HNHacker News
TopNewBestAskShowJobs

briandw

1,107 karma · joined May 25, 2009

submissionscomments
briandw··on Tesla changes meaning of 'Full Self-Driving', gives up on promise of autonomy
Lidar is the first thing brought up in these discussions. Lidar isn’t that great of a sensor. It does one thing well and that’s measure distance. A visual sensor can be measured along the axis of spatial resolution (x,y,z) temporal resolution(fps) and dynamic range(bit depth). You could add things like light frequency and phase etc. Lidar is quite poor in all of these except the spatial z dimension, measuring distance as mentioned before. Compared to a cheep camera the fps is very low, the spatial resolution in x and y is pathetic 128. in the vertical, higher horizontal but its not mega pixels. Finally the dynamic range is 1 bit(something is there or not). Lidars use near infrared and are just as susceptible to problems with natural fog (not the theatrical fog like in that Roper video) and rain. Multiple cameras can do good enough depth estimation with modern neural networks. But cameras are vastly better at making sense of the world. You can’t read a sign with lidar.
briandw··on Protobuffers Are Wrong (2018)
The crappy system that everyone ends up using is better than the perfectly designed system that's only seen in academic papers. Javascript is the poster-child of Worse is Better. Protobuffs are a PITA, but they are widely used and getting new adoption in industry. https://en.wikipedia.org/wiki/Worse_is_better
briandw··on MIT Study Finds AI Use Reprograms the Brain, Leading to Cognitive Decline
This is standard response to any new technology. Socrates call books the death of knowledge, in the 19th century there was a moral panic about girls reading novels etc etc.
briandw··on SpaceX's giant Starship Mars rocket nails critical 10th test flight
A starship mission to the moon will land over 100tons of cargo. Saturn V could get roughly 5tons to the surface. Its an entirely different class of operation.
briandw··on ADHD drug treatment and risk of negative events and outcomes
This was an observational study, not a randomized control trial. Access to care is one plausible confounder. Individuals who receive treatment may differ systematically in socioeconomic status, healthcare access, parental advocacy, comorbidities—from those who do not, affecting outcomes.
briandw··on "None of These Books Are Obscene": Judge Strikes Down Much of FL's Book Ban Bill
Very few people that I talk to are truly in favor of free speech as a principle. They are either in favor of free speech in cases like this, where christians are banning books that they like, and not in favor when it’s about political correctness (misgendering, banning words, hate speech etc) or vice versa. Free speech for me and not for thee seems to order of the day.
briandw··on Learning Is Slower Than You Think
I wasn't explicitly quoting anything, but I'm sure I've read something similar somewhere.
briandw··on Learning Is Slower Than You Think
I home schooled my kids during the pandemic. It was amazing how quickly we got through the material for the year. We did all of 5th grade math in 3 months of 40 minutes a day. It not just my experience, 1 on 1 tutoring has been shown to be dramatically more effective than classroom instruction.

This article sounds like the usual ideological objections, lots of vague claims that amount to “I don’t like it”.

There is nothing more threatening to a failing institution than a solution that delivers results.

The US education sector is quick to embrace any new fad that sounds good but doesn’t work, building thinking classrooms is the latest. Productive struggle is another.

Yes learning takes time, but it doesn't have to be painfully slow and unproductive.

The education system in US is a disaster and getting worse. The response from schools like the San Francisco school district has been to lower standards and remove higher level material.

I see tremendous potential in Ai tutoring. I use chatgpt to help me learn new material daily. Why should school be any different?

briandw··on It's time for modern CSS to kill the SPA
Sounds like the major complaint here is trying to make the web into something it's not. It's something like 17 years since the iPhone app store. Web apps were not comparable to native apps then, and while they've made progress, web UX is still much poorer than a good native app. The Youtube mobile app has its problems but is so much better than the web interface.

Native and web have different strengths, that's ok.

briandw··on Exhausted man defeats AI model in world coding championship
Guess I should read more than the summary :)
briandw··on Exhausted man defeats AI model in world coding championship
This is a real modern day John Henry story, except John Henry dies in the end.

https://en.wikipedia.org/wiki/John_Henry_(folklore)

briandw··on Grok 4 Launch [video]
O3 was just reduced in price by 80%. Grok4 is a pretty good deal for having just been released and being so much better. The token price is the same as grok 3 for the not heavy model. Google is loosing money to try and gain relevance. I guess i’m not sure what your point is?
briandw··on Grok 4 Launch [video]
Grok 4 helped me solve a problem with inconsistent behavior in running lldb via python. Had differences in docker and my local linux box. Turns out to be a differences in how address sanitizer works in the slightly different environments. O3 didn’t catch it. So far i’m impressed.
briandw··on Why LLMs Can't Write Q/Kdb+: Writing Code Right-to-Left
This is something that diffusion based models would capable of. For example diffusion-coder https://arxiv.org/abs/2506.20639 Could be trained on right to left, but it doesn't seem like they did.
briandw··on Thesis: Interesting work is less amenable to the use of AI
Don't think what I described was ping-ponging. But if you want to see it that way, go ahead.

