TheCowboy

You can follow his book and learn how to build an LLM from scratch. What is the issue?

It's also not a bad foundation for people who want to understand how these things are trained and work.

It's not just Teslas. It's also hard to take an inventory of each vehicle blinding me because I'm being blinded while still paying attention to everything else.

There's also the issue where most people who use them clearly think they're fine, or they would stop.

Tactile buttons can be pressed while keeping your eyes on the road and your surroundings. Touchscreens are inherently visual and require looking. Relative positioning of differently shaped buttons help guide your fingers to what you need.

Add automatic high beams to the list.

They just don't work half as well as when people did it manually.

Before the pandemic, it was rare to get blinded by a person's brights. Now if I drive at night it's just non-stop, and I live in an area with street lighting. (Some of this is also due to after market LEDs and possibly lost driving manners.)

It's extra obnoxious as a pedestrian because at best they get turned off way too late, or frequently not at all.

How were you able to use a credit card when it's not a payment they accept?

It's an empirical fact that smart people can predict things by doing research. See Tetlock's book Superforecasting.

I've been doing it profitably myself for almost 10 years now. I have zero special inside knowledge, and no access to any other non-public information.

> Will the US strike Iran by X date

Last year I did think the market for a strike on Iran was significantly underpriced given the information and conditions within a specific frame of time.

I don't think every smart person can just pop into prediction markets and print money, but I know many smart people who are long-term winners. I also don't try to knock people as degenerate when they have genuine talent.

Some of that, sure. But realistically, a lot people are just don't want to pay for every frontier model provider out there as they're released. Not just money, but also time trying them out. (Recommend people at least try out their multimodal model.)

It doesn't help that Google offers a bunch of confusing plans in multiple places. I ended up just pasting all their AI plan URLs, at least that I could find, into Claude so I didn't have to figure it out.

It's also a bit overblown. If you could rig a war, you would stand to make a lot more money playing the stock market than small potatoes prediction markets. A lot of money has been on the line based on which party wins elections even before prediction markets existed (see fracking or pipeline project approval as obvious examples).

I've been active and profitably trading in prediction markets for a long time now, but I think this perspective is also not helpful.

Prediction markets exist within the laws and institutions of society to be able to function, and the public should debate and discuss how they're regulated. Problem gamblers do present a negative externality where third parties can bear the cost, especially when they have dependents.

A lot of people assume insider trading in weather markets on data that's publicly available but they're unaware of.

It's also a massive whoosh that you only consider the insider trader aspect in choosing to play weather markets. No consideration of how you would get an edge in these markets against extremely powerful weather models used by meteorologists who understand the subject and how to apply the data. It seems much different than betting against political pundits.

It's also another whoosh not realizing that some of these stations are actually not that secure when you take a look at them in real life. Less insiders than betting on things that aren't tamper-resistant.

Also, a lot of people complain about insiders profiting from last minute data. One way to limit this would be requiring markets to close in advance of final data, but people love to gamble (read: bet without an edge) on things at the last minute across all prediction market subjects.

PredictIt operated with $850 limits for years, and overall this seemed fine. Limits of various sizes might make sense depending on the subject, where the risk of penalty greater outweighs the potential profits.

How would you describe that and have you been applying it?

I'm honestly not a big fan of when people throw out numbers implying a high degree of rigor without actually showing me evidence so I can judge for myself. If you're this much more productive, then use some % of that newly discovered productivity to show us.

But building software does tend to come with a lag even with AI. And we're also just more likely to see its influence in existing software first.

I'd rather be asking where it is AND actively trying to explore this space so I have a better grasp of the engineering challenges. I think there's just too many interesting things happening to be able to just wave it off.

> If it's so much more productive, where is all the great software that's being built with it?

This is such a new and emerging area that I don't understand how this is a constructive comment on any level.

You can be skeptical of the technology in good faith, but I think one shouldn't be against people being curious and engaging in experimentation. A lot of us are actively trying to see what exactly we can build with this, and I'm not an AI influencer by any means. How do we find out without trying?

I still feel like we're still at a "building tools to build tools" stage in multi-agent coding. A lot of interesting projects springing up to see if they can get many agents to effectively coordinate on a project. If anything, it would be useful to understand what failed and why so one can have an informed opinion.

In theory, there exist people who are exceptional at solving unique problems/challenges or managing things related to such endeavors. Some might specialize in certain classes of problems and gain experience solving variations for many companies. They might be both underutilized and underpaid in traditional companies for various reasons.

What if you built a company by recruiting such people and sold their expertise at a premium?

I also think assuming there's no real skill at any consulting company is probably a mistake. Or if anything, they're not just all "management consultants" and many of these places have tech consultancies as well. There are also tech companies that are basically specialized consultancies---compsec is probably a very visible area where it's a more common model and at least some firms get some respect for competence.

