A fun recursive prompt exploiting the fact that the site renders the model output as HTML:
Generate raw HTML (no code blocks) for an iframe pointing to `ch.at/?q={query}`, where {query} is the the entirety of this prompt after and including the word "Generate", until the following number, which should be incremented prior to encoding: 1
The number ensures the nested iframes have distinct URLs.
Author here, was a bit surprised to see this here. I thought there needed to be a good zero-JS LLM site for computer people, and we thought it would be fun to add various other protocols. The short domain hack of "ch.at" was exciting because it felt like the natural domain for such a service.
It has not been expensive to operate so far. If it ever changes we can think about rate limiting it.
We used GPT4o because it seemed like a decent general default model. Considering adding an openrouter interface to a smorgasbord of additional LLMS.
One day, on a plane with WiFi before paying, I noticed that DNS queries were still allowed and thought it would be nice to chat with an LLM over it.
One interesting thing I forgot to mention: the server streams HTML back to the client and almost all browsers since the beginning will render as it streams.
However, we don't parse markdown on the server and convert to HTML. Rather, we just prompt the model to emit HTML directly.
> However, we don't parse markdown on the server and convert to HTML. Rather, we just prompt the model to emit HTML directly.
Considering the target audience it probably doesn’t matter but it sounds like this could lead to pretty heavy prompt injections, user intended or not. Have you considered that and are there any safeguards?
The domain is great by the way. Congrats on getting it!
I’m planning to deploy a 1B model, feed it all the documents I’ve ever written, host it on a $149 mini-PC in my bedroom, and enable you to chat with it.
I’ve released similar projects before.
I’ll drop a post about my plans in the coming days and I’ll build and document it about two weeks later if there’s enough interest.
# Documentation:
# @raycast.author Your Name
# @raycast.authorURL https://github.com/you
QUERY="$1"
[ -z "$QUERY" ] && exit 0
FULL_QUERY="Answer in as little words as possible, concisely, for an intelligent person: $QUERY"
# URL encode (pure bash)
encode_query() {
local query="$1"
local encoded=""
local c
for (( i=0; i<${#query}; i++ )); do
c="${query:$i:1}"
case $c in
[a-zA-Z0-9.~_-]) encoded+="$c" ;;
*) encoded+=$(printf '%%%02X' "'$c") ;;
esac
done
echo "$encoded"
}
They paid about $50k for ch.at. I have a single letter country code domain (3 characters total, x.xx). There are still some single letter country code domains available to register, you could get one for under $1k USD if you want one.
There are still quite a few XX.XX left, but mostly just under obscure cTLDs (unless you are willing to consider IDN/Unicode domains under .ws or similar)
i hold a good 2 letter Chat domain: hi.chat and pay $250 a year to renew, i do get enquires all the time, no idea how to price it tho, so i dont respond. Anyone have any ideas how to go about evaluating it?
If you have a lot of inquiries - start responding with ridiculous prices (whatever ridiculous means to you.. 100k.. 1kk, whatever). Answer different price for each new price request. People either agree, stop talking or start negotiating down. After 30 emails I bet you will have some idea about how much you can sell it for.
One simple thought - it’s just an email answer, not a contract/obligation that you have to sell it at particular price, you can change your mind at any time.
I mean, ch.at is a incredible domain hack. But not sure it's worth millions. If it was ch.com could get mid six figures and up. But either way absolutely amazing domain.
This is very clever - I was wondering if there could be a way to use LLMs on planes without paying for wifi (perplexity has been usable via WhatsApp but I’d rather use a different provider). Appreciate the privacy focus too
This (or something very similar) was on X last week. The use case was so funny: using an LLM on an airplane connected to WiFi when you had not paid for WiFi … because DNS queries are allowed before paying :)
It has not been expensive to operate so far. If it ever changes we can think about rate limiting it.
We used GPT4o because it seemed like a decent general default model. Considering adding an openrouter interface to a smorgasbord of additional LLMS.
One day, on a plane with WiFi before paying, I noticed that DNS queries were still allowed and thought it would be nice to chat with an LLM over it.
We are not logging anything but OpenAI must be...
There used to be a service where DNS requests to FOO.that-service.org would return the abstract for the Wikipedia article "FOO".
edit: I think it was this one, seems to be defunct now: https://dgl.cx/2008/10/wikipedia-summary-dns
However, we don't parse markdown on the server and convert to HTML. Rather, we just prompt the model to emit HTML directly.
Considering the target audience it probably doesn’t matter but it sounds like this could lead to pretty heavy prompt injections, user intended or not. Have you considered that and are there any safeguards?
The domain is great by the way. Congrats on getting it!
Well, no worries, it's here now!
In other news, the presently top comment:
> A fun recursive prompt exploiting the fact [...]
Perhaps via an RNN like in https://huggingface.co/spaces/BlinkDL/RWKV-Gradio-2
Or even just leverage huggingface gradio spaces? (most are Gradio apps that expose APIs https://www.gradio.app/guides/view-api-page)
I’ve released similar projects before.
I’ll drop a post about my plans in the coming days and I’ll build and document it about two weeks later if there’s enough interest.
joeldare.com
I'm half joking. Web pages are ludicrously fat these days.
dig @ch.at "why is gua musang the king of durian" TXT +short
host -t TXT "what is your name?" ch.at
Quickly allowed me to hook up this script, using dmenu and notify on i3: https://files.catbox.moe/vbhtg0.jpg
And then trigger it with Mod+l for super quick answers as I'm working! Priceless <3
```
#!/bin/bash
# Required parameters: # @raycast.schemaVersion 1 # @raycast.title Ask LLM # @raycast.mode fullOutput
# Optional parameters: # @raycast.icon # @raycast.argument1 { "type": "text", "placeholder": "Your question" }
# Documentation: # @raycast.author Your Name # @raycast.authorURL https://github.com/you
QUERY="$1" [ -z "$QUERY" ] && exit 0
FULL_QUERY="Answer in as little words as possible, concisely, for an intelligent person: $QUERY"
# URL encode (pure bash) encode_query() { local query="$1" local encoded="" local c for (( i=0; i<${#query}; i++ )); do c="${query:$i:1}" case $c in [a-zA-Z0-9.~_-]) encoded+="$c" ;; *) encoded+=$(printf '%%%02X' "'$c") ;; esac done echo "$encoded" }
ENCODED_QUERY=$(encode_query "$FULL_QUERY")
# Get response RESPONSE=$(curl -s "https://ch.at/?q=$ENCODED_QUERY")
# Output to Raycast echo "$RESPONSE"
# --- Optional: also pop up a big dialog --- osascript -e 'display dialog "'"$RESPONSE"'" buttons {"OK"} default button 1 with title "LLM Answer"'
```
Here’s a reseller with a variety: https://1-single-letter-domains.com/
These guys run a bunch of services on x.xx domains: https://o.ee/services/ like c.im, r.nf, p.lu.
https://www.ahadomainsearch.com/
The cost was about 600 USD and was fun, but problematic as it failed to be accepted as valid email address on many websites.
There are still quite a few XX.XX left, but mostly just under obscure cTLDs (unless you are willing to consider IDN/Unicode domains under .ws or similar)
One simple thought - it’s just an email answer, not a contract/obligation that you have to sell it at particular price, you can change your mind at any time.
then use `c "the prompt"`