JoeCortopassi

Agreed. My frontend/react builds use esbuild, and rebuild fast enough that it feels like hot module reloading

The big money is going to the OpenAI/Anthropic types producing foundation models that have to raise billions on a regular basis. This is money that would normal be spread across the startup ecosystem instead of concentrated in a handful of massive companies. When it finally hits IPO, I'd bet that you see it start to get freed up for new investments

Just to drive the point home, in 2019 the total VC market was ~$300 billion. To date, roughly $235 billion is tied up in just OpenAI ($168b) and Anthropic ($67b)

I think you missed the parents point. They were saying they don't have the resources for proper defense, my guess is parent would agree about their offensive capabilities

This is some weird takes...

    - music apps and gapless playback: basically all major car manufacturer software is trash, I'd challenge you to find a car software that's better than Tesla
    - stretches without superchargers: to be clear, not stretches without places you can charge the car, right? Just places where Tesla's fast charging stations exist. Tesla has an order of magnitude more of these type of stations than any other similar network
    - Service centers: I have no knowledge here. From how silly the other arguments are, I kind of assume that this isn't fully true, but I'll give you the benefit of the doubt and say it's true. Not awesome, definitely an area they need to get better in. But lets not forget, major manufacturer service centers are not exactly know for being quick/good/cheap
    - Color options: I guess it would be cool to have truly custom colors, but basically every mainstream car ever has had a limited set of colors to choose from, and individual dealerships almost never have every color of that limited set available

Yeah, it really came from scratching my own itch in wanting a quicker/easier way to think through different companies offers. They always present them in slightly different ways

I really like the idea of a side by side comparison. I'll build that out next!

Just like writing normal software: have it work on different parts of the codebase that don't have overlap, or do the work sequentially. AI is a time compressor, but it's still constrained by the same problems as normal engineers.

If one task changes the button to green, and the other one changes it to red at the same time, something will have to decide how to reconcile that difference

I read this book because I misunderstood a recommendation and thought it was a business book talking about how companies get people to forget negative information (e.g. dump news on Friday etc). Boy was that first chapter a wild ride until I figured out it was not in fact a business book

Fun story, strong recommend

https://en.wikipedia.org/wiki/Paternoster_lift#Safety

Their overall rate of accidents is estimated as 30 times higher than conventional elevators

Germany saw an average of one death per year due to paternosters

...the new Gemini "replacement" is still by all accounts a disaster

Small nit: the initial rollout of Gemini was a dumpster fire. It is now quite good. For my use cases, I can't get any other current LLM to give me better results than Gemini 2.0 Flash. It's also free

But even that kind of proves your point, right? Pretend you 100% believe me without verifying. That means in a year or two the winner has transferred between ~3 companies. This is not a cheap mantle to keep passing around. The AI wars are going to get heated over the next year or two

You have to remember, HN is fully supported by YCombinator, and has a huge leg up on any possible similar sites. There will always be content thats encouraged to be posted here first from YC startups, and it will always be a valuable PR asset for those companies.

It doesn't have any pressures to make money, in fact it's basically a free place for PR and advertising jobs for portfolio companies

I've noticed that while a bunch of developers have played with LLM's for toy projects, few seem to have any actual experience taking it to prod in front of real users. I’ve personally had to do so for a few startups, and it's like trying to nail Jell-O to a tree. Every random thing you change, from prompts to models, yields massively different/unpredictable results.

I think because of this, a bunch of companies/tools have tried to hop in this space and promised the world, but often times people are best served by just hitting OpenAI/GPT directly, and jiggling the results until they get what they want. If you're not comfortable doing that, there are even companies that do that for you, so you can just focus on the prompt itself.

So that being said, help me understand why I should be adding this whole system/process to my workflow, versus just hitting OpenAI/Anthropic/Google directly?

I think the best way to understand/appreciate javascript as a language is to A) separate it from the DOM, and B) take a functional programming focus

Too many times, when people complain about javascript, what they are really complaining about is the hectic nature of web development. You have three trillion dollar companies (Google/Chrome, Microsoft/Edge, Apple/Safari) actively fighting over agreed upon standards for how the dom is rendered and manipulated. Try and learn any language in that environment, and you'll have a very frustrating time

Why functional programming? Javascript was one of the earlier mainstream languages to support passing around functions as first class citizens, and it's often an aspect of the language that gets downplayed. By and large, javascript sits at the sweet spot of allowing you to do cool functional programming concepts, without all the strictness and verbosity of a language like Haskell. Is that a good thing? It depends, obviously. But for the sake of "Change my mind", this is where I bet you'll start looking at it in a different light

I'm not saying it's the best/one-true language. It's just another cool tool to have on your tool belt

Hey HN, want a side project idea that's easy, and would be a huge hit?

Make a web crawler that hits publicly traded tech companies careers page, once a day, and tracks how often their job listings change. Make a big line chart at the top of a landing page that compares the different companies job openings (by count), over time. Penalize listings that list/delete/re-list job openings to always make it seem like they are hiring. Maybe have a max time a job post can be open before it's no longer part of their count (3 months?)

