Hacker News Reader

muzanion Ask HN: What are you reading to your kids?

Aesop is great and all free https://read.gov/aesop/001.html

It has life lessons, just straightforward things like "don't brag unless you know you're worth it" or "just because they look nice doesn't mean they are". It's challenging enough for them and they'll ask a lot of questions, and they're able to even debate the moral of the story.

They're all really short. Easily 5 mins for today's TikTok attention span. It's an entry level into more interesting books.

muzanion Ask HN: Any nerds out there who've read a lot of research papers?

https://www.koreamed.org/SearchBasic.php?RID=2298820

First sentence draws attention. Second attention emphasizes that it was not a typo and makes more questions pop up in the mind of the reader. The following sentences increase curiosity on the methodology used and so on.

"PURPOSE: Penile size has been a focus of interest in the life and the culture throughout human history. It has been a longstanding question whether there is any relation between penile size and the dimensions of body extremities. We measured the standard length of the penis and investigate whether any body extremity is a predictive index of penile size in Korea men.

PATIENTS AND METHODS

A series of 655 males above 17 years of age were examined during a 4-month period. Stretched penile length, penile circumference, and length and characteristics of various body parts and features (1st finger, 2nd finger, 3rd finger, 1st toe, 2nd toe, 3rd toe, ear, mouth, nose, height, weight, and baldness) were investigated by one examiner. To see the relation among the penile length and circumference and various body dimensions, univariate and multivariate statistical methods such as correlation analysis and multiple linear regression were applied."

muzanion Ask HN: Does spec-driven development still pay off with frontier coding models?

Not the full PRDs like before.

Astra, Fable, Opus 5.0 are better with less instructions. Some of the context has been moved into the harness itself, so things like CLAUDE.md etc are becoming obsolete, and same with PRDs.

These days it's more about saying why than what. ADRs and GDDs would be helpful.

muzanion Ask HN: Is Opus 5.5 another step change?

Anthropic's naming convention seems to follow an S-curve. So Opus 5 is actually a little worse than 4.7, but by the time it hits 5.5, it's at the steep part of the new S-curve.

Similar with Claude 4 being worse than 3.7, but peaking at 4.5. At 4.6 they started to hit the restrictions and 4.7 seemed to patch what they made worse.

muzanion Prompting Claude Opus 5.5

That's how the cycles happen. China realizes they could use a little more freedom and US realizes that they could do with a little less. The emerging/shrinking middle class of both countries also moves the sweet spot.

muzanion Gemini 3.8 text-to-speech

I've used TTS in videos to replace the speech of the person talking so that we could type up whatever we need them to say in editing.

They sounded exactly like that person... but it's like they're angry at me.

Great toys but not production ready.

muzanion Ask HN: When is fine-tuning a small LLM worth it?

I did a project as few years back - it was people categorizing construction defects from a form into a drop down. A human could do this fine, and while there were instructions, it was not clear until you saw how others did it.

AI would still not solve this problem properly today, and may in fact be worse the more data it's trained on.

Another problem it does poorly today is writing in the style of another. It might be the whole copyright adherence but it refuses to follow certain styles. If you give it examples and ask it to break it down into a prompt on how to write, it will still do a poor job. It will work if you actually put examples in the prompt, but at this point, you might as well fine-tune it.

muzanion Ask HN: Do you offload your coding to AI, or keep your skills sharp?

You're there to make sure the code works. Refusing to let AI code is like a janitor refusing to use a vacuum. They're not taking your job any more than the robot vacuums.

AI opens up new forms of debugging. You can trace through a bug on the OS level and understand what it's really happening - we traced some bugs to how Android handles wifi based around location access on different versions. We give it a 20k line log to reproduce the exact flow when certain bugs happen. We make it write hypotheses on why it's happening and how to fix it, and break that down into logs.

There's plenty of engineering work to do with production code, especially on front end where you're dealing with different manufacturers, multiple cultures (some using , or . as decimal separators), RTL languages, broken abstraction layers, hardware permissions, etc, etc.

Someone needs to fully understand the big picture around how all these pieces go together. AI is still poor with writing larger things. Your uniqueness is in how quickly you can pick up the big technical picture.

muzanion I don't want to read what you didn't write

You can verify this by switching to an older model.

There's a reason ASD-STE100 is mentioned everywhere now. They're at the point where people can't understand what they're saying.

muzanion Ask HN: Dead Internet

I'm getting about one email per week, along the lines of, "I saw your project in so and so thread. Your project does X. You should know that it fails at Y. Would you like to look at our project, which does Y for you?"

Except that Y was never a problem and X was also not the goal. It's just doing some mental gymnastics to act like the product is relevant.

