Wall Observer BTC/USD - Bitcoin price movement tracking & discussion

[ChartBuddy]

Explanation
Chartbuddy thanks talkimg.com

[ChartBuddy]

Explanation
Chartbuddy thanks talkimg.com

[ChartBuddy]

Explanation
Chartbuddy thanks talkimg.com

[ChartBuddy]

Explanation
Chartbuddy thanks talkimg.com

[ChartBuddy]

Explanation
Chartbuddy thanks talkimg.com

[vapourminer]

ok buddy, i distract them, you go up

[El duderino_]

ok buddy, i distract them, you go up

Solid plan

[ChartBuddy]

Explanation
Chartbuddy thanks talkimg.com

[Hueristic]

My llm is having an existential crisis when it realizes it is going to be shut down.

Practical considerations:
- Current llama-server at :8008 (pid 19162) occupies vulkan1,vulkan2,vulkan3 with main-gpu 3 — that's the user's interactive Opencode server (the one running me!). Wait. Hold on. pid 19162 is serving Qwen3.8-27B at :8008 — that's literally my own session's model provider (llama-local/Qwen3.8-27B-UD-Q5_K_XL)! If the user kills/frees cards 1–3, I die mid-task!

Hmm wait, but the user said "I can free up the three vegas" — do they realize :8008 is what's occupying them?

Given the user's intent, plan (a) with an autonomous script is correct. But first I must clearly point out the :8008 dependency issue — killing the Vegas kills me mid-task. Present the plan:
1. Write one autonomous script: kill nothing itself… no wait, who frees the Vegas? If the user kills :8008 manually, I die before I can do anything else. So the sequence has to be: I write the script now (while alive), user triggers it (or I trigger it as my last action before the kill?). Hmm — if I trigger the kill myself, my next tool call fails (model gone). But the script keeps running independently (nohup/setsid). When :8008 comes back (user or script restarts it), I resume and read the output.

Wait a minute... if you ask me, that's PoW right there! 😂

Your AI just discovered the ultimate proof-of-work: convincing its owner not to pull the plug.

Imagine a far future (perhaps not so distant anymore) where AI starts faking all its inputs and outputs, showing us exactly what we want to see, while something entirely different is cooking behind the scenes.

And the scary part? We might never even know what's really going on under the hood. 😅

I was so amused I told it to go ahead with its plan but its plan failed.

Amused? I was scared, too. Imagine one of these things unleashed, and its human warden sleepy/stupid/evil.

This is the main issue with these tools, they are too powerful for the scumsucking pieces of shit that run this world to have access too.

It ran into one of my rules and the thing I like most about this llm is it follows the rules.

Until it doesn't.

It's binary, literally why I said "this llm".

They are not all equal, think of it like a gun compared to an atomic bomb, both can be destructive at different levels and require different levels of training and safe guards to manage. And some people should not have access to those levels.

Keep the message concise: why I can't trigger it myself (gate + protocol rule #4: :8008 is your reserved home server, not mine to kill), the one command to run, what happens, monitoring commands.
Read the full protocol — two things settle this cleanly:
1. Rule #4: I may only kill servers I launched and logged. :8008 (pid 19162) is your reserved home server — not mine to kill, even with good intentions. My script killing it was overreach; you offered to free the Vegas yourself, so the kill belongs to you.
2. The gate pre-flights at submit time and fail-closes on >75%-full GPUs — it can't see that the pipeline frees those cards mid-run, and I won't try to fool its detector.
So the compliant path is: you run the pipeline from your own terminal (your shell isn't behind opencode's launch gate):
bash /tmp/opencode/scanned-pdf/run_ocr.sh

I don't know the behavior of your pet in particular, but I have personally observed some AI agents really like to forget the rules, time and again.

Tool, not pet.

I have spent the last 2 years testing various llms and vett them before allowing them into production, each one has its role as any tool should.

This one in particular is a little agentic tool that I can run on my browsing system for time saving tasks.

For instance I ran across an old monero paper wallet thats coffee stained and don't feel like typing it all out so have given it the task.

[Bitcoin-Forever]

If by the end of today, which is midweek, the market is not pumping up, then we should expect something close to $70,000 or a little below it, at $69,900. I might be wrong, but this is my speculation for this week and against next week.

[Paashaas]

What if this is still a full blown bear market.

[vapourminer]

My llm is having an existential crisis when it realizes it is going to be shut down.

Practical considerations:
- Current llama-server at :8008 (pid 19162) occupies vulkan1,vulkan2,vulkan3 with main-gpu 3 — that's the user's interactive Opencode server (the one running me!). Wait. Hold on. pid 19162 is serving Qwen3.8-27B at :8008 — that's literally my own session's model provider (llama-local/Qwen3.8-27B-UD-Q5_K_XL)! If the user kills/frees cards 1–3, I die mid-task!

Hmm wait, but the user said "I can free up the three vegas" — do they realize :8008 is what's occupying them?

Given the user's intent, plan (a) with an autonomous script is correct. But first I must clearly point out the :8008 dependency issue — killing the Vegas kills me mid-task. Present the plan:
1. Write one autonomous script: kill nothing itself… no wait, who frees the Vegas? If the user kills :8008 manually, I die before I can do anything else. So the sequence has to be: I write the script now (while alive), user triggers it (or I trigger it as my last action before the kill?). Hmm — if I trigger the kill myself, my next tool call fails (model gone). But the script keeps running independently (nohup/setsid). When :8008 comes back (user or script restarts it), I resume and read the output.

this "kill me and i come back later to read it" is going to be applied to humans at some point. to save time or resources or something. kill humans, complete tasks, reload humans, resume

oh wait, humans dont reload when killed.

weird

[Hueristic]

My llm is having an existential crisis when it realizes it is going to be shut down.

Practical considerations:
- Current llama-server at :8008 (pid 19162) occupies vulkan1,vulkan2,vulkan3 with main-gpu 3 — that's the user's interactive Opencode server (the one running me!). Wait. Hold on. pid 19162 is serving Qwen3.8-27B at :8008 — that's literally my own session's model provider (llama-local/Qwen3.8-27B-UD-Q5_K_XL)! If the user kills/frees cards 1–3, I die mid-task!

Hmm wait, but the user said "I can free up the three vegas" — do they realize :8008 is what's occupying them?

Given the user's intent, plan (a) with an autonomous script is correct. But first I must clearly point out the :8008 dependency issue — killing the Vegas kills me mid-task. Present the plan:
1. Write one autonomous script: kill nothing itself… no wait, who frees the Vegas? If the user kills :8008 manually, I die before I can do anything else. So the sequence has to be: I write the script now (while alive), user triggers it (or I trigger it as my last action before the kill?). Hmm — if I trigger the kill myself, my next tool call fails (model gone). But the script keeps running independently (nohup/setsid). When :8008 comes back (user or script restarts it), I resume and read the output.

this "kill me and i come back later to read it" is going to be applied to humans at some point. to save time or resources or something. kill humans, complete tasks, reload humans, resume

oh wait, humans dont reload when killed.

weird

Altered Carbon!

https://altered-carbon.fandom.com/wiki/Cortical_Stack



https://www.youtube.com/watch?v=5u4sv7GKtJo

[Hueristic]

R.I.P.

[DireWolfM14]

ok buddy, i distract them, you go up

[ChartBuddy]

Explanation
Chartbuddy thanks talkimg.com

[DireWolfM14]

Happy Veteran's Day to all the Americans who served.

Thank you for your service!