Patrick Laverty

machinist, tinker, free thinker.

[Patrick Laverty @PatrickLaverty0]

@SecRubio My guy, you're getting clowned on in the comments.

[John of Truth @JohnOfTruth]

@SecRubio This is what your regime change asset and his cohorts got up to in Hong Kong.

[Patrick Laverty @PatrickLaverty0]

@CanadianQuoter @MKartalaei @SecRubio Shut the fuck up you retarded snow back ugly bitch.

[Patrick Laverty @PatrickLaverty0]

This is the second US politician I've seen today posting "China Bad" slop. Get ready my Chinese comrades, the propaganda machine has been pointed your way.

[Secretary Marco Rubio @SecRubio]

The U.S. strongly condemns the Hong Kong Court’s decision to once again sentence Joshua Wong for exercising his freedom of speech. He has used peaceful means to defend the rights promised to all individuals in Hong Kong.

We call for his immediate release and that of all unjustly detained persons.

[M. Kartalaei @MKartalaei]

@SecRubio Marco, you’re against freedom of speech on US college campuses but suddenly championing it in China?

Democracy is just a tool for you guys.

Take care of your own backyard first.

[Patrick Laverty @PatrickLaverty0]

@klawonn_devin

[Patrick Laverty @PatrickLaverty0]

@RepRoKhanna Lol. Here we go. You can't stop Chinese releasing free open weight AI to the masses so you're going to crank out the "Chhh Chhh CHHHUHHWHINEAHH BwwwwAaaad" rhetoric. Cry harder.

[Jujunori @DumondinJ]

Salut à tous !

Je tenais à vous présenter mon travail concernant la version Saturn de Wipeout.

Je vous ai pondu une vidéo comparative de la version d'origine et ma version, J'ai mis un point d'honneur à quasiment effacer le clipping du jeu.

[Patrick Laverty @PatrickLaverty0]

@Hashgraphian @chaojidigua @RepCasar LaMo UhHhH yeah, you fucking retard. Read the post before making retard noises. Multiple Chinese labs releasing RSI. You can't stop it. Cope.

[Superman @thesupermannx]

China cracked recursive self-improvement 🤯

Tencent dropped 27B model that rewrites, retrains, and refines itself.. with zero human in the loop.

It’s called "Terminal-RSI" and it solves one of the biggest problems with AI.

They are incredibly lazy.

Usually, an agent will make partial progress, claim it succeeded, and completely fail the final environment verifier check. this new pipeline forces agents to actually finish the job.

here’s how it works:

they built a recursive loop with two alternating phases..

- RSR (recursive self-rewrite): the model fine-tunes itself on its own successfully rewritten trajectories

- RPR (recursive policy-refine): it then runs reinforcement learning (ppo) using a "critic" inherited directly from the previous stage to lock in those discovered behaviors

- Loop: the training sequence leapfrogs itself: Base → RSR-1 → RPR-1 → RSR-2 → RPR-2

and the numbers are insane..

they applied this to a qwen3.8-27b base model. across the four "terminal-bench" suites, the agent's average success rate jumped from 20.68% all the way to 50.58%.

on the absolute hardest benchmark suite (terminal-bench 4.0), the base model scored a flat 0.00%.

after terminal-rsi?

it hit 19.84% (putting it right on par with reference models like gemini 3.8 flash).

the secret sauce? they fixed the reward system.

instead of giving the ai "dense rewards" for just trying or passing semantic checks, they switched to a "success-prioritized reward" (spr).

it mathematically separates strict success from partial credit.

basically, the agent only gets the massive reward payout if the code actually passes the final terminal gate.. so it stops faking success.

the gap between partial local repairs and strict terminal verification is closing fast.

we are watching agents build their own training environments and level up autonomously.. the future is wild.

[Beff (e/acc) @beffjezos]

Diffusion of models, private ownership of hardware, and decentralized control of capabilities is the backstop against authoritarian tyranny of centralized institutions merging with the govt and banning AI for everyone else

If they try to ban private AI it's war.

[Patrick Laverty @PatrickLaverty0]

i cant make this shit up

[Paisley Circus @PaisleyCircus]

@peterAjensen Satire? lol

[Chris Dukes @clduab11]

@peterAjensen Not only are you all dead wrong, but I'll be doing what I can to partner with Indiana law firms to enjoin the City of Bloomington from enforcing this law.

You can't even DEFINE "superintelligence", let alone know how it works. See y'all in court.

https://t.co/ZgoQJvLq7x

[Chris Dukes @clduab11]

@peterAjensen If anyone knows any Indiana lawyers; I've drafted a draft Complaint and accompanying Memorandum of Law showcasing why @POTUS should send the Federal authorities to Bloomington, IN. Find the material here for free.

https://t.co/gdt2VpJXv7

[Patrick Laverty @PatrickLaverty0]

@peterAjensen @Chaos2Cured yeah real intelligent reply, a single emoji to that mans nuanced argument. The real clown show starts when you retards think you'll be able to march door to door confiscating peoples hardware. Please send me a notification when you do and live stream it.

[Peng @pengsonal]

4 FREE UNCENSORED AI MODELS THAT RUN ON 8GB RAM. NO GPU NEEDED

run locally with just your CPU + RAM

1. Dolphin 3B → Lightweight chat, ~8GB RAM
https://t.co/Ur3j5HM5rQ

2. Qwen3 4B Abliterated → Fewer refusals, ~8GB RAM
https://t.co/Q3BUDpvplI

3. Qwen3 4B Heretic → Non-thinking chat, ~8GB RAM
https://t.co/6P5xdc46X5

4. Dolphin 8B → Chat + coding, ~16GB RAM
https://t.co/FjWx4EIBa1

setup:

• Install llama.cpp: https://t.co/NQlj1ShsjC
• Download a Q4_K_M GGUF file
• Run: "llama-cli -m model.gguf -ngl 0 -c 2048 -cnv"

replace "model.gguf" with your model's file path.

works offline. CPU generation can be little slow

[Patrick Laverty @PatrickLaverty0]

@maxwinga @ONagel33303 @Gregory_C_Allen The coping is from you retards. You think China is going to stop advancement to appease a bunch of retarded American libshits? You people really are delisional.

[Patrick Laverty @PatrickLaverty0]

lol love to see it. cope antis

[Ori Nagel @ONagel33303]

A bill to Ban Superintelligence is unlikely to pass the 60-vote filibuster in the Senate.

—@Gregory_C_Allen (Former Senior AI Advisor, CSIS)

[CrazyShyyt @CrazyShyyt]

China built a 27B model that matches Claude Fable 5 on coding.

and it's 1000% open-source and free.

[Davey @daveeyyy_]

Jensen Huang: Nvidia already depends on AI to design its chips and write its software.

“We can’t design a chip anymore without AI.”

“At night, our AIs are exploring design spaces, uh, vast and wide, that we would never do ourselves. Because it costs too much money to explore it.”

“We can’t write software without, without AI anymore.”

“I want everybody, every organization in our company to use AI very aggressively.”

“I want, I wanna turn Nvidia into a one giant AI.”