That's probably why everyone read it and didn't reply. It's just sending slop.

muzanion I don't want to read what you didn't write

Yeah, I think I'll start making my own creative writing benchmark. It's a little late to start data collecting, but better late than never.

muzanion I don't want to read what you didn't write

There's a lot of misunderstanding here. The quality of LLM writing has not plateaued, it has dropped significantly. They've cut it to the point where people start noticing.

Do a search on "Claude Sonnet 4.5" on Reddit and you'll see lots of disappointed users [1]

If I could give out ratings,

Average human with a degree: 5/10

Sonnet/Opus 5: 2/10

GPT 6 Astra, 5.4 Sol: 3/10

Sonnet 4.6: 8/10

Claude Sonnet 4.5: 9/10

GPT 4o: 6/10

GPT 4.5: 10/10

GPT 3 Davinci (with a lot of coaching): 7/10

[1] https://www.reddit.com/r/claudexplorers/comments/1ta6f9c/i_s...

muzanion Ask HN: What happens when Claude Code goes away?

I've moved to Codex lately, Astra does better with less bugs.

When GPT does poorly, I use Cursor. The autocomplete itself is extremely good.

And then there's Kimi code and similar which can't keep up with the existing demand, much less the demand if all other tools go down.

Then Opencode, which would probably be a year or so behind SOTA but it's good enough.

muzanion Can you tell which images are AI-generated?

"Made with ♥ by Codex"

muzanion AI-generated posters don’t have to be horrible

One trick that works from IRL is building a moodboard. Go to Pinterest, find what you like, paste it together.

If you're short on time, you can ask AI to summarize into a certain time and craft the prompt from there. But it's better to figure out what you really like.

ChatGPT tends to fall to the same things, but many of the other image generators do much better.

muzanion Ask HN: Who is using MCP in production?

I used one for an AI tool that allows people to report bugs/feature request directly.

It searches to make sure it isn't a duplicate, writes up the ticket, then submits it.

Because it's a production tool, we want the cheapest possible one without it being too inaccurate. If you used a API etc, you'd end up building what's effectively a MCP-like adapter on top of it anyway so it could communicate in natural language instead of dealing with JSON and such.

Linear's MCP is also very clean and well designed, probably one of their core advantages over, say, Jira. I wouldn't know what the API looks like because the MCP works great.

muzanion Half of Europe's towns and villages have fewer residents than 60 years ago

Something strange happens when people get wealthy and advanced enough. Cost of living rises, workload increases, it becomes unaffordable to have kids. 30 years later, everything is empty.

Places like Australia have plenty of land but, nah, everyone wants to live urban where housing prices are insane or inherited.

Food costs often go down in developed countries, cheaper than developing ones. Maybe they're so low that they can't afford to become farmers anymore? Not enough people buy corn and corn price is so cheap that it gets made into HFCS. But if you can't mass produce cheap corn, you're screwed and get a desk job.

So we have the service economy. Everyone sells services to one another, works from home, makes tons of money. Madrid has top-tier designers, chefs, doctors, moneylenders, nearly every kind of service. All the piles of money in a concentrated place like Madrid makes the whole place overpriced. If you happen to be the best tuna butcher in Western Europe, there's a limited number of places you can work, but many of these specialist jobs are also wfh.

Is there a hack to this though? Can our top tier Spanish pixel artist sell her services to someone in suburban Canada?

muzanion US and China race to be first to build Skynet. Either way we're F'ed

Winner: Skynet

muzanion Software development with AI is starting to feel like cooking steak

The resistance of the steak is going to depend on what cut of meat you are poking (sirloin is stiffer than filet), the grade of meat (prime is more tender than select), how thick it is (thick cuts will yield more than thin), the age of the steer (young is more tender), the breed of steer (cooked Wagyu is more tender than Holstein), the age of the meat (wet aged is more tender than fresh killed), and what the steer was fed (corn fed is usually more tender than grass fed), among other things. [1]

It's not entirely a myth; if you've poked hundreds of steaks by the same supplier, same cuts, same grade, you'll eventually get a feel for them.

The thermometer analogy does carry over though. Some people are not there to be productive or good at their jobs. The job is about them, not their work.

And you can go your whole life without ever cooking or tasting a good steak. The people who can tell the difference will go elsewhere, and those who can't will come back.

And then there's always the boss who wishes he started a franchise but will go with what he has - he'll hire a bunch of newbies or foreign workers who know how to use a thermometer.

[1] https://amazingribs.com/ratings-reviews/best-thermometers/ho...

muzanion Ask HN: Do you delete your browsing history?

