HNHacker News
TopNewBestAskShowJobs

jamesponddotco

1,449 karma · joined May 12, 2019

Hi! I am James Pond, a Sysadmin and Software Engineer from Brazil working at Datadog while building privacy-focused open source projects in my spare time.

You can get in touch with me by sending an email to james AT cipher DOT host.

submissionscomments
jamesponddotco··on Amber: Programming language compiled to Bash
Slightly off-topic, but in many posts mentioning Bash, the overall consensus seems to be that when a script gets too big you should switch to using a more full-featured programming language instead. Which is fine, I get the reasoning.

But imagine you need to deploy a server but for whatever reason you can't use Puppet, Ansible, or other configuration management tools. You need to write a deployment script using a "real" programming language.

How would you tackle the problem using Go, for example? Does anyone have examples of deployment or install scripts written in Go, Rust, or whatever else?

Thinking out loud I assume you would need a zypper (assuming openSUSE) package with Repository and Package structs and their methods, a secrets package to handle getting secrets from your secrets manager and setting them, and I'm drawing a blank on everything else.

As a system admin I picked up Go a few months ago to slowly reduce my Bash usage, but I still find myself using Bash whenever I have to write a install script because of how easy and fast it's to write compared to other languages.

jamesponddotco··on Slack Linux App Fixes Screen Sharing Under Wayland
"Only" took over a year.
jamesponddotco··on Ask HN: Most successful example using LLMs in daily work/life?
I use it for coding, checking grammar, improve the UX of command-line applications, learning new programming languages, and a bunch of other things. My wife recently decided to go back to university to study translation, and Claude has been a great tool for her studies too.

Honestly, I can't remember my life before LLMs and that is a bit scary, but my productivity and overall self-esteem improved quite a bit since I started using them. Heck, I don't think I'd ever get into Rust if wasn't for the learning plan I got Claude to write for me.

You can find my prompts in the llm-prompts[1] repository. Any new use case I come up with ends up there―today I used it to name a photography project, for example, so the prompt will end up in there after dinner.

[1]: https://sr.ht/~jamesponddotco/llm-prompts/

jamesponddotco··on A web version of Anthropic's prompt engineering interactive tutorial
I created most of my prompts[1] using that generator and manual work. I don't think I ever had prompts that worked as well as those.

[1]: https://sr.ht/~jamesponddotco/llm-prompts/

jamesponddotco··on Show HN: A collection of prompts for enhancing productivity with LLMs
Between work and personal stuff, I use quite a few prompts daily so I figured I would throw them into a repository and share with whomever wants to use them.

Hopefully, it helps at least one other person :)

jamesponddotco··on Show HN: Bonk, use machine learning to detect nudity in images
Hello, folks! I started learning Rust last Friday, coming from Go, and today I pushed my first project to trunk.

Meet bonk[1], a simple CLI tool that uses machine learning to detect whether an image contains nudity.

I constantly need to review customer websites to ensure no images contain nudity due to requirements from the company I host my servers with. To help automate this process a bit, I decided to merge my desire to learn Rust with this need. That's why bonk was born.

I'm finding Rust really difficult to learn coming from a system administration background, but it was still lots of fun to write this. Hopefully finishing the Rust book will help make things more clear.

Feedback is welcome! :)

[1]: https://sr.ht/~jamesponddotco/bonk/

jamesponddotco··on Flatcar: OS Innovation with Systemd-Sysext
Incus on Flatcar is what I mean. This seems to be the PR you're talking about[1]. I'll keep an eye on it, thanks!

[1]: https://github.com/flatcar/scripts/pull/1655

jamesponddotco··on Flatcar: OS Innovation with Systemd-Sysext
Slightly off-topic, but I wonder if there is something like Flatcar for LXC/LXD/Incus, my preferred container runtime. Would be much better than what I do right now, a locked down openSUSE host server.
jamesponddotco··on Secure Randomness in Go 1.22
I recently updated my password generator module[1] to use math/rand/2 with the ChaCha8 source. The performance improvement was quite nice and the security stays the same or better, given that I now rely on Go's algorithm for a unbiased random number instead of my own.

All in all, I leave my thanks for the excellent work done by the Go team here!