What you'll end up showing is similar to this article: A lot of company's stock is evaluated on growth, and part of the growth estimation is based on how much they are hiring in this down market. Some companies know this, and are trying to game the system

Anyways, I bet a decent amount of people would watch that like a hawk, and the honest companies would love it because it would show how great they really are doing

A lot of people here haven't integrated GPT into a customer facing production system, and it shows

gpt-4, gpt-4-turbo, and gpt-4o are not the same models. They are mostly close enough when you have a human in the loop, and loose constraints. But if you are building systems off of the (already fragile) prompt based output, you will have to go through a very manual process of tuning your prompts to get the same/similar output out of the new model. It will break in weird ways that makes you feel like you are trying to nail Jello to a tree

There are software tools/services that help with this, and a ton more that merely promise to, but most of the tooling around LLMs these days gives the illusion of a reliable tool rather than results of one. It's the early days of the gold rush still, and every one wants to be seen as one of the first

If you are a hobbyist reading all this, and trying to compare it to arduino or raspberry pi, do yourself a favor and buy one of these: https://www.adafruit.com/product/5325

$12, fantastic documentation and tutorials, and you can literally just plug it in via usb and edit the python code as if it's just a text file on a thumb drive. No programmers, special IDE's, or specialty equipment

Microcontrollers are fun again

planes/helicopters have a fuel source that is orders of magnitude less volatile, and are also able to safely land without power

The case is not trivial, not crazy, but not trivial. It's also the least hard part about putting together a similar package. A non-EE type person could probably get some dev boards from sparkfun/adafruit, and connect them up to a raspberry pi with some sort of display with some buttons. But that last mile of getting them to play nice, not be a massive battery hog, and have simple/usable software is man-months of work. This also assumes you have a passing knowledge of circuits and are decent enough programmer to get a job, neither of which are low bars

This is likely @pierat's point: sure, it's possible, but your diy version will be objectively worse in a variety of ways while also taking you a pretty long time to put together.

Fun fact: part of the licensing agreement to use USB, is to have the usb symbol on top of the connector. So unless you're using a cheap unlicensed cable, look for the symbol facing up and you'll always be correct

To expand on this, if software was like hardware:

    - when your function calls don't have enough white space between them, they'll some times mix up their return value (crosstalk)
    - the more deeply your if/else statements are nested, the more random their results end up being (voltage/IR drop)
    - when the user makes a selection, your bound function is called somewhere between once and a dozen times. Each time with a different value (bounce)
    - their is no `true` or `false`. You just are given a float between one and zero, where one is true and zero is false. Sometimes `true` is as low as 0.3 and sometimes `false` is as high as 0.7
    - the farther apart your variable declaration is from it's use, the less it represents what you actually declared. If it's two lines apart, a 'foo' string is still 'foo'. Hundred lines apart though, and 'foo' might become 5.00035 (attenuation)

Bitcoin is cryptographically secured distributed accounting ledger, that created a novel way to incentivize participation

"Speed of light" makes more intuitive sense when you think of it more as the speed of causality i.e. the fastest physical speed a cause can have an effect

The speed of light in a vacuum happens to be the best representation of the maximum speed of causality

Which also makes more sense why you can't do things like travel faster than light (your effect would precede the cause), and why two protons going past each other in opposite directions don't violate this law

OpenAI will absolutely be able to raise money again, but it will likely never be on the same scale and will also likely have some serious safeguards in the contract language.

Whether you agree with investors agreeing with firing Sam or not, future investors will absolutely be nervous about sinking serious money in a company that split it's board without talking to key partners/investors first

Honestly, this is the wrong question, but I understand the intent. What you really should be asking is "Where is the online community for your hobby/interest?"

The thing with HN, is it works for the startup/hacker/tech community, and that community has evolved here over time. If you want a link/topic sharing site for other interests, the answer is almost always going to be a subreddit.

But if what you really want is to be immersed in a community, those communities will be in weird corners that you have to discover. The reason for this is that a community almost always wants to be semi-private: discoverable by others that share the same interest, but just far enough out of the public eye to not get bombarded with bots/trolls/obnoxious people

Some examples I know of:

  - Voron (3d printer)? You want to be on their discord
  - BigGreenEgg (kamado smoker)? eggheadforum.com
  - MMA? sherdog
  - Electronics/Robotics diy style hacky projects? hackaday and go to the comment secion

Practical Electronics for Inventors[0] is the go-to book for what you're looking for. Has the right balance of practical application without dumbing things down too much. That being said, if you need help getting familiar again, Adafruit's tutorials[1] are a fantastic resource to have a 'recipe' type project with an associated parts list

[0] https://www.amazon.com/Practical-Electronics-Inventors-Fourt... [1] https://learn.adafruit.com

The standard answer is:

    - Hello World
    - Calculator (add/subtract/multiply/divide)
    - Todo List (create/delete todo items, mark them as done, view completed and current)
    - Blog (user facing blog posts, a place to sign in and create/edit/delete posts)
    - Forum (user account handling, posting, moderation, etc)

Not sure I follow. ChatGPT can solve leetcode questions, and leetcode for better or worse is what everyone has to study for big corps.

I'd agree with your main premise: big corps are using leetcode questions to see what hoops you'll jump through, but blaming this tool for that feels silly

Super fun to use. I hate having to “give up” on a question when practicing and this gives me a bit more options rather than just pass/fail

He co-founded and was president of Y Combinator, the startup accelerator that Hacker News is the community site for. Not famous in general, but very famous here

This is the most naive take on HN. The only secure computer is one that's been unplugged and buried in six feet of concrete, everything after that is a compromise

The real world security issues that companies face are things like:

  - users that re-use tiny passwords written on post-it note that's attached to their monitor
  - regulated industries that don't allow them to actually lock a user out, which causes leaky social engineering flows for their help centers
  - users that constantly forget their passwords, and have terrible forgotten password questions/answers
  - passwords they share with a friend/partner, that they then have a falling out with

"But Joe, having a more secure system isn't that much harder on the user and is infinitely more secure". I promise you, it is that much harder. Most users can barely understand/handle SMS 2FA. Remember, we have to force users to not use trivially simple passwords like 'password'. Shoot, companies like AOL still derive monthly subscription fees because it's too hard for people to figure out how to change email providers

  "If you think the Twitter hack was bad, wait until you realize that back in March every company had to suddenly support remote access to their networks/tools virtually overnight"