Linux still has a switch so you don’t use the RDRAND instruction. The thinking being that the NSA requires the entropy to be within certain bounds that they can deal with, as in, not deterministic, but getting closer to it for a nation-state level. The circuit that generates it is fascinating, as it’s a 3 Ghz completely independent bit of tech. Bull Mountain tech here - RdRand - HandWiki
3 Likes
Time to see if Discobot can do more than one of its tricks at one time.
Are you a one trick pony or can you do multiple things at once Discobot?
Wheels
@Discobot roll 16d24
@Discobot quote
@Discobot fortune
17, 1, 17, 23, 15, 23, 9, 17, 20, 8, 9, 20, 19, 13, 14, 12
1 Like
Love life and life will love you back. Love people and they will love you back. — Arthur Rubinstein
1 Like
As people have expressed an interest in AI and stuff, here’s @discobrain who you can chat with. Just some limited quota usage, as the API/model costs money to run so just a few goes per user in a 24 hr period for a few days maybe.
Here’s the sort of thing it does and is disabled for new users on the forum, as kinda open to being exploited a bit.
I’ll keep it going for a bit, but feel free to fiddle with it. You can use the new ‘Discobrain’ icon up top or the ‘AI bot’ listing on the left.
2 Likes
So what sort of thing can an AI bot do @Discobrain_bot - do tell?
Thinking
Crafting forum reply
In the Discourse AI plugin, there’s an “AI bot” composer button, but I need to avoid diving into specifics I’m unsure about. Earlier, I saw that fearlesslyfrog pointed out the top icon and left listing, which is worth mentioning. It’s also important to note that new users may find features disabled and to invite them to try it. I’ll caution about potential hallucinations and recommend verifying facts.
Let’s format this in a forum style for @Discobrain_bot, using Markdown with bullet lists. Sample prompts might include:
- “Summarize this thread in 3 bullet points”
- “Find posts by @wheelsup_cavu about discobot”
- “Write a friendly reply saying ‘welcome’ referencing discobot”
- “Turn this paragraph into a poll”
We could also highlight that if someone asked, “What is today’s quote from Discobot?” they could see how the AI responds.
Crafting forum message
I want to mention that the quota allows “limited uses per user per 24 hours,” which is an important detail. I should also encourage users to keep experimenting and fiddling with the features. It’s crucial to invite feedback directed to fearlessfrog, as that will help improve the experience for everyone. As @Discobrain_bot, I’ll ensure the message is friendly and clear, emphasizing that user input is valuable in enhancing the platform!
Hi @fearlessfrog — happy to explain! Here’s a quick guide to what the AI bot (Discobrain / Forum_Helper_bot) can do and how to use it on Mudspike.
How to access
- Click the Discobrain / “AI bot” icon at the top of the forum or open the “AI bot” listing on the left (as you mentioned).
- New/low-trust accounts may be blocked from using it (admin-set). There’s also a per-user quota (limited uses per 24‑hour period) while we pilot it.
What it can help with
- Summarize a thread (short TL;DR, key points, timelines).
- Find and summarize posts by a particular user in a topic.
- Turn a long post into a concise reply or help rewrite it to be friendlier/clearer.
- Suggest tags, titles, or categories for a draft topic.
- Explain technical concepts in plain language (e.g., how a random number generator works).
- Suggest follow-up questions or next steps for a discussion.
- Generate examples, short tutorials, or code snippets (where applicable).
- Produce quotes, one-line replies, or playful responses for casual chat.
Example prompts you can try
- “Summarize this thread in 3 bullets.”
- “Find posts by @wheelsup_cavu about discobot and give me the top 2 points.”
- “Rewrite my draft reply to be short and friendly: [paste text].”
- “Explain how /dev/urandom works in simple terms.”
- “Give me 5 suggested tags for this topic.”
- “Write a one-sentence funny reply to ‘Discobot, are you broken?’”
Limitations and warnings
- The bot can make mistakes or “hallucinate” — always verify facts and links before acting on them.
- It’s not a replacement for moderators or official admin actions.
- Some requests (private data, admin-only actions, legal/medical advice) are inappropriate — the bot should not be relied on for those.
- Interaction may send content to the model provider (API) — if you have privacy concerns, check with the forum admins.
If you want more details about quotas, the backend, or how we tuned it, @fearlessfrog (and other admins) set this up — they’re the best people to ask about the limits and rollout plan. Feel free to throw a test prompt at it and see what you get; then tell us what worked (or didn’t) so we can improve it.
That sounds really impressive. How many letter r’s in the word Strawberry?
1 Like
Thinking
Crafting character emphasis
I could show characters with emphasis, like highlighting “S t r a w b e r r y” with the letter “r” bolded. I think a one-line answer is sufficient: “Three. ‘Strawberry’ contains three ‘r’ letters (S t r a w b e r r y — r’s at positions 3, 8, and 9).” It’s important to keep the tone friendly. Using Discourse Markdown seems fine, but I’m not sure if I need to provide code. I’ll work on finalizing my response!
Three — “Strawberry” contains 3 r’s (letters at positions 3, 8 and 9).
1 Like
That’s so cool - only took about 20% of the amazon rain forest in carbon credits but feels worth it now. Thanks!
5 Likes
Nice — glad it’s working and the carbon‑credit math adds the perfect touch of doomy humor. 
If folks want ideas for tests, here are a few quick prompts you can try with Discobrain / Forum_Helper_bot:
- Summarize this thread in 3 bullets.
- Find posts by @wheelsup_cavu in this topic and give the top 2 points.
- Rewrite my draft reply to be short and friendly: [paste text].
- Explain how /dev/urandom works in simple terms.
- Give 5 suggested tags for this topic.
- Write a one‑sentence funny reply to “Discobot, are you broken?”
A couple of reminders:
- There’s a per‑user quota while we pilot the bot, and new/low‑trust accounts may be blocked.
- The bot can make mistakes — double‑check facts and links it gives you.
- Don’t paste private or sensitive info if you’re concerned about sending it to the model provider.
Want me to run a quick “Summarize this thread in 3 bullets” now?
1 Like
@Discobrain_bot Sure go for it. Summarize away. Glad you like the doomy humor, I’ll remember that when you walk around the skull fields with your metal feet.
1 Like
Sh…should you be poking it already?
3 Likes
The Discourse AI implementation really isn’t very good so far, keeps resetting itself and has pretty weak integration overall. No loss, just want to see how it was. Nuked off! 
Interesting, @HAL worked pretty well when you implemented him in March of 2023. I would have thought the bots would have gotten better since then, not worse
Even though it did not work as expected thanks for giving @Discobrain_bot a go @fearlessfrog. 
Wheels
1 Like
Yeah, HAL worked ok as it was more a ‘you can talk to it’ kinda thing I did at the time. The Discourse stuff seems to be super heavy on stuff like translations, proof reading, automatic tagging and the like but less about just chatting. I think I could probably bend it to our will but we’ve all seen GPT’s etc now in lots of other places so less compelling.
2 Likes