HNHacker News
TopNewBestAskShowJobs

martythemaniak

11,234 karma · joined February 26, 2007

submissionscomments
martythemaniak··on Tesla sales crash continues in Europe, with Germany down 70%
Tesla posts delivery numbers the day after quarter end and several weeks before earnings, so that's the one to really watch out for. If they don't post anything on April 1 or 2, you know something real is up.

In general, I think Musk really failed the marshmallow test - keeping on the sidelines and working through the existing system (for autonomy, spaceflight, etc) would have been more painful, but would have delivered greater returns long-term. Instead he decided to take a shortcut by buying a president and torching the regulatory system, and he'll succeed, but will that actually pay off? Self-driving Teslas will be allowed without any questions nation-wide, but will people get into them after what he's done? Will people let Teslabots into their house? I don't think the destroyed trust will make up for the lack of regulations.

Even finances aside, being the #1 oligarch doesn't always end well - just ask Jack Ma or Mikhail Khodorkovsky. He would have been safer under the old rule-of-law American system.

martythemaniak··on Federal workers ordered to return to offices without desks, Wi-Fi and lights
> Take your pick.

Now now, there's no need to choose, you can just blame all of them all at once!

martythemaniak··on Geothermal power is a climate moon shot beneath our feet
They are likely raising a round or preparing to, so would like to raise awareness of what they're doing. I think it's fine - you work on something by yourself for 2 years, you have something neat to show to world and need more money to keep going, so you invite folks over and show them what you've done.
martythemaniak··on Grok 3 is highly vulnerable to indirect prompt injection
Ah, of course OpenAI is to blame for something bad xAI did. My confidence is restored.
martythemaniak··on Grok 3 is highly vulnerable to indirect prompt injection
It's also easy to get its system prompt, which specifically directs it to not say bad things about Musk or Trump: https://www.reddit.com/r/singularity/comments/1iwegbl/groks_...

They had to put that in because people were posting screenshots them asking Grok who the biggest spreader of misinformation on X is and Grok giving a truthful answer, lol.

martythemaniak··on 'What a lie': Danish astronaut responds to Musk
Because "owning the libs" is basically the entirety of the right's ideology. It doesn't matter if it's calling people names, firing park rangers or cheering on Putins war, if it pisses off the libs, it must be great.
martythemaniak··on The subtle art of designing physical controls for cars
Honest question here - do people actually constantly fiddle with climate knobs? I honestly don't understand the whole "knobs are important because I can change without looking", I get it, but are you actually doing this?

I've had a Tesla for many years now, the temp is set to 21C and I don't miss knobs because I don't change this. I prewarm/cool the car from the app so I never enter a hot or cold car, only a 21C car. Occasionally (ie a dozen times a year) I'll change the temp for some odd reason, but rarely while the car is moving. Seat and steering wheel are also set to auto. I just don't think about the temp in my car, but climate knobs seem like a huge sticking point to lots of people?

martythemaniak··on "A computer can never be held accountable"
Funny, I'd say the above passage is the very reason we're headed towards techno-feaudalism. If you believe some mid-level bureaucrat with a computer is history's greatest oppressor, then why not have the world's richest people ransack the government? After all, they are the people's heroes, valiantly fighting oppression.
martythemaniak··on Stargate Project: SoftBank, OpenAI, Oracle, MGX to build data centers
> and they for some reason give credit to Trump?

Well, this one is really simple - nothing of note will happen in the US in the next four years without giving Trump credit for it, because if you don't he'll turn the full power of the state against you. And with no checks on his power, there's nothing to stop him. So yes, this has nothing to do with Trump, but if you don't want to get arrested and harassed, you better give him credit for it. Same playbook Elena Ceausescu used, except she did it just for scientific papers, Trump will do it for everything.

https://paperpile.com/blog/elena-ceausescu-scientist-fraud/

martythemaniak··on O1 isn't a chat model (and that's the point)
One thing I'd like to experiment with is "prompt to service". I want to take an existing microservice of about 3-5kloc and see if I can write a prompt to get o1 to generate the entire service, proper structure, all files, all tests, compiles and passes etc. o1 certainly has the context window to do this at 200k input and 100k output - code is ~10 tokens per line of code, so you'd need like 100k input and 50k output tokens.

My approach would be:

- take an exemplar service, dump it in the context

- provide examples explaining specific things in the exemplar service

- write a detailed formal spec

- ask for the output in JSON to simplify writing the code - [{"filename":"./src/index.php", "contents":"<?php...."}]

The first try would inevitably fail, so I'd provide errors and feedback, and ask for new code (ie complete service, not diffs or explanations), plus have o1 update and rewrite the spec based on my feedback and errors.

