seanw265

8 ms·

31.
by seanw265 1y ago
How do you prevent bad actors from using your tools as a feedback loop to tune models that can evade detection?
32.
by seanw265 1y ago
Sure, React has overhead. No one disputes that. But pointing to a few red squares on a synthetic benchmark doesn’t explain the actual user experience on GitHub today. Their entire stack has evolved, and any number of architectural choices a
33.
by seanw265 1y ago
> Now they rewrite their frontend in react and it feels sluggish and slow. I find this is a common sentiment, but is there any evidence to find that React itself is actually the culprit of GH's supposed slowdown? GH has updated thei
34.
by seanw265 1y ago
Knowledge cutoff isn’t a big deal for current events. Anything truly recent will have to be fed into the context anyway. Where it does matter is for code generation. It’s error-prone and inefficient to try teaching a model how to use a new
35.
by seanw265 1y ago
Hey, I really liked the post and especially the title. Quite surreal but also very fitting at the same time. The writing was great too. Hope you keep going. I’d love to read more.
36.
by seanw265 1y ago
Coffee shops are fine for a change of scenery, but they get expensive fast if you're there all day. I also switched to a coworking space but mainly for these two reasons: 1. It’s much cheaper if you’re regularly working 8-hour days out
37.
by seanw265 1y ago
Last time I looked into Morph, I noticed you weren’t yet on OpenRouter. I see that’s changed, but it looks like only an older model is listed. Any plans to be more active there? Also, are there any benchmarks comparing your fast apply model
38.
by seanw265 1y ago
A word of caution on import maps. My app, Easel, previously relied on import maps for running user and AI-authored code in an iframe and it caused many many issues for us. Import maps combined with ESM CDNs (like jsDelivr, jspm, or ESM.sh)
39.
by seanw265 1y ago
I used to maintain inventory software used across dozens of cannabis retail locations, where every product had to be tracked “from seed to sale” to comply with strict state regulations. Every item had a barcode. Every movement had to be log
40.
by seanw265 2y ago
I was thinking less about self-imposed code reviews and more about regulatory frameworks—principles borrowed from architecture and construction, like mandated documentation, reviews, and inspections. There's some precedent for this: so
41.
by seanw265 2y ago
I attended the University of Delaware around the same time, where the CS honors program also started with Racket. As someone self-taught with experience in imperative languages like Obj-C, Java, and Haxe, most intro courses would have been
42.
by seanw265 2y ago
It's a fascinating comparison—I've seen this happen at companies too. Makes you wonder if imposing something akin to building codes for software development could prevent this kind of sprawling complexity.
43.
by seanw265 2y ago
I got charged by Squarespace the other day, and it immediately raised red flags—I've never done business with them before. Then it clicked: this was for an old domain I’d purchased through Google Domains. I knew Google had sold its dom
44.
by seanw265 2y ago
I received a 500 error when attempting to submit a product. I also attempted to log in (with Google) and it doesn't seem to have taken. I still see the "Login" button at the top of the page. I kind of regret giving you my dat
45.
by seanw265 2y ago
Related: there is a small college in NYC named "LIM", which used to stand for "Laboratory Institute of Merchandising." A few years back they updated their branding and now the name quite literally stands for nothing. T
46.
by seanw265 2y ago
Based on the title, I was expecting this to be about quines [1]. If you aren't familiar, quines are programs which produce their own source as their only output. They're quite interesting and worth a dive if you haven't explo
47.
by seanw265 2y ago
Tangentially related, I was once handed a single PDF between 2 and 5 GBs in size and asked to run inference on it. This was the result of a miscommunication with the data provider, but I think it's funny and almost impressive that th
48.
by seanw265 2y ago
I'm having trouble finding table mode in the demo you linked. Where can I find it?
49.
by seanw265 2y ago
I imagine that's orders of magnitude more expensive than paying these rates for an exploit.
50.
by seanw265 2y ago
Your claim that 40% of architects piqued my curiosity. I wonder if this would have an impact on the success of tools like ChatGPT in the architecture industry. Do you have a source for this stat? I can't seem to find anything to suppor
51.
by seanw265 2y ago
Is it? I'm not so familiar with the specifics of bug bounty programs, but it seems like this issue could cause much more than 3k in damages if it were to be exploited. Similarly, I'm kind of shocked that Google is only offering 30
52.
by seanw265 3y ago
I'm not so familiar with Signal, but could you explain why you would expect Signal notifications to happen out-of-band with normal push notifications? Assuming Signal sends push notifications of some sort, as most messaging services do
53.
by seanw265 3y ago
Can't comment on whether it works or not, but it sounds like there's a mix-up here. On iOS, it would be their bubble that is either blue or green when messaging you, not yours. To an iOS user, everyone else is a grey bubble.
54.
by seanw265 3y ago
I like a lot of what Deno does, but I'm having some trouble understanding why this (along with Deno's KV and Queues offering) is part of the Deno runtime. The blog post touts this as a feature to reduce the amount of code to get u
55.
by seanw265 3y ago
As other commenters have pointed out, accuracy is not a great metric to use here. A confusion matrix [1] is needed to fully capture the results. I'm surprised that one was not prominently displayed in this article. They are standard pr
56.
by seanw265 4y ago
What do you say to potential customers who are concerned with the legality and privacy challenges associated with session replay? I recognize how tools like this could level up product/marketing-level decision-making. But with litigati
57.
by seanw265 4y ago
This was very helpful. Thank you!
58.
by seanw265 4y ago
Please see my follow-up comment with another simulation that shows how the expected score can rise well above .296. https://news.ycombinator.com/item?id=33025759
59.
by seanw265 4y ago
Diving into this further, I've found another situation that appears to interfere even more. Our baseline (once again): Agent 1 (nash distribution): .296 Agent 2 (nash distribution): .296 Agent 3 (nash distribution): .296 Another cas
60.
by seanw265 4y ago
There must be something I'm missing. I borrowed the author's choice probabilities and wrote a script to verify the probability of winning by simulating 10 million rounds of gameplay. It seems it's true that when all three age

More ›