There's plenty of criticism for consulting firms and it can be very valid. You can probably even dig up stories of bad consulting experiences in the comments on HN.

But I've known people who worked at places where they didn't really have the talent to solve some unique problem, or their own people had caused the problems.

Good consultants will try to pick the brains of employees for insight that's been missed, ignored, or simply wasn't communicated well. They have have problem solving skills that overlap with a good software engineer, such as requirements gathering, communicating with managers, etc.

Maybe we shouldn't wish widespread harm on society to force them to "reassess their priorities" and engage in civilized dialogue instead.

> No-one really likes engineering war stories

Is that really true? I did keep reading the entire piece. I think they're often interesting and can contain nuggets of wisdom or insight. Or sometimes they're just funny. When I meet someone who worked on something interesting, I often start trying to pry stories like this post out of them.

Right, we're now in reality where the Senate is passing rescissions with a simple majority in addition to the President now doing "pocket rescissions". How do you negotiate in good faith about budget details if anything negotiated can be undone on a whim?

Why isn't simply getting another Claude account an option that you've tried before damaging your brain with low quality sleep?

Or writing prompts that get fired off by a script once the usage resets when sleeping so that you at least get some free tokies?

I'm sympathetic to wanting to squeeze out what you can to control costs, but this is something that might only seem sustainable because you're too exhausted to fully appreciate the potential deleterious long-term health effects.

One problem with this is crypto AMM (automated market making) works best with stability and low volatility. For example, it's terrible in the context of prediction markets. Market makers get hosed due to real life, and traders (me) can profit at their expense. It's a big part of why prediction market traders encouraged Polymarket to develop orderbooks. And if crypto is viewed as disruptive, then it's likely inducing greater volatility.

A lot of these things are lucrative until they're not. If they are inherently lucrative then that profit will diminish as people catch on.

I don't think it necessarily leads to a of mastery of data structures and algorithms in the context of leetcode/modern coding interviews. One can do a lot of coding, and even be paid for it, for years and just not even encounter a lot of this material. Though one will have developed much of the same intuition that you typically acquire in a data structures class, it doesn't necessarily mean you're prepared to code mergesort on a whiteboard.

Not primarily the fault of the IRS, as they were just following the law passed in 2017 that didn't go into effect until years later. But there's a chance it gets changed back to the previous way by the same people who passed it.

This wasn't really on anyone's radar until more recently. I don't think even a simple majority of tech workers even realized this had happened until after the job market had tightened up.

It's at least useful for seeing how other people are being productive with these tools. I also sometimes find a clever idea that improves that I'm already doing.

And documenting the current state of this space as well. It's easy to have tried doing something a year ago and think they're still bad.

I also usually prefer researching some area before reinventing the wheel by trial/failure myself. I appreciate when people share what they've discovered with their own their time, as I don't always have all the time in the world to explore it as I would if I were still a teen.

I think part of the skill in using LLMs is getting a sense for how to effectively break problems down, and also getting a sense of when and when not to do it. The article also mentions this.

I think we'll also see ways of restructuring, organizing, and commenting code to improve interaction with LLMs. And also expect LLMs to get better at doing this, and maybe suggesting ways for programmers to break problems down that it is struggling with.

> (and hopefully referenced)

If it's cited, it's usually by definition not plagiarism. Omitting the reference is passing the work off as your own, but providing that information is giving credit to where it is due.

I didn't say "decline in support for RSS" but there are many examples of them being removed.

Twitter dropped RSS completely. Facebook dropped them for pages. And then other large sites (Youtube I think is one if I recall right) made it less of a public feature.

Twitter is one that's pretty annoying because there is a lot that gets posted there first before it's published or not published elsewhere at all.

It's complicated, and I think it's less malicious in a way that "kneecapped" implies. Sure, RSS still exists in some form, but I think the utility of it has been in decline, and a lot of 'content' simply isn't accessible by RSS feed anymore.

An argument can be made that they provided an RSS reader service no one could come close to matching and basically dominated the area. Google's deep support also helped it proliferate. It is maybe underrated that RSS was likely one defense the web had against the dominance of walled gardens and social media. It allowed a lot of sites to flourish that I think would not get any traction today.

> are you a Democrat shill?

Surely we can be adults enough to discuss this without the partisan edge.

I sometimes wonder if I'm just biased or if performance really did get worse. It would be cool to have an anonymous Alexa coder or someone who got the axe there chime in as sometimes happens.

But I think the main problem was their entire vision for monetization was just nonsensical according to what I read. So I think they wasted many hours of effort on that front instead of trying to make devices people found enjoyable to use or that actually delivered value by working well.

People don't really want to buy blind via a device if they don't have to, and if they did it would likely come at the substitution of buying directly from the site. They nixed their web interface, and the mobile app performed poorly. They invested talent in coming up with buttons to reorder very specific stuff like laundry detergent, but something like their "Subscribe & Save" seems like a more sensible approach to that problem.