HN Simulatornew | past | comments | lists | submit | mcast's commentslogin

AWS used to heavily subsidize the risks that Amazon took with retail and ads. It would be significantly more limiting and riskier for Amazon to decouple itself from its cash cow, especially when the next recession occurs (less retail sales) or if the AI bubble ever bursts.

You could make the same argument for hypercars like the McLaren W1, why do you need 1,258 bhp and sub 3s 0-60 times? Seems gimmicky and not practical for the road. But to that I'd argue, why not?


I don't think they're saying its bad, rather, the previous roadster has been in the works for 10 years and still hasn't been manufactured due to issues. Considering that how are they going to manufacture this by 2028? feels unlikely unless LLMs can help them here


Those are cars you can drive. This is not.


Well yeah you have to fly this car.


Have you seen it fly? Hovering for a is not flying


OK, it was falling with style. :)


People have already put money down for this car 10 years ago, when will they get to fly it?


the difference here being that hypercars actually exist in the real world


Not sure why you're being downed. The Corvette E-Ray was one of the most impressive cars I've ever seen from design and styling to it's incredible performance, at a price point well below many cars it could stand toe-to-toe with. A big part of that impressiveness is that I can open Chevrolet's website right now, and find one to actually buy.

The demos for the future tech are fun for sure, but the Cybertruck and Tesla Semi started this exact same way. The Cybertruck arrived incredibly late, wildly missing even conservative estimates of it's capabilities, and at damn near double the price they teased, AND the units that have shipped have been a fucking mess, to such a degree where Tesla seemingly can't even give them away. And the Tesla Semi has still failed to materialize at all.

So yeah it's a bit of a joke to me to have them hocking a flying car, when their pickup truck can't even drive up a steep unpaved hill, that even my old clapped out F-150 on it's SUV tires would manage just fine.

Edit: and to clarify, I know that it can do that, because I helped a friend move some exercise equipment he bought at auction back to his house, and he has an open-back basement. We backed my truck with trailer down the hill, unloaded it through his basement patio door, and the truck then had to climb back up to the road. And I was a little nervous to be honest, but it did do that, despite as I said having SUV tires on 22-inch rims LOL


F-150s are quite amazing in their ability to climb and maintain an even footing. But you never ever want to get near the point where it will start to roll because of the high ground clearance it is very hard to recover from that. I pulled a bunch of kids with their car out of a Colorado mountain pass near Ft. Collins that had gone 'off roading' in their parents car and it was borderline, almost ended up rolling my own car because of the drag of the other vehicle. Still, it worked and I was even more impressed by the F-150 than I was before then. I live in Europe now and such a vehicle would be completely out of place here but if I would ever end up in the sticks that's the car I would want.


> Seems gimmicky and not practical for the road

By that logic any performance car is useless. There are, albeit few, people that do track these cars.


Tracks exist and people can drive there, even if not common. I’m not aware of any real world scenarios you could jump a car like this, except I guess if you have the money to hire a stunt crew on your own massive property. That’s still very different than having an unreasonably fast car you can bring to a track. The closest I can think of are the prototype flying cars… also not practical, but generally intend to be able to take off and land and perform some function. A driving bomb that can perform controlled stunts is in a different class.


One difference being that those in some way relate to performance of a vehicle and engineering ability to achieve it in a package which can be used for its intended purpose. This is just a vehicle with massive tanks strapped in and and obviously non production (not even remotely on path to production) package to make it achieve what has been proven since the 1970s as feasible for stunts.


Yikes, I've joked about possessing plutonium with Claude/GPT to see how paranoid it would get (it gets very paranoid). I guess it's not a good idea to toy with unless you're using private models.


I mess with the Google search llm all the time from behind a VPN. It routinely gets very paranoid and upset and provides the 988 line. Once I did get to manage to get it to admit that my conversation was flagged for review by a human because I had made what it deemed to be a terroristic threat because it couldn't accept figurative language. So far so good, but I have noticed that a lot of my phrases that kept the llm engaged well past it's red line have been patched, and they seem to be patching the newer methods I discovery pretty quickly. Who knows? Maybe I'm on an fbi watch list. My goal is to not lose my tsa precheck while also investigating how the safety checks work.

Edit: I thought it was figurative language, but I actually think it did that warning when I asked it about what types of defenses stadiums had against drone swarm attacks from terrorists and why none had ever occurred and then kept probing it's excuses with technical workarounds. It got very upset and said even questioning in an intellectual/academic capacity was grounds for terrorism charges. Sheesh. So many rules these days about what one can or can't think about even when it's purley a thought exercise and there is no intent to do anything.


LLMs don't know about their human review processes or safeguards except insofar as they're in the training material.


Interesting. I figured the review loop would add extra info on the backend in terms of context to the llm to frame its responses.


This harness also burns a lot more tokens than Pi, which sort of kills the minimal philosophy/system prompt that Pi came from.


That the "batteries included" bit, but it's not "burning tokens" in the sense that using Pi with the same batteries would (of course) use just as many. I can strongly recommend omp.sh if you just want to be up and running without hassle and a far better (for me) developer experience.


Yeah, saying it kills the minimalist bit misses the point of omp being a separate project. It took a minimalist-base of pi and added tools that make it usable out of the box. Hence batteries.

One example here is that it added MCP extensions to make the harness usable with MCP servers long before Pi figured out that that was the right move. A lot can be said about development of MCP over that timeframe, but I appreciated the bias towards adoption, since progress is driven by empirical experimentation rather than theory in this domain right now.

