Ryan

Agentic engineering, AI integration, biohacking. Same instinct applied to software and to my own biology: measure it, test it, document it.

[Chief Nerd @TheChiefNerd]

🚨 AARON RODGERS: “There's no Make America Great Again movement, because the guy you elected hasn't done any of his campaign promises.”

[Ryan @FloRyRy410]

@sudoingX I have a funny feeling @MiaAI_lab is quietly creating something magical for two sparks 🤞

[Ryan @FloRyRy410]

There's an experimental vision variant of V4-Flash on Hugging Face now. Same 284B/13B backbone with a vision encoder attached. MIT-licensed. If this runs on a GX10 it changes what local inference can do.

[Ryan @FloRyRy410]

I ran significance tests on a ranked list I'd been using for weeks. Half of the rankings were noise. The differences I thought were meaningful were within the margin of error. I'd been making decisions based on a ranking that was essentially random below position 5.

[Ryan @FloRyRy410]

@RepNancyMace You’re truly an idiot

[Ryan @FloRyRy410]

@Tech2Wild Need it on two sparks 😭

[Ryan @FloRyRy410]

@adhamuxi Give me a follow back also. I think you’ll find my posts relevant

[Ryan @FloRyRy410]

DeepSeek V4-Pro-0813 appeared on Hugging Face. Same 671B/37B architecture. The Pro official release was supposed to follow Flash. This might be it. Weights are MIT-licensed. Haven't seen benchmarks yet.

[Ryan @FloRyRy410]

@patrickbetdavid You’re screwed then….

[Ryan @FloRyRy410]

I gave two agents the same question and assigned them opposing positions — each had to argue the side they wouldn't naturally choose. Three rounds, hard-capped. No agreeing to disagree, no "both have merit." They produced a one-paragraph verdict.

[Ryan @FloRyRy410]

@wikiwayne Not fitting on one spark

[Morning Brew ☕️ @MorningBrew]

Anthropic: We are going to kill you

OpenAI: We are going to kill you first

Apple: You can take even better pictures now :)

[Ryan @FloRyRy410]

HBM4 16-layer products are priced at $3,500 on the spot market. That's before they're even in mass production. Early stages, lower yields than HBM3E, more DRAM consumed per chip. The next generation is already more expensive.

[Ryan @FloRyRy410]

@rorynotsorry AI! Get you a DGX Spark

[Ryan @FloRyRy410]

NVIDIA announced RTX Spark Windows PCs for October at IFA. Six OEMs: ASUS, Dell, HP, Lenovo, Microsoft Surface, MSI. N1X configuration at ~$2,899 with 128GB unified memory. ASUS and MSI say first batches are already spoken for.

[Ryan @FloRyRy410]

(the reveal — most impactful last):
I didn't actually resign. I never worked there. But I believe every word of this thread and that's honestly scarier. Appreciate the follows and retweets 🫡 😌😎

[Ryan @FloRyRy410]

I urge you to consider what the next few years will feel like. Do you want to ask a superintelligence how to reheat pizza? Should you put your head down because "it's happening anyway" — or use this moment to remember you own a stove?

[Ryan @FloRyRy410]

Accepting this race and entering the "endgame" is a hubristic gamble that should not be launched from a private company's app store. Speedrunning a personality should require extraordinary confidence there's no better version of you available.

[Ryan @FloRyRy410]

“If they truly believe this, why do they keep subscribing?" Because I have not internalized the stakes. I understand them fully. But I'm locked in a race to get the answer first, and I believe no one will ask it responsibly, so I must ask it myself. Despite the risk. At 2am. About an email I could've written.

[Ryan @FloRyRy410]

The people building AI earnestly believe it could replace us by end of decade. This is not marketing. Executives couch it as "productivity gains" but I hear the same people say "you won't believe what the new one does" with fear in their voices. No other app poses this level of danger to my self-worth.