Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
the_arun
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
14 ms
·
91.
▲
by
the_arun
10mo ago
I wish there are tests written to cover functionalities (& race conditions) instead of covering lines/branches.
92.
▲
by
the_arun
10mo ago
I liked the thread sharing feature of BluSky.
93.
▲
by
the_arun
10mo ago
Yep, we are part of the system.
94.
▲
by
the_arun
10mo ago
When you read this story - your heart warms & your eyes gets filled. It is crazy nice feeling. You feel like this world is such a better place. Yet, it hurts - to know there are so many homeless that our system needs corrections.
95.
▲
by
the_arun
10mo ago
Isn’t Open Policy Agent (OPA) and Zanzibar not good enough to be in the article or author talking about specific permission controls?
96.
▲
by
the_arun
10mo ago
Curious - How did you find the image is not optimized? Is there a tool to find it?
97.
▲
by
the_arun
10mo ago
They also have “promotions” tab listing all promo content. I wonder is this real or mock data.
98.
▲
by
the_arun
10mo ago
I felt like changing HN down page to show top 30 posts from this week before or after the generic message.
99.
▲
by
the_arun
10mo ago
Just curious - if we remove stop words from prompts before going to LLM, wouldn't it reduce token size? Will it keep the response from LLM same (original vs without stop tokens)?
100.
▲
An MCP that lets you play DOOM in ChatGPT
(old.reddit.com)
1 points
by
the_arun
10mo ago
|
0 comments
101.
▲
by
the_arun
11mo ago
Without paywall - https://archive.is/Eiedt
102.
▲
by
the_arun
11mo ago
How does a family of 3 or 4 use this room according to this logic?
103.
▲
by
the_arun
11mo ago
I didn't think he is a bad person. I felt he is not brave enough to take the challenge.
104.
▲
by
the_arun
11mo ago
Great project! What I liked most is how you credited every individual along the way in the article. That shows how authentic you are. Kudos!
105.
▲
by
the_arun
11mo ago
They do good marketing for sure. I would like to see same from Toyotas & Hondas of the world.
106.
▲
Async QUIC and HTTP/3 made easy: Tokio-quiche is now open-source
(blog.cloudflare.com)
2 points
by
the_arun
11mo ago
|
1 comments
107.
▲
by
the_arun
1y ago
4 Crashes in a drive of 250k miles. How good/bad is this with human driving? I couldn't compare this with human driving. Though I found a report on fatal crashes here - https://www.iihs.org/research-areas/fata
108.
▲
by
the_arun
1y ago
Also decline in Security engineer by 0.35% doesn't make sense by conventional wisdom. Shouldn't it be increasing due to increased demand for security in all ai integrations?
109.
▲
Benchmarking My Laptop
(shishir-iyer.medium.com)
1 points
by
the_arun
1y ago
|
0 comments
110.
▲
by
the_arun
1y ago
It is a good analogy.
111.
▲
by
the_arun
1y ago
I think diagrams are opinionated. I’d like to show animations instead of just colors. Data flow for example. Yes arrows help, but animations help better. It is all about personal preference. There is no right or wrong.
112.
▲
by
the_arun
1y ago
Future - Is it possible to run swift built OS to run on Android devices 100%? Is that a threat to Android as an OS?
113.
▲
by
the_arun
1y ago
I think Sphere is a outlier for this context.
114.
▲
by
the_arun
1y ago
At that height if windows are damaged enough to hurt captain or pilot, would the flight lose balance because of air coming in? How did they land in that situation? There is no mention of that in the article.
115.
▲
by
the_arun
1y ago
I was content even without AI. I’m good with whatever we have today as long we use them to change the life in a positive way.
116.
▲
by
the_arun
1y ago
What is the cost of running Neo4j on aws vs using aws Neptune? Related to disk I/o?
117.
▲
by
the_arun
1y ago
Shouldn’t we compare profit? Instead of revenues?
118.
▲
by
the_arun
1y ago
That is what I am looking for. a) LLMs are trained using compressed text tokens and b) use compressed prompts. Don't know how..but that is what I was hoping for.
119.
▲
by
the_arun
1y ago
I wish there was a way to send compressed context to LLMs instead of plain text. This will reduce token size, performance & operational costs.
120.
▲
by
the_arun
1y ago
How about some policies to ensure customers have the control to see what they want to see? For example - I don’t support to see ads ahead of actual search results?
More ›