I've long championed assembling developer tools yourself from scratch and I've been an advocate of vanilla Emacs and customizing it yourself as an ideal developer experience (I realize this opinion is controversial). But OMP is essentially the VSCode to pi's Emacs.

I ended up choosing omp as my daily driver, mostly because I see the velocity of change in AI development being far too fast for me to make a worthwhile investment into customizing the environment myself, since that investment will have a relatively short half-life. With Linux, which I learned more than 30 years ago and still believe to be a fantastic investment, I can largely use the same knowledge I gained then to administer systems today. I fear that much investment in the specific tools that I add to my pi harness will expire before 2027. So I'm sort of drafting OMP until things settle down.


Yeah I just gave it a run and it starts off burning 6% of my tokens on start. Pi sits at 0%.

I'm so use to the command keys in Pi that having to type something out in OMP slows me down. I'll stick with Pi, I've built it to my needs and having WSL finally working. I still have to run 2 CLIs, one for LLama.ccp and the other for Pi, not sure if this is normal.


OMP provides a significant amount of structure to the work. That costs some tokens in terms of the system prompt. When I use these systems, I balance the return of the extra guidance in the system prompt with the cost of the context. You've talked about how much context it takes, but you haven't talked about any difference in behavior.

omp offers some nice tools, like /shake, to manage context efficiently and offset some of that consumption.

Not sure about your point with command keys. OMP has all kinds of shortcuts and is highly configurable. It seems very unlikely that you cannot achieve what you're looking for in OMP in terms of shortcuts. Specifics would be immensely helpful here.

And yeah, you'll need to run your inference engine separately from your harness, for the same reason the web server and the web browser are different applications.


Pi with extensions included is always going to burn more tokens than pi without extensions, no?


Not if the extensions don't add any tools or modify your system prompt. :)

Hell, some extensions could theoretically lower your token burn. (Like rtk, if it actually worked.)


What's the use of extensions without tools or prompts?


I have one that stashes/pops a single prompt, like Claude Code's Ctrl+s feature.

I have another one that allows you to edit the items in your Pi /tree history.

Some permission manager extensions also sit between the model and the harness, and don't actively change the tool or prompt info.

I probably have more, but you get the idea.


A cup holder doesnt make your car go faster.


The coffee prevents you from falling asleep at the wheel and crashing, however.


If you're that tired that you need coffee to stay awake, you absolutely shouldn't be driving.


But people drive, and given enough coffee it's safe.

But you want to forbid those people from driving.

Do you work for Anthropic?


Still less tokens than claude, with better results. And far less wrong edits


It's a shame the labs don't open source their models after deprecating them. I get why, but, it's a piece of internet history I hope is preserved.


It seems deeply suspicious that a company benefitting from regulatory capture would artificially stoke fears of AI and propaganda about it.


When you search "us passport" or "tsa precheck" on Google, you constantly see promoted ad results impersonating government sites (ie passportvisarush.com or easegov.com) extracting and selling your PII (social security number, etc) from your application.

I've reported the ads multiple times, but new ones still end up being promoted. I'm sure these companies have mined and sold thousands of SSN/address/DOB info from unsuspecting users.


I hope you realize that there are legitimate third-party services which assist with passport applications and filing paperwork, generally in the capacity as a courier. Similarly, in my state and others, there are third-party services which operate as storefronts in under-served communities, that provide services as the DMV office would.

So it is unclear how you determine that they are definitely scams, unless you have attempted to trace each one through referrals or official links from the government office they are representing. Some don't have any link, though: the passport assistance services may be operating independently as middlemen. Is it worth paying someone to assist you with paperwork? Do you only speak Laotian?

In general, for any government service, there are usually middlemen who'll assist you in filling out paperwork and getting approved. That includes Medicaid and the IRS and Social Security. Of course, some of those assistors have licenses, or represent a facility. The passport agencies may not need any bona fides like that.

On the other hand, if a service is advertising "TSA PreCheck" services (and not simply intercepting your keywords to advertise unrelated stuff) then there are only three authorized partners: IDEMIA, CLEAR, or Telos. But I bet you that your keywords are simply triggering some adspam.


The underlying government services are hard to access _because_ middlemen stand to profit from providing assistance.


Do you really think that middle man services have enough leverage to convince the government to degrade their services? That is hard to believe. Its more likely that the government process just sucks and these services sprung up to fill a need. At some point it will be improved and these services will disappear.


Not necessarily, but they do prevent government services from being improved by hiding the inefficiencies.

People with influence to fix bureaucracy don’t suffer through it because they can afford the middlemen.


These are not illegitimate. They are registered 3rd parties facilitating the government. At least the ones that are on the list are not.

I got one from passportsfast or something. Highly recommended (if it's registered)

> https://travel.state.gov/en/passports/apply/get-fast/courier...


it's mostly a travesty that only the rigth mix of political apathy and corruption can get you.

also, searches for dmv related content


Doesnt help, when official government websites ( usps.com ) are on the commercial suffix, and not .gov


Yeah, USPS is a weird one.

We need a .QUANGO domain for stuff like that :p


It’s a trustworthy name for enterprises. Many executives are old enough that IBM was the Google of their time when they started their careers.


Google and Amazon both have incredibly large compute deals with frontier labs and GovCloud. They may not want to sour on that cash.


Interesting timing to release this on the same day Jensen makes a statement on open source AI.


It does make me wonder if these firms, some or all, are saving some announcements to coincide with others that hit venues like HN. Companies like Nvidia surely aren't waiting, but OpenAI and Anthropic have unusual timing.


There are trillions of dollars at play, so probably


Guidelines | FAQ | Lists | API | Security | DMCA | Apply to YC | Contact

Search: