pants2

10 ms·

211.
by pants2 7mo ago
If the hardwired chips are magnitudes faster couldn't they be manufactured on an older process and still be competitive?
212.
by pants2 7mo ago
You might be surprised, with NVMe swap 8GB is surprisingly capable. ~1.6GB/s Read/Write.
213.
by pants2 7mo ago
Nice! Is there a similar option for Logitech Webcams?
214.
by pants2 7mo ago
The privacy angle is not that interesting to me. - You can find inference providers with whatever privacy terms you're looking for - If you're using LLMs with real data (let's say handling GMail) then Google has your data any
215.
by pants2 7mo ago
This really highlights the impracticality of local models: My $3k Macbook can run `GPT-OSS 20B` at ~16 tok/s according to this guide. Or I can run `GPT-OSS 120B` (a 6X larger model) at 360 tok/s (30X faster) on Groq at $0.60/
216.
by pants2 7mo ago
This itself is against the rules: > Please respond to the strongest plausible interpretation of what someone says > Please don't post shallow dismissals Personally I've posted comments with glaring typos that everyone thankf
217.
by pants2 7mo ago
Religion always seems like the default explanation for anything without an obvious use and it seems lazy. Maybe it was a game, a rite of passage, a boundary marker, or perhaps there was a Peruvian Mr. Beast running a competition. Anyone els
218.
by pants2 7mo ago
I understand Amazon wants to protect their customer experience / advertising potential and avoid being commoditized, ie. "hey perplexity order me the cheapest option between Amazon and 17 other online retailers." However I th
219.
by pants2 7mo ago
Or how about the curb weight of the car? Higher mass means you're doing a lot more damage in an accident. People might think twice about buying an F250 for their grocery getter.
220.
by pants2 7mo ago
Three hundred dollars per hour? How is that acceptable? Certainly there's some sort of maximum or grace period?
221.
Why is GPT-5.4 obsessed with Goblins?
17 points by pants2 7mo ago|9 comments
222.
by pants2 7mo ago
The HN post[1] eight months ago about the Scale acquisition is full of comments questioning Zuck's choice to have Wang lead their AI division. I guess HN was right this time. 1. https://news.ycombinator.com/item?id=4426
223.
by pants2 7mo ago
"New jobs will be created that we didn't even know could exist"
224.
by pants2 7mo ago
San Diego is really a hub for health wearables! Tijuana to an extent too. Oura - based in SD. Dexcomm - based in SD. Omnipod/Insulet - major R&D hub in SD & TJ.
225.
by pants2 7mo ago
Wow, and Google's response to this was "unfortunately AI models are not perfect" That's a bit worse than 'imperfect'
226.
by pants2 7mo ago
Can't imagine what one needs more than 16GB for unless it's local LLMs. I regularly do front end dev while I'm editing 10-bit 4K60 footage in Da Vinci Resolve, runs smooth as butter.
227.
by pants2 7mo ago
TikTok has a drug-like effect on the brain. Multiple studies show a clear link between excessive TikTok engagement and increased levels of anxiety, depression, and stress. Maybe it is time we regulate it like a drug?
228.
by pants2 7mo ago
You can put pickle juice in your cereal too
229.
by pants2 7mo ago
Curious, why was it $20? I would think that's close to an hourly rate for first level support and calls are mostly resolved in ~10 mins?
230.
by pants2 7mo ago
I agree -- we're on the ChatGPT Enterprise plan at work and every time someone complains about it screwing up a task it turns out they were using the instant model. There needs to be a way to disable it at the bare minimum.
231.
by pants2 7mo ago
When I joined my last job I noticed that their email settings were misconfigured... EVERYTHING was going straight to the inbox, not even the most basic of spam filters were in place. When I got filtering on observe-only mode I saw users wer
232.
by pants2 7mo ago
No need for Reddit, their status page has it just fine: https://status.claude.com/
233.
by pants2 7mo ago
Certainly not the "one way to write it" idea. Java has a ton of language features.
234.
by pants2 7mo ago
I remember the total uproar about the name! Everyone said it was too close to a Feminine hygiene product, or too close to iPod. Nobody's complaining any more.
235.
by pants2 7mo ago
Indeed - I remember MKBHD praising the initial OnePlus thanks to that approach. Today he's got 20M subscribers. If Motorola gets big tech reviewers on their side it's a big market. Not to mention the more techy people in a family
236.
by pants2 7mo ago
No, but for industrial and business uses yes. You'd start to see a lot more robots used for warehouses, deliveries, restocking, cleaning, preparing food, etc.
237.
by pants2 7mo ago
I use them heavily in production because OSS 120B on Groq/Cerebras is one of the fastest models available with enough smarts for basic tasks
238.
by pants2 7mo ago
OpenAI actually does have two excellent OSS models. Not Anthropic. Not that OpenAI is 'open' per se, but more so than Anthropic. Also see the Codex vs Claude Code extensibility.
239.
by pants2 7mo ago
Ghostty calls itself "feature rich" but only added cmd+F / find functionality a few months ago. Makes me wonder what other basic functions it's missing.
240.
by pants2 7mo ago
very avant-garde

More ›