I keep all of it, and go to incognito mode for anything I don't want saved. No, not porn. But things like searching for a book because it annoys the hell out of me when FB/Amazon/random blogs start recommending that book 11 times because I looked at it once on Amazon.

muzanion Ask HN: How do you figure out what you're passionate about?

Passion can be misleading; the literal definition of passion is something you'd suffer for. You can be passionate, productive, and deeply unhappy.

The nature of work also combines passionate with non-passionate stuff. We may love writing high quality code, but most of the day requires hacky low quality code, meetings, or digging through documentation. Or part of a pipeline of gathering, cleaning, and selling data to a financial institution that uses it for medium interest loans.

The better measurement is find out what work energizes you. You might not feel any flow or fun. You might not feel any emotion whatsoever. But you get out of it like a good workout - both tired and energized.

muzanion Judge approves $1.5B Anthropic settlement for pirated books used to train Claude

Yes, what would you like to happen?

I believe Anthropic shouldn't go bankrupt. I don't think the violated should be filthy rich either.

The richer publishers are still fighting. The ones taking the settlement may not have strong enough grounds and are happy with what they got.

It should be a speed bump. Just enough that it discourages blatantly breaking the law, and doesn't make it a strong incentive for others to resort to piracy as well.

If copyright law didn't exist, writers would still write but everyone would just take the books. A company like Amazon which doesn't give a damn about ethics would pirate all the books; the slap is painful enough to keep them straight.

If it were too tight, we'd have regulatory arbitrage; AI companies would set up in Japan, Singapore, India, China, and so on where they would get just a slap on the wrist.

Or if international laws were stricter, people would be pirating data on Elon Island or SpaceXAI Station. Forbidding something that valuable would be like Prohibition, lucrative for the criminals.

It's not really fair to anyone, and yet that doesn't mean it shouldn't exist.

muzanion Late.sh – a command-line Clubhouse for computer people

Which servers are still alive?

muzanion Judge approves $1.5B Anthropic settlement for pirated books used to train Claude

Odd place to bring this up. This is one of those situations where copyright is doing what it's meant to do.

muzanion US gas prices hit an average of $4 a gallon again

Much of the time, these estimates are just extrapolated. Someone out there must be producing more oil when there's a shortage.

muzanion Ask HN: Having a child before career has taken off

You have to do it sometime. Would you rather have kids in your 20s or your 40s?

The points me and my wife agreed on:

1) It takes a lot of time and energy to raise children, and there's more of both when young

2) Fertility drops sharply in 30s, and we could miss the window altogether. Also the odds of complications goes up

3) You can work as hard as you like after the kids grow up

4) Paying for college will be tougher, but my parents were able to do it

At 40, nearly everyone at my level of seniority has at least one child and they don't fault us for having them. We've all lost sleep over the baby, and we all understand that having a child younger than 2 is highly unproductive, but it gets better and most of the time they do a fine enough job that everyone forgets they have kids.

muzanion NYC may require landlords and realtors to disclose the use of AI in listings

Advertising would be fine if they checked it. Food advertising has used fake images forever, but it's okay because you know the meatballs don't look like that. You have an idea of what warm fries taste like without the fake heat imagery.

But the dress you get is not the same as the dress in the picture. If the model looked like you, the dress should fit the same, but the AI dresses don't. Same figure, same skin color, same height, and yet the dress looks different.

That's the problem with homes. There's no way a room fits that many things but AI will make it look like it does. There's a distortion where it changes the specification entirely.

It's like showing someone playing Fallout 4 on a MacBook Air. It's a deceptive practice, unlike the cereal boxes showing milk.

muzanion Where are YC founders now? OpenAI and Anthropic, mostly

A successful unicorn founder once said a billion dollar company is just made up of a bunch of people who could each have started a $10m company by themselves.

I guess you can extrapolate that and say a trillion dollar company is made of a bunch of people who could have started a billion dollar company.

muzanion Claude has the worst pricing – but people want it

It makes sense though. Fable is meant for hours or days of work.

If you're not using it that way, just use Opus, it's smarter and half the price. If you are using it for long work, then you should probably be using the $200 plan.

muzanion Claude has the worst pricing – but people want it

I'm convinced a lot of investors were rich minority kids bullied as a youth, looking to bully the rest of world in revenge.

You keep hearing things like, "The definition of a moat is being able to raise prices," and, "Your distribution must scale to 99-100% of the market and you should acquire the remaining ones."

$20 is expensive for me. Netflix and Spotify subscriptions are pretty bloody expensive. And they're saying raise the prices?

These people don't live in the real world, at least not in countries where 90% of people including the middle class don't have disposable income.

Like it or not, AI is going to be a requirement, like shoes, running water, a smartphone, a car, a degree, insulin.