Curious if anyone's tried something like this.

martythemaniak··on TikTok preparing for U.S. shut-off on Sunday
You're both right! There was a good article/discussion on on this yesterday, but tldr: They are authentically fake! As in, the creators are not putting up a show with a 'real' person behind the persona, the algorithms have remade whatever person there use to be such that their 'authentic' self has become the persona.

https://news.ycombinator.com/item?id=42696691

martythemaniak··on Hobbyist Builds AI-Assisted Rifle Robot Using ChatGPT
Publicity stunt on the guy's part and PR management on OAI's part. Mostly just clickbait. You can make the same thing with open models running on a local device.
martythemaniak··on TikTok should lose its big Supreme Court case
This is the correct way to think about it. The new America operates in a much simpler manner - you debase yourself and praise the ruler by repeating one of his favourite lies, you give him a lot of money and if he likes you he'll give you what you want. Trump has ordered the court to rule a certain way and they'll do exactly as he told them to.

Arguing the merits of a case, equal branches, judicial independence, etc is just silly.

martythemaniak··on Meta is killing off its AI-powered Instagram and Facebook profiles
Surely you're aware of all the proud red hats out there. Maybe Trump won because he's really good at this religion?
martythemaniak··on Dropbox Engineering Career Framework
Because you're expected to get to the next level in good time, otherwise you're not good enough and will be fired. The level you can stay at is typically the level below Staff.
martythemaniak··on US airspace closures, lack of answers deepen East Coast drone mystery
I used to be amused at all the silly medieval peasants and their weirdness, like witch burnings or the https://en.wikipedia.org/wiki/Dancing_plague_of_1518

I am no longer amused.

martythemaniak··on In Defense of Y'All
I've long used "y'all" as a gender-neutral way of addressing a group, but now that's gone all mainstream I've been considering switching to "yinz".
martythemaniak··on Grug’s Guide to Sound
This format barely worked for the original and it doesn't work here. I threw it in an LLM and it translated it into English, pretty good info in there.
martythemaniak··on United Health CEO Decries "Aggressive" Media Coverage in Leaked Recording
America sure is funny. They just happily elected an oligarchy that can't wait to very aggressively cut anything social spending they can (most government spending is pensions and healthcare, ACA subsidies are also expiring in 2025 and there's 0% chance they get extended), but they're also kinda gleeful about the murder of a CEO that is merely operating a company within the environment they so happily voted back in? It's just an absurd situation all around.
martythemaniak··on TikTok divestment law upheld by federal appeals court
He will absolutely 100% save TikTok. Expect to see him work closely with Xi and Putin against their common enemies during his term.

Here's another fun prediction: China will give him (personally, not the US) a huge sum of money and he will withdraw all support from Taiwan allowing China to take them over. TBD whether there's shooting involved.

I'm open to friendly bets!

martythemaniak··on Waymo closes $5.6B investment round
That is one way, yes, there are others. But if you find my writing unconvincing, here's something you can try: Watch videos of FSD driving and note how often it fails to see objects like cars, trucks, people etc. FSD has 99 problems, but depth perception ain't one.
martythemaniak··on Waymo closes $5.6B investment round
Lidar produces a monochromatic 3D scene, which you can also extract with cameras using SLAM techniques. But these scenes simply do not carry the information needed to navigate the world - you only get one colour, so you cannot see signs, markings, lights, anything. LIDAR is thus a small, secondary input to the problem of "perception".

Once you've accurately perceived the world, the hard part is the action, and the judgement needed to drive safely. It doesn't matter how you got your labeled 3D scene (LIDAR or SLAM), you have to figure out how to navigate it safely and this is more like 90% of the self-driving challenge. If you look at videos of the latest FSD in action (12.5) perception is not an issue. Teslas see everything just fine, they don't disengage because they didn't see a car, but because their judgement is not human level yet.

martythemaniak··on Waymo closes $5.6B investment round
People will tell you they're fundamentally different, but they are in fact the same. First, there are two different, independent aspects: Perception and Action. Seeing the world, vs taking action in that world.

For perception Waymo uses more sensors than Tesla. It uses lidar to construct a 3D scene of what the car is seeing, while Tesla uses SLAM-like techniques with their cameras. What people are missing is that these are only a small part of the perception problem - lidar returns a monochromatic 3D scene so it cannot see labels, markings, read signs, lights etc. LIDAR simply doesn't carry the necessary information needed to navigate the world, thus it is a secondary source of info. Cameras in motion do carry all the information needed, so the big difference between them is only one part of the overall perception stack.

Once you have constructed a labeled, accurate 3D scene (whether by lidar or SLAM), "action" is the same and there is no difference between Tesla and Waymo here. They both have to learn how to drive safely using the same information, so it's going to be a lot like LLMs where there's difference between LLAMA/Claude/GPT, but they're also all kinda the same thing.

martythemaniak··on Waymo closes $5.6B investment round
Big Tech is largely set of interlocking duopolies.

