muzani

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.

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.

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.

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.

What always worked for me is something like, "Hey, I read your book!" or "I just heard your interview on the radio!"

It's especially effective if it's a timely event. People do talks and interviews to meet people and get some attention, so you're basically shaking hands with someone who's ready to shake yours.

Anecdotally, I find Composer 2.5 to be useless. I do use light LLMs like Claude Haiku and some of Cursor's older free models, but Composer is negative productivity for me.

Yes, the only way to get reliable input/output is hosting it yourself

Extremely valuable. It's the best way to transfer knowledge among engineers and control damage.

It's more vital in bigger teams and teams that have something in production, used by a lot of people. A bad push breaks things on multiple levels, and the damage is multiplied by number of users. One person making two build breaking mistakes a year in a team of about 25 would break the build every week - CI/CD and auto review catches some of this, but PRs are good at catching the obvious.

Plus regressions. I've been commenting that "this line prevents the logo printing bug for PNG in HN-1345". Often people don't know why the line is there and why it looks funky. But the new era hates comments. So we're reliant on humans who have faced bug HN-1345 to remember. In some cases, they've faced similar bugs in other jobs.

I think most places overdo it though. 80-20 rule. Less than 20% of time should be spent on this. Most should be automated.

I just check that it works. That it does what it should. That there's no gotchas ("triBeep() rings at 3 frequencies, not three times"). That it's not an architectural problem ("we don't put strings in the async because it leads to a memory leak and/or makes translations harder in the future")

There's some quality control, but it often involves the above - don't name functions triBeep() and so on. Code style should be controlled by the linter.

I never comment on whether it should be onChange vs onChanged, this is all bikeshedding and doesn't prevent bugs or increase productivity.

In the last 12 months?

Antigravity started the workflow where you give a list of things to do and it goes off and does those things without supervision, including drafting and testing edge cases. It can even spin up images. Fable is the latest form of this workflow.

Gemini 3 Pro is actually at a mid designer level. None of the others are even at a junior human design level.

Sonnet 4.5 was, for a brief moment, creative brilliance. But we're talking coding, not writing right? They ditched it all for 4.6.

Opus 4.6 and 4.8 are extremely good for coding. I use them to reliably go through logs that are like 15k lines long. It can read my code, plot out the logs that should happen, check the logs for what actually happens, and from there, form hypotheses, and set up the logs needed to validate these.

Codex/ChatGPT is probably second best in all of the above.

They're different beasts. It's like comparing a donkey to a steam train.

GLM 5.2 can run for hours. Fable was supposedly designed for something that takes a human up to weeks. It's also more for straightforward tasks, so it'll do poorly in tasks that other models are great at.

Anthropic is a company which claims that the biggest threat AI has is inequality. This was communicated from the very beginning, before Claude and other public models.

OpenAI has also held the same stance in the past, hence the original plan to make everything "open".

Now ChatGPT and the $20/month ($10/month in developing countries) are a good way to make these tools accessible. Anthropic did this. This was fine. It hurts the poor to pay, but they can still pay for it.

But the new proposal to remove subscriptions is worrying.

Fable itself is worrying. It was terribly communicated to the media. It seems more like a model below Opus which is designed to run for several days straight, effectively replacing human oversight.

It's not cheap because it's not designed to be cheap. It's built to be cheaper than humans.

My experience with these places is they don't improve because they don't have room to, and they don't have room because they're too busy being inefficient.

For example, one place I worked used emails as source control. They'd email what was done at the end of the day to the manager. The emails had a limit of ~100 MB. So, emails bounced.

They used emails for all communication too. Lots of people were always busy all the time, working overtime, and completely ignoring anything through email. I complained to the CEO that people were just ignoring emails, and he scolded me and told me that when working, you're supposed to always CC their manager and the CEO, otherwise they won't do it.

Since the PM was spammed with code, all emails to her bounced and she didn't know what was going on. It's possible this was deliberate.

I tried to get them on (free) Slack and git. They thought it was nice, but never had time to actually adopt it.

Project management was done through Excel. One manager opening a single excel sheet on his laptop. Every day, they opened it and went through every item and asked if it was done and when it would be done. After meetings, he'd have nothing to do, so he'd talk to the devs asking if they were done yet, or random questions like, "What is DevOps?"

I talked to the CEO about improving hiring. He told me they usually just have one applicant who fits the salary range and had the qualifications. The rest of the applicants probably couldn't unzip a file.

Kids grow up being lied to about everything. Back when I was a kid, my mom would threaten to call Batman if we were misbehaving. Today, there's TikTok videos we "call" where a "policeman" picks up and asks for the details of the misbehaving child.

After 40, people become gullible again. FB is full of people who think an actual Ghibli clip is AI generated.

No offense, but it's these people who can't tell the difference who are most afraid of it.

The kids, lied to all their lives develop a kind of immunity to it. I can easily tell a poorly photoshopped picture, because my brain can recognize inconsistent shading and the repeated background effect when something is erased.

