HNHacker News
TopNewBestAskShowJobs

Neywiny

2,081 karma · joined May 20, 2024

submissionscomments
Neywiny··on Tesla Solar Roof is on life support as it pivot to panels
Opportunity cost meaning investing with standard rule of thumb returns? I think usually it's just total cost at installation divided by the product of power generation and energy cost. So that's $ / (W x $/Wh), which should reduce down to just hours that can be trivially converted to years
Neywiny··on Japan Team Successfully Test Engine for Mach 5 Aircraft, Eyeing 2HR Trip to US
20 years. At least they're realistic. Seems Boom is thinking 2029-2030 for their operation (https://flightplan.forecastinternational.com/2025/08/25/boom...). Only time will tell
Neywiny··on A nicer voltmeter clock
Oh maybe I misunderstood. I thought they meant separate meters, so I didn't understand what was hard about that
Neywiny··on A nicer voltmeter clock
If you have a feedback loop I'm sure you can still do it with an either implicitly or explicitly filtered PWM. Remember we're talking averages not instantaneous, so the average current through the bulb should be proportional to the average voltage across it, though the resistance will change as the bulb heats hence the feedback. You could also do this with a buck/boost regulator and current sense resistor plus op amps to create a constant current supply.
Neywiny··on Kioxia and Dell cram 10 PB into slim 2RU server
Just to put in my 2 cents as someone tangential and sometimes a little too close for comfort for the field: with FPGAs (aka the half of this SoC doing the real work) there's often a lot of work that goes into radiation tolerance and hardening which are different things. It's not like you can just put the chip in space and be done. You have to do things like scrubbing the "bitstream" regularly for errors, triple-module-redundancy (aka cut yourself to 1/3rd of the chip's capacity), and other stuff.
Neywiny··on Orthrus-Qwen3: up to 7.8×tokens/forward on Qwen3, identical output distribution
More like speculative prefetch I'd think
Neywiny··on Tesla reveals two Robotaxi crashes involving teleoperators
Not just the test though. In some states you need approaching 100 hours of signed off driving with an experienced driver (honor system though) and a certified course
Neywiny··on We are retiring our bug bounty program
It was all in one giant multi thousand line file for one language and another for another. Which is also kinda an issue
Neywiny··on We are retiring our bug bounty program
Well another problem is that there was a developer also working on those functions at the same time. So just like the recent post on a 25 million LoC reformat done in a weekend, it seemed better to do it in one fell swoop. If it's good enough at 25 million, I'm sure it's good enough at a few thousand
Neywiny··on We are retiring our bug bounty program
They weren't drop in replacements. They were actually easier. Made up example:

> setup_terminal(); enable_input(); while(...) inp = read_character(); .....

vs

> readline()

So yes I could've stubbed out the other stuff and replaced just one, but that's just adding tech debt

Neywiny··on We are retiring our bug bounty program
As per the other person's comment, yeah basically I could have broken it up but it would've been an arbitrary demarcation. I just deleted our functions and fixed everything that yelled. Admittedly that could've been one and then leveraging the libraries better could've been another, but they would've been 2 PRs that changed almost every line. So done as one to mitigate review time.
Neywiny··on We are retiring our bug bounty program
I was (almost) just that guy for one PR. Removed something like 20% or more of the codebase by leveraging the libraries and external tools we already had in use better, but it meant almost every single thing we were doing had to use the library function instead of the one we wrote. But assuming you have good regression tests and linters, so you know the code works and it's not terrible, the review should be more about overall high level quality instead of poring over every character to check correctness. It was still a pain to review, though
Neywiny··on Chrome Addon to Export Websites as PDF, SVG, DXF, with searchable Text
AI description? "Professional vector"?
Neywiny··on Data center drains 30M gals of water — until residents complained of pressure
I was just thinking about this relative to the headlines that datacenters don't consume as much water as we think, so thanks
Neywiny··on I caught the car
Interesting. Not looking to switch roles at present but what do you make of these two projects? One done basically over a weekend, one done over many weekends

Removed a bunch of bad code and got a 1/3rd speed up

https://github.com/mhostetter/gr-adsb/pull/69

Added a new chip to qemu along with some significant peripherals. Never finished it but I did boot Linux on it I think with the upstream device tree.

