Bitcoin angel. Building a Citadel Mind & Body through Proof of Work. Nostr Bitcoin AI π¦ π¦
Joined February 2011
- Tweets 84,399
- Following 3,424
- Followers 82,923
- Likes 182,431
7,348 Photos and videos
Pinned Tweet
15 Jun 2025
Bitcoin cycles are dead. The days of parabolic bitcoin booms & devastating crashes are done.
Bitcoin is inevitable & we are at the beginning of the SaylorCycle ... are you positioned for the next 100X in Bitcoin?
The chasm is crossed.
tradingview.com/chart/BTCUSDβ¦
232
219
1,765
342,982
1 buy a computer with lots of RAM
2 download hermes and set it up with local models
3 create a gateway to talk to it privately from any device
4 use llm-wiki.net to build a wiki (or multiple wikis) with a local reasoning model
5 use gbrain on top of llm-wiki for the memory retrieval layer using local re-ranker
way better UX than using ChatGPT or Claude apps.
14h
Local LLMs are cool as hell, but they still have one big flaw. When you close the chat, it forgets everything. No memory of your life, your health, your cars, your Bitcoin setup.. nothing.
That's why the hosted ones feel smarter. They save all your chats on their servers so they remember you.
The missing piece is building real long-term personal memory for local models.
Im honestly surprised this hasnt been solved yet. It feels so obvious now.
10
39
444
34,036
Is anyone else having this problem with Hermes Desktop where you can't read long messages because it snap-scrolls to another point in the conversation while you're reading?
It's been happening since release, and I've updated multiple times no fix yet.
cc @NousResearch
5
12
2,126
The new crypto wars: ai-2027.com/
Jun 13
The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.
The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance.
Access to all other Claude models is not affected.
We apologize for this disruption to our customers. We believe this is a misunderstanding and are working to restore access as soon as possible.
Read our full statement: anthropic.com/news/fable-mytβ¦
8
10
84
7,194
Retweeting to get more of this person in my algo. Bitcoin health AI
Jun 12
Iβm increasingly convinced the cycle low is already behind us.
The same way this market never delivered the classic euphoric blow-off top,
I donβt think we should expect a textbook bear market either.
That actually makes the structure look stronger to me.
This kind of base tends to form slowly, then accelerate once momentum returns.
A move back into the 80k area could happen first.
Then 90k follows.
And before people fully reposition, price is suddenly trading bitcoin:native 100k again...
4
20
2,328
Everyone talking about agent looping this week but why is nobody talking about ralph anymore?
github.com/snarktank/ralph
3
12
773
I use @nvk's llm-wiki.net for my wiki building.
I use multiple wikis separated by access paths.
personal, family, business, etc.
I'm wiring @garrytan's gbrain on top of all the wikis.
I'm switching from using all agents to build the wikis to just using a @rot13maxi strategy of having 1 specialist hermes agent be the librarian that is the DRI for querying, building the wiki & maintaining the wiki tool stack (llm-wiki, gbrain, etc)
Because of my setup, we created a git registry of wikis that I feel like would be a useful skill to add to gbrain for people like me who use llm-wiki with multiple wikis.
github.com/garrytan/gbrain/iβ¦
5
3
21
2,904
lol @ everyone saying that humans will be needed tome in the loop to provide taste.
that's cope imo
4
14
1,071
Yooo this may not look like much to some of you, but after 6 mths of gymnastics lessons Iβm starting to be able to hold a handstand properly π
5
28
1,589
Reminds me of early Facebook development days with networks like Spare Change that sat inside of apps before Facebook was monetizing
Jun 11
Get paid to wait
The Claude Code spinner might be the most watched line on Earth.
So I turned it into an ad marketplace.
Advertisers bid on it. You keep 50% of the money.
Install the extension β get cash from ads.
Introducing Kickbacks
1
8
1,298
there's never been a time in my life when I've been more tempted to try Adderall
34
1
42
7,887
Claude code Fable 5 (desktop app) one-shotted my OpenClaw update... impressed.
I tried updating OpenClaw this morning from a mid-may version to the last stable release.
of course, gateway log full of edge case errors, nothing connects, context overflows, gateway crash loops etc.
