cryptoz

15 years later and still no word from Google if they will use the barometers in Android devices to assimilate surface pressure data. It has been shown that this can improve forecast accuracy. I think IBM may be doing it with their weather apps, but Google/Apple would have dramatically more data available.

Apple even bought Dark Sky, which purported to do this but never released any information - so I doubt they really did do it. And if they did, I doubt Apple continued the practice.

Been waiting a long time to hear Google announce they'll use your barometer to give you a better forecast. Still waiting I guess.

Montreal public transit times used to be on some kind of like, 28-hour clock. Bus times after midnight would be labelled 27:30 or something. Suuuper confusing. It sounds so bizarre in fact, that I'm doubting my memory a bit, but I'm certain it was like that (say around 2006 or so).

They're testing specific things with no need for full orbit, although I think they reach verrrry close to orbital velocity. They want the payload dummies to 'de-orbit' quickly (from a suborbital trajectory). They could easily have gone orbital if they wanted to. I guess we'll see orbital demonstrations after a few splashdowns of v3 stack early next year.

The video and info: https://www.spacex.com/launches/starship-flight-11

(Liftoff is around 33 mins in)

I'm working on Code+=AI: https://codeplusequalsai.com/

It's an AI-webapp builder with a twist: I proxy all OpenAI API calls your webapp makes and charge 2x the token rate; so when you publish your webapp onto a subdomain, the users who use your webapp will be charged 2x on their token usage. Then you, the webapp creator, gets 80% of what's left over after I pay OpenAI (and I get 20%).

It's also a fun project because I'm making code changes a different way than most people are: I'm having the LLM write AST modification code; My site immediately runs the code spit out by the LLM in order to make the changes you requested in a ticket. I blogged about how this works here: https://codeplusequalsai.com/static/blog/prompting_llms_to_m...

Computer Use models are going to ruin simple honeypot form fields meant to detect bots :(

I wonder if this stuff is trained on enough Hallmark movies that even AI actors will buy a hot coffee at a cafe and then proceed to flail the empty cup around like the humans do. Really takes me out of the scene every time - they can't even put water in the cup!?

A way for people to build LLM-powered webapps and then easily earn as they are used: I use OpenAI API and charge 2x for tokens so that webapp builders can earn on the margin:

https://codeplusequalsai.com

I've really got to refactor my side project which I tailored to just use OpenAI API calls. I think the Anthropic APIs are a bit different so I just never put in the energy to support the changes. I think I remember reading that there are tools to simpify this kind of work, to support multiple LLM APIs? I'm sure I could do it manually but how do you all support multiple API providers that have some differences in the API design?

Opening the App Store to download a bunch of apps - in general - is probably the #1 thing people are doing when they open the App Store. Of course, installing a specific app is a top use case. But I think you're just not the average user. Lots of people open the App Store frequently to just check out what's available.

~10 years ago I would do this all the time. It's fun, kind of like surfin' the net was back in the old days, but in a walled garden of applications.

Had a lot of fun trying to break this. Turns out you can screenshot real easily by zooming out. Maybe there are other ways but I stopped trying :)

Oh my God. That domain is parked and for sale for $125,000?!?!

Wild.

Gonna be some wild conspiracies some day in the future, when humanity has altered the moon visibly but 'good old phones from way back in the day' take photos that "clearly" show no change to the moon.

Did you do anything fancy to get the whole Pico docs into context? I see on their website that their docs are split up across a bunch of pages, did you copy+paste each one into a large prompt document? I was hoping to see a single-page docs download but I didn't find one.

Oh gosh sorry, I do try to contribute positively to HN and write quality comments. I'll expand: I've been in circumstances where I've been rented a company car in a foreign country, felt that I was a good driver, but struggled. The road signs are different and can be confusing, the local patterns and habits of drivers can be totally different from what you're accustomed to. I don't doubt that lots of humans could drive most roads - but I think the average driver would struggle, and have a much higher rate of accidents than a local.

Germany, Italy, India all stand out as examples to me. The roads and driving culture is very different, and can be dangerous to someone who is used to driving on American suburban streets.

I really do stand by my comment, and apologize for the 'low quality' nature of it. I meant to suggest that we set the bar far higher for AI than we do for people, which is in general a good thing. But still - I would say that by this definition of 'full self driving', it wouldn't be met very well by many or most human drivers.

Many humans couldn't.

I think you’re referencing the tornado EF scale, not the hurricane Saffir-Simpson scale.

Oh I don’t use Gemini! I did try it out and admittedly formed an opinion too narrow on cli agents. But no way do I actually use Gemini.

Oh that's wild, I did suspect that but didn't know it outright. Mind-blowing Google would release that kind of thing, I had wondered why it sucked so much haha. Okay so what is a good representation of the current state of coding agents? Which one should I try that does a better job at code modifications?

I really think the current trend of CLI coding agents isn't going to be the future. They're cool but they are _too simple_. Gemini CLI often makes incorrect edits and gets confused, at least on my codebase. Just like ChatGPT would do in a longer chat where the context gets lost: random, unnecessary and often harmful edits are made confidently. Extraneous parts of the codebase are modified when you didn't ask for it. They get stuck in loops for an hour trying to solve a problem, "solving it", and then you have to tell the LLM the problem isn't solved, the error message is the same, etc.