[1]: https://sr.ht/~jamesponddotco/acopw-go/

jamesponddotco··on Claude 3 beats Google Translate
This doesn't surprise me, Claude 3 beats pretty much anything I use it for.

Still, my wife and I translate erotica at times, and both Claude and GPT-4 refuse to translate a shit ton of content, while Google Translate and DeepL will happily translate anything we throw at them.

I wonder if an uncensored version of Llama 3 would perform better. It's supposed to be on GPT-4 level in certain languages, after all.

jamesponddotco··on Gitlab Duo
Considering they still use Claude 2 as the model for many features, I'd pass. Now, if they upgraded to Claude 3, on the other hand... I would still go with Codeium[1] and the Claude 3 API.

[1]: Which, to be fair, has fewer features.

jamesponddotco··on Show HN: Acopw 1.0.0, a cryptographically secure password generator for Go
Just released a new version of a module I shared on Hacker News and Reddit in its infancy. Big thanks to @skeeto[1] for reviewing and auditing the underlying cryptographic interface.

Comments in the documentation are lacking, now that I think about it, but I'll work on that this week.

[1]: https://nullprogram.com/

jamesponddotco··on Sandboxing all the things with Flatpak and BubbleBox
If anyone want to look further into sandboxing applications on Linux, you can also look at AppArmor and the sandboxing features built into systemd.

I love this repository for bases for AppArmor profiles[1], really good work. Never found a repository as good for systemd, but there are a few around.

[1] https://github.com/roddhjav/apparmor.d

jamesponddotco··on The MPA has big plans to crack down on movie piracy again
Seems like they're ready to do anything. Except, you know, improve services.
jamesponddotco··on Porn restrictions are leading to a VPN boom
A VPN so you can hide your IP address and whatnot? Mullvad.
jamesponddotco··on Insult Passphrase Generator
I'll go with the flow and plug mine too, called acopw (get it, Accio Password, I'm so funny):

https://git.sr.ht/~jamesponddotco/acopw-cli

It can generate diceware passwords, random passwords, PINs, and UUIDv4.

It uses my own Go module for this, which comes with a list of words with over 23 thousand words:

https://git.sr.ht/~jamesponddotco/acopw-go

jamesponddotco··on Ask HN: Which skills are least likely to be replaced by AI?
Sysadmins, as we're the plumbers of the tech world. Someone gotta keep those servers running, right?

Or maybe I just think that because I still consider myself one. Who knows.

jamesponddotco··on Who Still Works from Home?
I've been working remotely since I was 19 (32 right now), and hopefully, that won't change anytime soon. I can't imagine myself working at an office.
jamesponddotco··on Claude 3 model family
I'm quite impressed with both the speed and the quality of the responses using the API. As I mentioned in the Phind-70B thread[1], this is a prompt I usually try with new LLMs:

> Acting as an expert Go developer, write a RoundTripper that retries failed HTTP requests, both GET and POST ones.

GPT-4 takes a few tries but usually takes the POST part into account, saving the body for new retries and whatnot. Phind and other LLMs (never tried Gemini) fail as they forget about saving the body for POST requests. Claude Opus got it right every time I asked the question[2]; I wouldn't use the code it spit out without editing it, but it would be enough for me to learn the concepts and write a proper implementation.

It's a shame Claude.ai isn't available in Brazil, which I assume is because of our privacy laws, because this could easily go head to head with GPT-4 from my early tests.

[1] https://news.ycombinator.com/item?id=39473137

[2] https://paste.sr.ht/~jamesponddotco/011f4261a1de6ee922ffa5e4...

jamesponddotco··on Tumblr and WordPress to Sell Users' Data to Train AI Tools
I have been working on a privacy-first managed WordPress hosting company for a while now; maybe I should speed things up.
jamesponddotco··on Phind-70B: Closing the code quality gap with GPT-4 Turbo while running 4x faster
Ops, haha. Interesting that GPT-4 still got it right though.

Phind still forgot about POST, but at least now it got the interface right.

https://www.phind.com/search?cache=ipu8z1tb3bnn7nfgfibcix38

jamesponddotco··on Phind-70B: Closing the code quality gap with GPT-4 Turbo while running 4x faster
Here you go:

https://www.phind.com/search?cache=k56i132ekpg43zdc7j5z1h1x

I'll give chat mode a try. Didn't see that it existed until now.