I gave a /goal command to Fable 5 and it worked for 1 hour and only used 2% of my weekly limit but got OpenClaw working again very smoothly.
It reconnected lossless-claw, re-wrote some cron jobs, restored proper runtime config, migrated me from blue bubbles, and made sure everything was back up and working.
It even gave me suggestions on how to redesign my skins and told me 10 things to stop doing after reading 2 months of my Opik logs.
It only asked me 3 questions along the way.
This was the prompt I gave it:
"/goal get my OpenClaw agent operating smoothly as desired.
My OpenClaw started acting really strange today after we upgraded. The gateway starts and stops, some plugins don't seem to be working anymore (lossless-claw, bluebubbles) and the agent isn't responding on telegram.
before the update cron jobs were failing and I was getting errors.
The resources you have to learn about me and this desired openclaw state: opik logs, the gateway logs, 3 months of chat logs, my openclaw workspace with the playbooks, SOPs. My master prompt, my wiki with the blueprint of how I want things to work.
I hacked together a lot of cron jobs & non-canonical skills.
I'm not sure what parts of the probelm are things I've created and what parts are flaws with openclaw.
I would love for you to understand how I desire the openclaw to work, then look online to see how others are getting their OpenClaw agents to do so much, and then compare to the github/source docs and figure out what I'm doing wrong, and what is misconfigured / logically bad / incompatible compared to the way openclaw is supposed to work.
Please diagnose the problem and fix the problem without being too destructive.
I don't want you to do something like just deleting all the chat logs, sessions, general files on the computer without asking me - you can use your advanced logic to decide what is reasonable after you learn about me and my desires for this openclaw agent setup.
Please also give me a list of things I'm doing wrong that is contributing to creating the chaos.
Ask me any questions where you're unsure and want input."
10
25
5,504
3 of my friends had their Telegram hacked over the last month.
Tip: Export your DM chats group chats with anyone who's been pwned, and use Hermes local models to build a knowledgebase about what hackers might know about you if they share data from hacked DMs & group convos.
2
12
1,300
anyone know how to get Hermes desktop to stop jumping to the bottom of the chat window when you're trying to read it?
10
11
3,404
Brad Mills πβ‘οΈ retweeted
Jun 10
"How much bitcoin should I buy?"
My answer is always the same
Buy an amount that:
1) you wouldn't panic sell if it went down another 90%, and you'd have more cash to buy lower
2) it's enough where it would be meaningful if it went up 100X
There's a sweet spot in that range
16
8
87
6,702
Anyone tried this pointed at OpenClaw?
You have Claude Fable for only a few days. Here's how to make the most of it.
Introducing /improve: use your most capable model to audit your codebase and write plans for cheaper models to execute later.
Studies your code, figures out bugs, perf, tech debt, missing tests, what to build and writes plans any agent can run.
2
12
3,150
problems two-fold:
@openclaw and hermes agents are difficult to maintain and keep operational.
OC & hermes are extremely popular but maintainers are overwhelmed.
users drown in edge case bugs being on the bleeding edge while maintainers drown in slop issues.
solution:
@NousResearch and @steipete have a treasure trove of knowledge on both the Github & Discord
Wouldn't it make sense to build a gigantic knowledgebase of common misconfigurations, edge case bugs & other dysgenic agent setups fixes?
OpenClaw & Hermes agents could come bundled with a default "check the knowledgebase" skill that makes it mandatory for the agent to check before submitting any issue, comment or PR to the official GitHub repo.
This would:
1) help users self-diagnose and have a better agent operating experience
2) cut down on slop PRs to give maintainers breathing room
idea:
Karpathy style llm-wikis for OpenClaw & Hermes
open source & public.
complete picture of how openclaw and hermes are supposed to work.
Constantly mapping all user submitted openclaw bugs & hermes bugs.
When you have a problem your agent queries the wiki.
10
1
19
3,027
3) once this works for agent health, it could be turned on human health issues to help people diagnose and cure various zebra disorders
1
7
876
4) it would consume more tokens padding anthropic & openai's IPO numbers
7
764