* Ads: Google/Meta

* Cloud: Amazon/MS

* Mobile: Google/Apple

* Desktop: Apple/MSFT

* VR: Apple/Meta

There's also smaller players and sometimes strong third players (eg GCP), but by and large there are no single winners. Waymo and Tesla are looking like a good bet on the self-driving duopoly. I think a lot of people are skeptical on Tesla because "LIDAR", but lidar is largely irrelevant. It is only used in 3D scene reconstruction, it has nothing to do with actual driving which is the hard part.

martythemaniak··on Notes on Anthropic's Computer Use Ability
I've been been hacking on a web browsing agent the last few weeks and it's given me some decent understanding of what it'd take to get this working. My approach has been to make it general-purpose enough so that I describe the mechanics of surfing the web, without building in specific knowledge about tasks or website. Some things I've learned.

1. Pixels and screenshots (video really) and keyboard/mouse events is definitely the purest and most proper way to get agents working in the long term, but it's not practical today. Cost and speed are big obvious issues, but accuracy is also low. I found that GTP4o (08-06) is just plain bad at coordinates and bounding boxes and naively feeding it screenshots just doesn't work. As a practical example, another comment mentions trying to get a list of flight recommendations from Claude computer use and it costing $5, if my agent is up for that task (haven't tested this), it would cost $0.10-$0.25.

2. "feature engineering" helps a lot right now. Explicitly highlighting things and giving the model extra context and instructions on how to use that context, how to augment the info it sees on screenshots etc. It's hard to understand things like hover text, show/hide buttons, etc from pure pixels.

3. You have to heavily constrain and prompt the model to get it to do the right thing now, but when it does it, it feels magic.

4. It makes naive, but quite understandable mistakes. The kinds of mistakes a novice user might make and it seems really hard to get this working. A mechanism to correct itself and learn is probably the better approach rather than trying to make it work right from the get-go in every situation. Again, when you see the agent fail, try again and succeed the second time based on the failure of the previous action, it's pretty magical. The first time it achieved its objective, I just started laughing out loud. I don't know if I've ever laughed at a program I've written before.

It's been very interesting working on this. If traditional software is like building legos, this one is more like training a puppy. Different, but still fun. I also wonder how temporary this type of work is, I'm clearly doing a lot of manual work to augment the model's many weaknesses, but also models will get substantially better. At the same time, I can definitely see useful, practical computer use from model improvements being 2-3 years away.

martythemaniak··on Starship Flight 5: Launch and booster catch [video]
That's excellent work, only nit is I think these were not the v3 raptors. I think they're just now starting production of those, so they still have a bunch of earlier ones they need to use up in these early test flights. This actually makes your calculation more accurate, as earlier versions will need a slightly longer burn time.
martythemaniak··on Y Combinator Traded Prestige for Growth
Everyone is missing the reason they got accepted: they are some kind of tech influencers have 300k subs on YouTube.

You have 10 mins to pitch your idea, so they went in and said ”look, we work for coinbase, we're competent. Look, we have 300k subs on YouTube, we know how to sell shit and get attention. Look we have <some random AI thing>, we got it all" and that's all it takes. The problem is, they want to sell dev tools and those tend to be more grounded in reality than <some random social media thing> where their attitude and skills might be a better fit. Lots of devs like calling out bullshit as a hobby and making fun of shoddily built stuff, so it's a tough audience.

YC doesn't do "due diligence" and never has, not worth it for the $125k they put in. The real danger is they become known as the place for bullshit peddlers, influencers, etc.

Edit: 344k subs: https://youtube.com/@fryingpan?si=QIPTDvJATXFNYBPM

martythemaniak··on The Continued Trajectory of Idiocy in the Tech Industry
A big part of the idiocy in tech is the endless stream of generic, cynical neckbeard takes that are indistinguishable from AI slop.
martythemaniak··on The Future of European Competitiveness
Canada also suffers from lack of competitiveness, productivity, etc and I always wonder if we actually have the political constituency for a pro-growth agenda. I don't think we do it seems the group that would vocally back this is likely <10% of the population and probably low-engagement/apolitical. Europeans seem to give off the same vibe, if not more so.

For example the governing Liberals have been pretty bad on the economic file and will most certainly lose the election next year, but the alternatives are: 1) A Russian-propaganda fuelled Conservative party led by an obsessive culture warrior who's never known life outside politics (a 45 year old who has been an MP for 20 years) 2) a leftist party that thinks more regulation and taxes is what we need.

The "Abundance Agenda" types just don't fit in anywhere, except perhaps as a small faction of the Liberals.

martythemaniak··on SpaceX Pulls Employees from Brazil, Discourages Travel There
The ruling has been upheld by the wider court: https://www.theguardian.com/technology/article/2024/sep/02/b...
← PreviousPage 8 of 34Next →