I think the future will be dashboards/HUDs (there was an article on HN about this a bit ago and I agree). You'll get preview windows, dynamic action buttons, a kanban board, status updates, and still the ability to edit code yourself, of course.

The single-file lineup of agentic actions with user input, in a terminal chat UI, just isn't gonna cut it for more complicated problems. You need faster error reporting from multiple sources, you need to be able to correct the LLM and break it out of error loops. You won't want to be at the terminal even though it feels comfortable because it's just the wrong HCI tool for more complicated tasks. Can you tell I really dislike using these overly-simple agents?

You'll get a much better result with a dashboard/HUD. The future of agents is that multiple of them will be working at once on the codebase and they'll be good enough that you'll want more of a status-update-confirm loop than an agentic code editing tool update.

Also required is better code editing. You want to avoid the LLM making changes in your code unrelated to the requested problem. Gemini CLI often does a 'grep' for keywords in your prompt to find the right file, but your prompt was casual and doesn't contain the right keywords so you end up with the agent making changes that aren't intended.

Obviously I am working in this space so that's where my opinions come from. I have a prototype HUD-style webapp builder agent that is online right now if you'd like to check it out:

https://codeplusequalsai.com/

It's not got everything I said above - it's a work-in-progress. Would love any feedback you have on my take on a more complicated, involved, and narrow-focus agentic workflow. It only builds flask webapps right now, strict limits on what it can do (no cron etc yet) but it does have a database you can use in your projects. I put a lot of work into the error flow as well, as that seems like the biggest issue with a lot of agentic code tools.

One last technical note: I blogged about using AST transformations when getting LLMs to modify code. I think that using diffs or rewriting the whole file isn't the right solution either. I think that having the LLM write code that modifies your code and then running that code to affect the modifications is the way forward. We'll see I guess. Blog post: https://codeplusequalsai.com/static/blog/prompting_llms_to_m...

Reminds me of the Improv Everywhere sketch where they did this exact thing.

https://youtube.com/watch?v=EKEeHREK2nQ

I really think people are approaching LLMs wrong when it comes to code. Just directing an agent to make you something you’re unfamiliar with is always going to end up with this. It’s much better to have a few hours chat with the LLM and learn some about the topic, multiple times over many days, and then start.

And ask questions and read all the code and modify it yourself; and read the compile errors and try to fix then yourself; etc. Come back to the LLM when you’re stuck.

Having the machine just build you something from a two sentence prompt is lazy and you’ll feel lazy and bad.

Learn with it and improve with it. You’ll end up with more knowledge and a code base for a project that you do (at least somewhat) understand, and you’ll have a project that you wouldn’t have attempted otherwise.

I have come to absolutely despise web form inputs with front end email validators that are broken. Input field hints to type your email, so you start typing. As soon as you type the first letter it goes red and says “error!!! Invalid email!”

Unbelievably frustrating.

You can use any phone with a barometer to make a scale. All iPhones since the 6, and all the Pixels, and Samsung flagships have one. You get a zip loc bag, blow some air into it, put your phone in running an app that shows the pressure in a big font (so you can see it through the ziploc). Then you put an object of known weight on it like a quarter (balanced carefully on top of the air-filled ziploc) and note the pressure change on the display. With that, I think the weight / pressure change scales linearly, so you can now weigh anything small that you can balance on the ziploc.

That will be done with a solar gravitational lens - there's a recent-ish NASA paper about it. Basically you send your probe to > 550 AU in the opposite direction of your target exoplanet, point it at the Sun and you will get a warped high-res photo of the planet around the Sun. You can then algorithmically decode it into a regular photo.

I think the transit time is likely decades and the build time is also a long time as well. But in maybe 40-100 years we could have plentiful HD images of 'nearby' exoplanets. If I'm still around when it happens I will be beyond hyped.

Fantastic feedback, my gosh, thank you. Yes the home content doesn’t do a good job of explaining what it is. Honestly that’s because a lot of the ideas I typed here are pretty fresh, and my ‘marketplace’ is only implemented on prod as a prototype. I do see the the mismatch in copy and the real utility.

I’ll work on positioning and update the copy to keep in sync with my vision …

And yeah, my target audience is developers, specifically those with not a lot of time but with a lot of ideas.

Gosh thanks again, that’s super excellent feedback.

Yeah this exact thing is happening to me also. Minutes of runtime and only errors. I guess I’ll try again later? I have billing up and I’m Tier 1. Wouldn’t expect to hit limits like this on the first prompt.

I'm building something like this. The value to you would be that you could earn a margin on the token costs. That is, the end user is charged 2x the token cost of the API call. The API provider earns the base cost, the platform owner earns 20% of the remaining cost, and the webapp creator earns 80% of the remaining price.

So for an API call that costs $0.50, the end user is charged $1; and from that AI API earns $0.50, the webapp creator earns $0.40 and the host earns $0.10.

I'm trying this out with https://codeplusequalsai.com right now but it's not clear to me yet that it will take off!

But clearly, the value to you should be that you could earn $ based on the token usage from end-users.

All life on Earth is going die. Humanity has never been content with staying put, why would we start now? And what do you mean "literally nothing there"? The universe has a loooooot of stuff in it.

I remember going into local browser cache folders and pulling out YouTube videos in full. Am I remembering wrong or did in fact the #1 video streaming platform simply just download the videos to your hard drive, same as you would have with right click save? Only difference is the default folder it goes to.