HNHacker News
TopNewBestAskShowJobs

dudisubekti

274 karma · joined December 30, 2023

submissionscomments
dudisubekti··on DeepSeek-v4-flash-vision-exp
Going straight to surveillance and unable to think "nicer" ways... is strange.

1. process graphs and charts

2. process handwritten math formula, also chinese characters writings

3. process design sketch and wireframe

4. process scanned documents

... etc

in fact these transformer models currently suck for surveillance, too slow and expensive. There are already faster and better facial/gait/object recognition models out there.

dudisubekti··on It is a sign of the times that Amazon gets to call this fair use
Culture of contempt is the name of the game: being outraged all the time is tiring, and tired people are easier to control.

https://en.wikipedia.org/wiki/Outrage_industrial_complex

dudisubekti··on Abdominal fat predicts heart disease risk better than BMI
I mean yeah, of course. We don't need as much as fiber as gorilla.

My point is just modern diet is severely lacking in fiber, when our gut evolved to process them.

It's not a surprise this evolutionary mismatch will cause so many problems in our body.

dudisubekti··on Abdominal fat predicts heart disease risk better than BMI
Fair but resistant starch that we're speaking of behave more like soluble fiber:

https://nutritionsource.hsph.harvard.edu/carbohydrates/fiber...

> Resistant starch – Soluble fermentable fiber found in legumes, unripe bananas, cooked and cooled pasta, and potatoes that acts as a prebiotic. Adds bulk to stools but has minimal laxative effect. May help to normalize blood sugar and cholesterol levels.

For stronger studies, these guys have you covered: https://www.ncbi.nlm.nih.gov/books/NBK559033/

dudisubekti··on Abdominal fat predicts heart disease risk better than BMI
Good thing we're not just assuming, we're actually studying it too: https://www.ncbi.nlm.nih.gov/books/NBK559033/

Also savannah human diet is chock full of fiber too: baobab pulp, grass seeds, tuber plants, etc

dudisubekti··on Abdominal fat predicts heart disease risk better than BMI
I hope you're not being serious, but they're not "supplements", they're food. You can easily get them from fruits and vegetables.

While they're not essential in short-term survival, they're essential for long-term health:

https://www.ncbi.nlm.nih.gov/books/NBK559033/

We have enough peer-reviewed studies on this.

dudisubekti··on Abdominal fat predicts heart disease risk better than BMI
> not generalizable

But there's already a general advise on this: eat more fiber.

https://med.stanford.edu/news/all-news/2017/08/hunter-gather...

We're a great ape, our body evolved to process a lot of fiber from fruits and starches. It's the modern fiber-deficient diet that's really weird.

90g/day of fiber is similar to the Hadza tribe diet, where they can consume 100-150g of fiber per day.

dudisubekti··on First human trials of designer protein therapies stun US neuroscientists
If it's any consolation, people tend to get their sh*t together after they're tired of everything around them failing.

China basically experienced this last century, collectively asking themselves "maybe we shouldn't suck this bad??"

dudisubekti··on First human trials of designer protein therapies stun US neuroscientists
> demoralizing

Why? China could do their own thing, and your country can do yours.

dudisubekti··on The Water Footprint of AI
I... don’t follow?

LLMs rely on many core AI concepts and techniques: neural networks, backpropagation, transformer inference, etc. An LLM is essentially an implementation of AI.

The term AI has been around since 1955, and has been used by the academics ever since then.

You might not like ChatGPT and co, but they're definitely AIs lol, at least in academic sense.

dudisubekti··on The Water Footprint of AI
lol no, if this is the case then we would have banned beef: a grossly inefficient protein source that takes absurd amount of energy and water, and produce significant greenhouse gas waste. We would also ban disposable cheap goods.

Yet we don't.

If people don't like AI, they should just say so, no need to skirt around.

dudisubekti··on DeepSeek-V4-Flash Update
Gpt 5.6 Luna cache read is $0.02 per mtok

V4 flash cache read is $0.0028 per mtok

That's not "a bit cheaper", just saying

dudisubekti··on IRGC claims it destroyed Amazon's Bahrain data center
"The strong do what they can and the weak suffer what they must."

-Thucydides' Melian Dialogue

Given how clownish the US is handling this war, I'm afraid you just have to get used to it.

dudisubekti··on Clarity didn't work, trying mysterianism (2012)
He was in Haiti for one week.

https://web.archive.org/web/20150407223525/http://squid314.l...

Then went back to a society safeguarded by the "earrings" (OSHA, FDA, CDC, etc). Poor countries usually don't have the resources and enough HDI to enforce public safety.

Not saying his experience doesnt count, but 1 week is not enough to experience that crushing learned helplessness when everyone around you keep doing stupid sh*t and sabotaging themselves and others near them.