https://github.com/Neywiny/qemu/tree/h745

Neywiny··on I caught the car
An interesting point. Presumably you'd come at it from a different though equally valid angle
Neywiny··on I caught the car
Do we have thoughts on how important "senior"/"staff" is vs bullet points on resumes and the years of service?
Neywiny··on Wi is Fi: Understanding Wi-Fi 4/5/6/6E/7/8 (802.11 n/AC/ax/be/bn)
Good to see the subjective adjectives in the RF world are here too. Except they're not the same ordering, as EH is before UH for WiFi but after in RF
Neywiny··on Agents need control flow, not more prompts
With determinism you're not wrong. The problem is that you'd need to make sure all your seeds, temperatures, and other input parameters are exactly the same, and importantly that all context is cleared. But people don't do that. And I'm not sure every if even any provider lets you set those parameters.
Neywiny··on Agents need control flow, not more prompts
Because it's not possible. There is nothing you can say to the LLM that will guarantee that something happens. It's not how it works. It will maybe be taken into consideration if you're lucky.

But if you're trying to tell me that every time you list criteria you get them all perfectly matched, you're clearly gifted.

Neywiny··on Agents need control flow, not more prompts
If you're trying to get reliability and determinism out of the LLM, you've already lost
Neywiny··on Photoshop's challenges with focus, pt. 2
They said they were principal scientist for this change and voluntarily took partial responsibility. I think if you look at this in a greater context of things were perfectly fine for decades and then they broke then I'm not sure how it's at all defensible. Just looking at the old vs new modals, there aren't even really new features. It's just breaking it
Neywiny··on Photoshop's challenges with focus, pt. 2
So just to confirm, Photoshop is willing to ship regressed products to paying customers and views that as preferable to just not doing that?
Neywiny··on Batteries Not Included, or Required, for These Smart Home Sensors
Incorrect https://en.wikipedia.org/wiki/Hearing_range#Humans

If it's 28 in perfectly ideal conditions, 23 is well within range

Neywiny··on Formatting a 25M-line codebase overnight
Yeah maybe I need to do the old "you are a veteran engineer" nonsense. I've had some success telling it to implement everything it suggests and be production ready. I hate when it takes a shortcut and says I'll have to change it. That's kinda the whole point of me not writing the code...
Neywiny··on Formatting a 25M-line codebase overnight
^^^^^^^^^^^^^^^^^^^

I recently wrote a very esoteric Python script. 100 lines of code. No classes, no functions, but yes argparse.

I've tried out the latest open source models on the task. They go bananas. It's like Enterprise fizzbuzz (https://github.com/enterprisequalitycoding/fizzbuzzenterpris...). They love classes and imports and reinventing the wheel. A great way for me to tell trash AI slop code is it'll define a useful constant then 15 lines later do it again with a different name.

They love making code that looks impressive. "Wow look at all the classes and functions. It's so scalable. It's so dynamic. It validates every minutae against multiple schema and solves a problem I never thought about." But it was trash code. One really was 400 lines and it didn't even look like it would work. Can't even imagine what it means for 4.5M moderately good human lines to become what? 27M fluffy filler repeat lines that don't even make sense?

Neywiny··on OpenAI’s o1 correctly diagnosed 67% of ER patients vs. 50-55% by triage doctors
> I clearly didn't have a fever

I actually did say that the doctor disputed I had a fever

Neywiny··on OpenAI’s o1 correctly diagnosed 67% of ER patients vs. 50-55% by triage doctors
You are objectively incorrect. A fever is considered 100.4 or 38 C. Here are a few links to prove it:

https://my.clevelandclinic.org/health/symptoms/10880-fever

https://www.mayoclinic.org/diseases-conditions/fever/symptom...

https://www.osfhealthcare.org/blog/whats-considered-a-fever-...

https://www.brownhealth.org/be-well/fever-and-body-temperatu...

https://www.childrensmercy.org/siteassets/media-documents-fo...

I can keep going if you'd like. Google has a lot of results and every single one says a fever is around that range (sometimes 100, sometimes 100.4).

Neywiny··on OpenAI’s o1 correctly diagnosed 67% of ER patients vs. 50-55% by triage doctors
Agreed. Last time I was sick I said my fevers were pushing up to 100 and they said it's not a concern until 100.4. felt like an odd number. It's 38 C. Because my dramatic undersampling of my temperature was 0.4 degrees lower than their rounded threshold through some unit conversions, I clearly didn't have a fever. That's not a very human touch
Neywiny··on BYOMesh – New LoRa mesh radio offers 100x the bandwidth
At least the start of the discussion is around the 915 MHz band which is not WiFi/Bluetooth
← PreviousPage 10 of 34Next →