Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
rao-v
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
7 ms
·
121.
▲
by
rao-v
6mo ago
Is there a meaningfully useful version of automatically write to an encrypted disk / RAM that could be used with a random cloud instance? Obviously the decryption key would be in RAM somewhere but as a short term best practice it might
122.
▲
by
rao-v
6mo ago
This is simply not true - if you look at this article’s excellent benchmarking, linear search falls behind somewhere around 200-400 elements. In general I love this article, it took what I’ve often wondered about and did a perfect job explo
123.
▲
by
rao-v
6mo ago
I messed with this a bunch (still have a prototype floating around somewhere). Add a food weight signal with a Bluetooth scale and you’ll get a much much more grounded answers. Standardized the output format, soft match against nutritional
124.
▲
by
rao-v
6mo ago
Cap the power to 400W you won’t see much impact
125.
▲
by
rao-v
6mo ago
I doubt you'd get the same sort of result on a modern-ish MOE or dense model via a more standard inference engine like llama.cpp or VLLM. I don't think MLPerf is a reasonable benchmark at this point. Edit: Here is a simple llama.c
126.
▲
by
rao-v
6mo ago
As Albert Hirschman observed in reflecting on his seminal "Exit, Loyalty and Voice": "an organization needs minimal, or floor, levels of exit and voice in order to receive the necessary feedback about its performance". D
127.
▲
by
rao-v
6mo ago
This is news to me and genuinely impressive. Putting extra work into ensuring your attempt at one of the few records that will last the duration of humanity is damn smart.
128.
▲
by
rao-v
6mo ago
Alright, apart from Instagram, WhatsApp, Llama 1 & 2 and somehow managing to sell nearly 10M less nerdy google glasses what has Zuck done for FB?
129.
▲
by
rao-v
6mo ago
Paging Dr. Susan Calvin!
130.
▲
by
rao-v
6mo ago
Ah
131.
▲
by
rao-v
6mo ago
This always bothers me because models will almost never see text that is mostly English with a little other language in training data (opposite happens of course) and certainly not in RL data. Why do they occasionally language switch?
132.
▲
by
rao-v
6mo ago
Could somebody package Tylenol with a sufficient amount of NAC to de-risk it? I suspect such a formulation would not require trials?
133.
▲
by
rao-v
6mo ago
This would sell on Steam with a light Godot reskin
134.
▲
by
rao-v
6mo ago
I don’t really think this reflects the current era of challenges? The “enforcement layer” is the hardest and most important part, and is barely addressed. - is the answer structurally / syntactically valid? - is it appropriately ground
135.
▲
by
rao-v
6mo ago
My experience is that this is largely true only of the biggest ISPs in the US and Europe that were around in the 90s and have IPv4 in plenty. In other countries or even with other ISPs I see fairly frequent turnover. Agree it’s not the bigg
136.
▲
by
rao-v
6mo ago
The article does seem to simultaneously claim that IPv6’s design is the result of wierd no longer current pressures but also that it’s perfectly fine and correctly designed.
137.
▲
by
rao-v
6mo ago
At a high level one of the sad things about IPv6 is that it surrenders a wierd, valuable and emergent property of IPv4 for the average home user in $random_country: IPv4 addresses in logs are not super helpful in tracking a specific person
138.
▲
by
rao-v
6mo ago
That’s different from solo gather and cook and repair which is the artificially inflated bar being set here. I know people from multiple parts of the world who grew up on family working farms - specializing very real (esp. gender based)
139.
▲
by
rao-v
6mo ago
I don’t think in 1926 more than 50% of global 15-20 year olds could: “acquire food from nature (farm, hunt, gather) and cook it for themselves.” Or “make and fix their own tools. Build and fix their shelter.” (Culturally, those tasks often
140.
▲
by
rao-v
6mo ago
Tables are the one thing in markdown where I’d prefer to emphasize edit ergonomics over good looking unrendered text. Making a quick manual change like adding column to a markdown table is just unfun. I’ve always thought a json like format
141.
▲
by
rao-v
6mo ago
A very nice example, and more kind and welcoming than most. When would you say it reached it's near final form? Git blame suggest about 7 months ago or 3rd Q 2025?
142.
▲
by
rao-v
6mo ago
This is very good! It captures the core idea that "As a contributor, you are responsible for the code you submit" that the other example doesn't cover. Perhaps a nit is that it's a little too OpenInfra policy specific. I
143.
▲
by
rao-v
6mo ago
A phenomenon I can not explain is the fact that this simple clean statement of a fairly obvious approach to AI assistance somehow took this long and Linus to state so cleanly. Are there other popular repos with effectively this policy state
144.
▲
by
rao-v
6mo ago
Got to say I was hoping for a phone browser compatible demo
145.
▲
by
rao-v
7mo ago
I got to say people also seem to be missing really simple tricks with RAG that help. Using longer chunks and appending the file path to the chunk makes a big difference. Having said that, generally agree that keyword searching via rg and us
146.
▲
by
rao-v
7mo ago
Wait, so they don’t have test parity with git? How do they know that they, umm … did the actual thing they were trying to do?
147.
▲
by
rao-v
7mo ago
I'll give it a shot later today, but basically you need a pretty specific seccomp profile (see my example - I pulled from the podman repo) to allow bubblewrap to run inside an unpriviledged docker.
148.
▲
by
rao-v
7mo ago
Hey - I'd love for you to add a documented / standard way to use this inside dockers so we can use build on it for various agentic efforts. I've solved getting bubblewrap to work inside a docker once for the nanobot project,
149.
▲
by
rao-v
7mo ago
Wait to usefully import and export STEP you need to be BREP based right? I thought SCAD’s engine was fundamentally incompatible (only really one open source BREP engine out there - OpenCascade)
150.
▲
by
rao-v
7mo ago
I’d love to see this. It was a frustrating learning curve for me to realize that I couldn’t STEP export work from OpenSCAD to something like Fusion. Build123d is much better (supports STEP export and import) but a tightly integrated CAD fro
More ›