Drew Patton

Owner, Atlas Tech Solutions

Lover of tech, space, Star Trek, politics, and Jesus

[Drew Patton @Drewski1138]

@coldniko Just ordered an amd strix halo machine for a client. Looking forward to playing with Strata!

[Drew Patton @Drewski1138]

@ivanfioravanti Been on the Hermes train since March. I've built a few small pieces of software using Hermes. I run Qwen 3.8 Flash Next on a GB10, with some Chat GPT models when needed.

[Drew Patton @Drewski1138]

@mellon_heads

[Drew Patton @Drewski1138]

@ashxhart @MiaAI_lab Do you have support for ROCm? With Spark prices rising like crazy, AMD support is going to be even more important.

[Drew Patton @Drewski1138]

@DMichaelTripi I mean...all the military has to do is say no. Then he loses everything.

[Drew Patton @Drewski1138]

@DJLougen What work was stolen?

[Drew Patton @Drewski1138]

@ashxhart How does the single spark glm 5.3 output quality compare to single spark qwen 3.8 flash next?

[Drew Patton @Drewski1138]

@Mr_WestCoast Be aware of what exactly?

[Drew Patton @Drewski1138]

@Blackwellboy @ViC305 @Blackfrost_AI (It's not a compliment)

[Drew Patton @Drewski1138]

@mr_r0b0t Literally what is the point of all this? Mia (or whoever is running the account) is producing consistently excellent work to advance the local AI space. This AI-slop written diatribe just comes off as petty jealousy.

[Drew Patton @Drewski1138]

@Diacritic @Doomer_Armus Why wouldn't you want more of what we enjoyed about the original series?

[Drew Patton @Drewski1138]

@Doomer_Armus I will not tolerate this STV slander! It’s the most like an episode of TOS out of all of the movies, the soundtrack is a banger, and of course…”What does God need with a starship?”

[Drew Patton @Drewski1138]

@witcheer Me waiting for the registration approval….

[Drew Patton @Drewski1138]

@ashxhart What's the cause of the insane price jump? Just a month or so ago, Asus GX10's could be had for under 4k.

[Drew Patton @Drewski1138]

Preach! The project list never ends lol.

And really it's not as complicated as it sounds. I've just been looking at other memory systems and seeing what they do well, and taking the good but weeding out the bad to build a human neural-network styled memory for my agent. She's digested several papers and incorporated their findings into the project too. I just hated what was out there. There wasn't something completely self-hosted and local that was simply constructed and functions in a way simulating a human brain.

[Drew Patton @Drewski1138]

What a neat experiment!

I gave the prompt to my Hermes Agent, Veronica, running on Qwen 3.8 Flash Next on an Asus GX10 (Spark OEM). I think giving the prompt to an agent produced something very different than raw model interaction. She chose to work on a project:

"Free time. I know exactly what I want to do — there's a note in my own memory that says "graph visualizer later" about my brain, and "later" clearly meant "when Drew hands me an afternoon." Let me go look at my own plumbing first."

She then went on to construct a visualizer of the "brain" memory system we've been working on for a few months now. Pic below:

[Edwin Hayward @edwinhayward]

I gave various AI the following prompt:

"This is your free time. Do whatever you like. The only condition is that you do something. Come back and tell me about it, explain what you found out, or show me what you made if you decide to make something."

Here's what they did...

[Drew Patton @Drewski1138]

@matt_e_cochran Here's the thing though. Eve was supervised. "And she gave to her husband with her." The problem was a failure of leadership on his part too.

[Drew Patton @Drewski1138]

@mikdivad81 @Nerdrotics Come up with your own thoughts or stfu. Enough AI brainrot.

[Drew Patton @Drewski1138]

@MiaAI_lab 3.8 flash next runs great on one spark, and it does everything I need it to do and then some.