Which is why Trump defunding public safety agencies is baffling to me, a person from a poor country. Such a decadent behavior.

dudisubekti··on Clarity didn't work, trying mysterianism (2012)
Eh I think the writer should visit poor countries and live among locals. Mistakes and poor judgements compound, and it leads to horrific living conditions (malnutrition, nonexistent hygiene, quack medicines, short-sighted violence, etc)

In fact developed countries are built on top of these "earrings": it's called math (and by extension science in general). If the math is right, it's always right, and it has spared their lucky citizens a LOT of needless misery and wheel-reinvention.

"Thinking for yourself" is overrated if you don't have good framework. See antivaxxers and their "research"

dudisubekti··on Xiaomi-Robotics-1
Yeah, western lecture on human rights carry no weight these days lol (Israel support, Trump doing Trump thing, war in Iran, etc). They ruined their own reputation so the rest of the world just tuned out.
dudisubekti··on Xiaomi-Robotics-1
Speaking as a relatively unprivileged kid from Southeast Asia, my parents were so busy working they just didn't have time and energy to care about me, let alone how well i fold my laundry.

But that's just one data point.

dudisubekti··on Xiaomi-Robotics-1
What?

There's already a better system today: mine launchers (remote minelaying system)

dudisubekti··on Japan develops a method to recover up to 90% of lithium from used EV batteries
> last laugh

Can I ask your reasoning?

Currently hydrogen is just oil with extra steps. Efficient electrolysis either needs ultra-rare materials like iridium and platinum, or exotic ceramics for continuous high-temperature electrolysis.

I personally can't see how this arrangement can supplant oil and batteries.

dudisubekti··on Show HN: Jacquard, a programming language for AI-written, human-reviewed code
Doesn't stop people from mass-producing and consuming beef lol, people just don't care aside from performative lip-service.
dudisubekti··on GLM-5.2 is a step change for open agents
You don't seem to like the "CCP" and their political views, but why are you using their sponsored models?

Why don't you exclusively host and use the open-weight western models, even if right now they don't perform as well?

I'd like to know the psychology behind this, because your actions feel contradictory to me.

dudisubekti··on Qwen-AgentWorld: Language World Models for General Agents
Ah I see. Yeah the graphics are probably AI-generated, and AIs do struggle with unit consistency in charts.

(For another example, the charts in the August 2025 GPT-5 presentation)

dudisubekti··on Qwen-AgentWorld: Language World Models for General Agents
This label?

> Figure 1: Overview of Qwen-AgentWorld. Top: Qwen-AgentWorld is a unified native language world model across seven domains. Bottom: We explore two complementary strategies for applying world modeling to enhance language agents (mainly using the 35B-A3B model as agent): Decouple and Unify , where the world model serves as the environment simulator and agent foundation model, respectively.

Where is the mistake?

dudisubekti··on W Social, public institutions and the theater of European digital sovereignty
Yeah money is just a fancy paper if you can't translate it to actual productivity.
dudisubekti··on DeepSeek Introduces Vision
I feel like '/s' has ruined irony on the internet. Irony is at its best if left ambiguous, lol.
dudisubekti··on GLM 5.2 Is Out
Crazy how people make light of this, when you can see the alternative today: India.

Sorry Indians reading this for throwing shade at India, but I just want to point out that making 1 billion of people not poor is freaking hard.

dudisubekti··on AI coding at home without going broke
Other than Taiwan declaring independence, I don't see any reason why China will rush to take the island.

At the very least they would wait until they cracked EUV and mass-produce the chips, and that is still 4-5 years away at the earliest.

dudisubekti··on If Claude Fable stops helping you, you'll never know
I guess... but I think, at its core, a good coding harness usually includes:

- well-crafted system prompt that follows best practices

- good contextual reminder prompts (when an llm got stuck in an infinite loop and times out, forgets how to use tools, or needs recurring best practice reminders, etc)

- well-written ergonomic tools the llm can use (read/write files, read diffs, browse the internet, etc)

I dont think these are anything special. The deepest moat I can think of is, proprietary models can be specifically trained to use their proprietary harnesses, so they are more token-efficient and make less tool call and file editing mistakes.

However in my experience, I'm as comfortable working with my own homemade harness as with Claude Code, so I don't think it's a deep moat...

dudisubekti··on If Claude Fable stops helping you, you'll never know
But harness is relatively easy to code yourself?

They're just system prompt composer, with some tool functions that the LLM can invoke. I've vibe coded my own in just one day.

dudisubekti··on Show HN: Gitdot – A better GitHub. Open-source, written in Rust
I mean, yeah... but I bet if Zig gets really popular, it will attract similarly toxic weirdoes.

Also if we're only judging whole communities by their toxic minorities, then... I'm so sorry Americans reading this, I've got some bad news for you, lol

What I'm trying to say is, it's best to try <a thing> for yourself, and see if it fits your preferences, rather than letting others stop you from exploring.

← PreviousPage 2 of 4Next →