Kids these days will spot AI in 10 seconds. Right now, it's the composition. Even if perfect, AI will enhance it in a certain way and focus from some angles, and there's particular art styles. There's some collector card variants that people hate because it feels AI, even though it's not AI, likely done from tracing/redrawing something AI generated.

Alternate theory: It's a best seller because it's so goofy. Much like brainrot and Dr Seuss.

Csikszentmihalyi did research around what gives people joy. From that, he introduced the concept of Flow in 1975, which is often misunderstood.

Flow is my definition of joy. You do something that needs intense focus. The focus makes you lose sense of self and time. You're too focused to be self-conscious. You don't overthink. You move fluidly.

I don't have fixed hobbies. I did boxing for half the year just because I love the feel of redirecting energy from my toes to my fists. But something like climbing or basketball works too.

Work can give flow too, it's just programming and writing the way we're usually taught is quite bad for it. One hack is to write at the speed of thought and then edit it afterwards, instead of doing block by block.

Meditation makes everything more pleasurable because you learn to silence all the internal voices and ambient thoughts. You learn to empty your mind and other emotions. You focus on the present. And once you can do this, you can apply the same to work or other less joyful activities.

As for purpose, it's the long term, the infinite game as described by Carse. Not just the afterlife, but what lies beyond death - family, community, country, humanity, ethics. Purpose has been solved many centuries ago, just RTFM.

Gmail product managers are scratching their heads reading this.

Yeah, and as much as people complain about Claude restrictions, it's a lot more generous than Antigravity and more transparent.

Claude seems to do perfectly fine with monorepos and navigates them better than humans. GPT will get lost and repair the wrong code. Most tools will also look at what you worked on recently.

If it's difficult for humans, then that should be improved.

It's practically a benchmark now. Some friends have been specifically training models to count the R's in "strawberry"

Typical interview types we see:

Make this thing that would be impossible without AI. The test is to see if you actually architect it properly and understand principles of how things connect together.

Make this thing that would be impossible without AI. Now make these modifications without any AI.

Make this thing. You may use low quality AI like Composer 3 or none at all, but if you use none, we'll probably think of you as some kind of boomer.

Here's a bunch of technical problems that we don't know the answer to. If you give answers or insights we haven't considered, then you're bringing value to the team (e.g. git/PR policy, microservices, feature flagging, localization, security)

Both should be flagged and violate the guidelines.

"Don't post generated comments or AI-edited comments. HN is for conversation between humans.

If a story is spam or off-topic, flag it. Don't feed egregious comments by replying; flag them instead. If you flag, please don't also comment that you did."

I'm about the age where I need a walking stick and a cyborg arm to keep up with all these leetcode artists. AI couldn't come at a better time.

This jumped out at me right away too. What happened to the days when a dedicated manager would manage 8 reports? What now? AI is going to double the communication bandwidth with these reports and further double the free time they have?

I do think the most efficient form of team is a "cell" of three people. One is a little unstable.

We can go back to web rings. Web rings were great.

Capitalism and the free market wasn't built for our economy. It was built for scarcity, not abundance. The idea was that competition would lower prices to the level of cost. This works for the price of iron or textiles, but since software, we've seen prices set to even lower than costs (pirated DVDs, open source).

It broke a long time ago, back in the modern industrial-agricultural revolution. People say communism is worse, but the problem is that communism was a response to capitalism. When communist revolutions trigger, then the higher ups decide that maybe all we need is a patch instead of a rewrite, and start tightening laws. But these are just patches. Things like ad-driven models and consulting on open source are patches as well.

Capitalism will continue to break every time there is an industrial revolution. Robots took jobs but added a few, computers added plenty of jobs, but with AI we might not be so lucky.

For impact most likely. Dynamic pricing is core to the budget airline industry and such a law would hurt them more, especially with the thin margins. It happens with games too, but the price of games doesn't affect how someone eats.

It's a relative's birthday too, and we always joke that Labor Day is about celebrating her mom.

Maybe related to this? https://news.ycombinator.com/item?id=47906253

My and my partner made an app where you stare at yourself in the mirror for 5-30 minutes. You'd record how you feel after.

It's not "scientifically proven" to work but it's quantifiable that there's a clear improvement. We tried to scientifically prove it - the scientists were incredibly offended, lol.

So it can't be proven, but there's data. We started quantifying it because corporations asked for data. Nobody wants to buy an app where you record yourself staring at a mirror, so we don't charge for that. We end up charging for the data reporting, anonymization, data handling, etc.

There's some interesting data, like we find people who are highly stressed will get all kinds of diarrhea, vomiting, and angry if they crank it up to 30 minutes. So we designed it to start at 5 min/day for a week, then 10, then 15, slowly up to 30.

Staring at a mirror is a little different to staring at a wall, but I'd imagine the results are similar.

Famine and disease always follows wars. The more wars, the more famines. It's not just supply chains. Very often things are delayed - shortage of fertilizer will impact things a year later.