The logic is basically: if model A achieves 55% on benchma
56.
by sinuhe69 3mo ago
In this book, Sanjoy Mahajan shows us that the way to master complexity is through insight rather than precision. Precision can overwhelm us with information, whereas insight connects seemingly disparate pieces of information into a simple
57.
by sinuhe69 3mo ago
With these heatwaves, many cities in France, Italy and England are at times *hotter* than cities in the tropical SEA! No wonder the bananas are starting to fruit.
58.
by sinuhe69 3mo ago
The title is misleading. The link led to a pricing page/token plan and not about the new QWen 3.8 model.
59.
by sinuhe69 3mo ago
Nope! GPT just cheated (by search the web). If you changed the text, export as image, GPT 5.6 Sol will fail. I tested with other text and even with a hint of a hidden text underneath, GPT 5.6 Sol could not see it.
60.
by sinuhe69 3mo ago
If they are not a frontier-lab, they would not need to submit their models for safety test before release. At least that is the proposal.

More ›