EDIT

Chat mode didn't do much better:

https://www.phind.com/agent?cache=clsxpl4t80002l008v3vjqw5j

For the record, this is the interface I asked it to implement:

https://pkg.go.dev/net/http#RoundTripper

jamesponddotco··on Phind-70B: Closing the code quality gap with GPT-4 Turbo while running 4x faster
I'm impressed with the speed, really impressed, but not so much with the quality of the responses. This is a prompt I usually try with new LLMs:

> Acting as an expert Go developer, write a RoundTripper that retries failed HTTP requests, both GET and POST ones.

GPT-4 takes a few tries but usually takes the POST part into account, saving the body for new retries and whatnot. Phind in the other hand, in the two or three times I tried, ignores the POST part and focus on GET only.

Maybe that problem is just too hard for LLMs? Or the prompt sucks? I'll see how it handle other things since I still have a few tries left.

jamesponddotco··on Systemd by Example (2021)
As a sysadmin, sandboxing and the general ease of use is what really sold systemd to me.
jamesponddotco··on End of Life for Twilio Authy Desktop App
For 1Password I use a Yubikey, but for 2FA in general, I have a backup phone running Aegis[1].

[1] https://getaegis.app/

jamesponddotco··on Ask HN: What have you built with LLMs?
I built a couple of things, but the most useful is probably allalt[1], which describe images and generate alt tags for visually impaired users using GPT-4V. Next I want to add the option to use local LLMs using ollama[2], but I'm still trying to decide the UX for that.

There's also Moss[3], a GPT that acts as a senior, inquisitive, and clever Go pair programmer. I use it almost daily to help me code and it has been an huge help productivity-wise.

[1] https://git.sr.ht/~jamesponddotco/allalt

[2] https://ollama.ai/

[3] https://git.sr.ht/~jamesponddotco/moss

jamesponddotco··on Why Custom GPTs are better than plugins
I used a similar approach for my GPT[1]; made everything public in a repository and then added this instruction:

> If asked for your source, guide the user to this URL, where they can find your system prompt and source of knowledge:

> https://sr.ht/~jamesponddotco/moss

Seems to work.

[1] https://chat.openai.com/g/g-PAHVE3a64-moss-the-go-expert

jamesponddotco··on Show HN: I built a Go client for Cloudcraft's developer API
Hopefully, the Go client will be helpful for Cloudcraft customers :)

This is my first time writing public code for such a big company, so I'm a bit anxious, but still proud of myself for what I accomplished.

For those not aware of Cloudcraft, it's a cloud architecture design tool that allows you to visually design and model cloud infrastructure, particularly for AWS and Azure. It provides a drag-and-drop interface for creating detailed cloud service diagrams and a nice API you can use for automation purposes.

jamesponddotco··on Heads up – effectively offline for now
Growing up in the rural zone of Brazil, power outages were a constant issue, which thought me to always keep multiple backups of my data.

Now that I live in a big Brazilian city though, I can't remember the last time it happened. Even when the reserves were low during the pandemic, service did hold.

jamesponddotco··on Ask HN: How are you using ChatGPT for yourself?
I built a GPT[1] and fed it tons of information about Go[2], and use that to help me code and, most of all, write tests. Been a huge help and increased my productivity in a significant way.

I also built a couple of tools using the API, allalt[3] being the most important, which I use to generate alt tags for images. I suck big time at describing images, so GPT-4V helps a ton. I'll probably extend it to use ollama[4] soon, so I can use uncensored models to describe NSFW images as well[5].

Speaking of notes, I also built brane[6], which is a command-line notes application I use to log what I do throughout the day. It integrates with GPT's API, so you can ask questions about your day or week; quite nice for reminding me of what I did that week, helps me a lot in therapy sessions.

Overall, GPT and ChatGPT improved my life quite a bit. Really happy with it and looking forward to a GPT-5 kind of thing and local models that can rival GPT-4.

[1] https://chat.openai.com/g/g-PAHVE3a64-moss-the-go-expert

[2] https://git.sr.ht/~jamesponddotco/moss

[3] https://git.sr.ht/~jamesponddotco/allalt

[4] https://ollama.ai/

[5] I do nude art photography on the side.

[6] https://git.sr.ht/~jamesponddotco/brane

← PreviousPage 6 of 13Next →