To clarify my process. 1) I have a problem in a new domain that I'm stuck on. 2) I work with the LLM to discuss my problem, think about solutions, get things to try. Not unlike StackOverflow or digging through documentation. However this process is much faster and I learn more without being called stupid by random people on SO (or HN). 3) The problem is fixed and I move on, or back to 1 or try something else.

The value here is that I have a problem to solve and I'm seeing it through to the end. I know what good looks like and have the agency and attention span to get there. The LLM doesn't and likely won't for quite some time.

briandw··on Thesis: Interesting work is less amenable to the use of AI
Just like anything else, stackoverflow, advice from a coworker or expert. If it doesn’t work, it will become clear that it’s not fixing your problem.
briandw··on LLMs should not replace therapists
If the llm could convince people to go to the gym, it would already be doing better than most therapists.
briandw··on Thesis: Interesting work is less amenable to the use of AI
I feel much more confident that I can take on a project in a domain that im not very familiar with. Ive been digging into llvm ir and I had not prior experience with it. ChatGPT is a much better guide to getting started than the documentation, which is very low quality.
briandw··on Can Large Language Models Play Text Games Well?
Have you seen https://www.aidungeon.com They started with GPT-2 in a google collab. You should put something together and try it, it's easier than ever to get a simple version of that working.
briandw··on Can Large Language Models Play Text Games Well?
Interesting to see but as the authors say a chat bot isn't trained to play text adventures. Instruction tuning doesn't seem to match the text adventure style very well. I think a very small bit of context engineering would allow it to play successfully. Reformatting past action response pairs from the history would certainly help, mostly to condense the context window and keep it from getting stuck taking about irrelevant topics. Also note that they used GPT-4 and not a reasoning model.
briandw··on CO2 sequestration through accelerated weathering of limestone on ships
Nuclear Explosions for Large Scale Carbon Sequestration: https://arxiv.org/abs/2501.06623 seems more promising, I love that there is more research and interest in accelerated weathering
briandw··on Tools: Code Is All You Need
Anyone else switch their LLM subscription every month? I'm back on ChatGPT for O3 use, but expect that Grok4 will be next.
briandw··on A proof-of-concept neural brain implant providing speech
Like in the book "Accelerando (Singularity)", the lobsters that get a protected sentience status
briandw··on Proton joins suit against Apple for practices that harm developers and consumers
Not on the Xbox
briandw··on Proton joins suit against Apple for practices that harm developers and consumers
I don't see people making the same arguments about Steam, Microsoft, Nintendo, etc, etc. Why is Apple special here? Other companies make stores that deliver application and developers don't get to tell them what to do. That's fine apparently, just not for Apple.
briandw··on The $25k car is going extinct?
my dyslexia strikes again
briandw··on The provenance memory model for C
The code blocks are very difficult to read on this page. I had ChatGPT O3 rewrite this in a more accessible format. https://chatgpt.com/share/68629096-0624-8005-846f-7c0d655061...
briandw··on The $25k car is going extinct?
They don't mention the amount of mandatory additions to the car from government regulations. Obviously not the only factor but it's certainly a factor. Cars now have 6 to 8 airbags, backup cameras, more high strength steel, automatic breaking systems etc etc. I love the safety that those all bring, but it's not free.
briandw··on More on Apple's Trust-Eroding 'F1 the Movie' Wallet Ad
The Apple of old had a deep respect for their users. We paid for a product that tried its best to sweat the details and deliver the best experience possible. UX was king. Apple made hard choices and delivered minimal, thoughtful and delightful products. The motto was "less but better".

Today we have an Apple that keeps pushing new poorly thought out features. More and more they don't respect the user. Constant interruptions that don't serve the user, a ridiculous onboarding process with far too many screens, forcing their own products like Apple Music on people, not making design choices and making the user pick an option. We are so far from less but better and it's only getting worse. I wish there was a way forward for Apple, but I think it's just going to slowly die.

briandw··on Honda conducts successful launch and landing of experimental reusable rocket
Rocket engines struggle to throttle down to low levels due to combustion instability, injector dynamics, and turbopump limitations. Here are some stats on minimum throttle levels:

SpaceX Merlin 1D: ~40% Rocketdyne F-1 (Saturn V): ~70% Space Shuttle Main Engine (RS-25): ~67% Blue Origin BE-4: ~20–25%

Falcon 9 does the "hover slam" where they have to turn off the engine exactly at touch down, or the rocket starts to go back up again. Throttle is too high for the weight of the booster at that point in flight.

← PreviousPage 6 of 9Next →