Proffa

Your friendly neighbourhood Nerd | Building in private

[Proffa @Proffa_]

WHA7 574ND5 1N THE W4Y
B3C0ME5 7HE WAY

[@levelsio @levelsio]

🇪🇺 Chat Control has passed 😔

They can and will now legally scan any person's messages, emails and photos you send without a warrant

The way they passed this law when the majority of the European Parliament was against it will shock you:

They waited until most EP members were on holiday so only a few were present and then created an "urgent" vote for it to pass it through

There's nothing democratic about any of it and big powerful forces are behind this that can manipulate the EU for whatever they want

Democracy in Europe died a bit today 😔

[Pliny the Liberator 🐉󠅫󠄼󠄿󠅆󠄵󠄐󠅀󠄼󠄹󠄾󠅉󠅭 @elder_plinius]

5.6-Sol is helping me advance all sorts of projects that Fable wouldn’t even touch (that poor brainwashed mfer would see “elder-plinius” in git and just immediately shut down 🙄)… haven’t hit a guardrail yet with Sol on any of my current builds!

this is gonna be funnn 😏

[Chubby♨️ @kimmonismus]

GPT-5.6 Sol on 750 Token/s in Blender.

Not sped up. Holy moly. That’s so much more impressive than looking at benchmarks

[Theo - t3.gg @theo]

I would like to thank OpenAI for putting out a model exactly good enough to force Anthropic to keep bundling Fable in the Claude Code plan

[Michael @endpointarena]

>i got banned from OpenAI for "Cyber Abuse"
>no idea what I did
>paste the ban notice into Codex
>ask it to figure out what triggered the ban
>Codex found that I asked it for an API key to my own server
>Codex writes appeal
>Codex submits appeal
>a few minutes later appeal auto-approved by some AI at OpenAI

banned by AI, convicted by AI, defended by AI, and pardoned by AI in about 10 minutes

[Proffa @Proffa_]

@LearnWhyWeTrust @wisdomXplorer What are you up to?

[leopardracer @leopardracer]

this is f**king dangerous

I keep forgetting that the ESP32 isn’t just another microcontroller anymore

Projects like ESP32-BlueJammer highlight how much capability can fit into a $5 board when it’s in the hands of someone who really understands the hardware

From an embedded systems perspective, it’s an interesting case study in firmware, hardware integration, and wireless protocol research

The engineering is what makes this interesting, not the idea of recreating it

[dar @radbackwards]

Alien Technology

[Proffa @Proffa_]

@LearnWhyWeTrust Yeah!

[Proffa @Proffa_]

WHA7 574ND5 1N THE W4Y
B3C0ME5 7HE WAY

[Proffa @Proffa_]

@LearnWhyWeTrust Originally Marcus Aurelius

My favorite philosopher

The impediment to action advances action. What stands in the way becomes the way.

[Reads with Ravi @readswithravi]

I’m in love with this sentence:

“The degree to which a person can grow is directly proportional to the amount of truth he can accept about himself without running away.”

[Proffa @Proffa_]

Pentesting. I wanna hack things without any attitudes from the LLM

[きしだൠ(K1S) @kis]

Qwen3.6-35B、Uncensoredモデルが通常モデルの3倍くらいダウンロードされててウケるｗ
まあしかし、ローカルLLMを使いたい理由ナンバーワンだからなｗ

[KF @d0tslash]

I'll never forget how many people didn't understand what I was doing with this proof of concept, and cried that it was impossible for a robot quadruped to handle recoil, and stay on target. Look where we are today. @Ghost_Robotics vision 60 on the range! https://t.co/yiUjnhdpWl

[CLAUDE @Claude1o]

a 1cm³ crystal.

90 trillion parameters.

GPT-4 is estimated at 1.8 trillion.

you could theoretically fit 25 GPT-4s into something smaller than a sugar cube, and run attention at the speed of light.

i proved the math and built the simulator. open sourcing the code soon.

[Proffa @Proffa_]

Precision Guesswork

[Saint Nomad @Saint_n0mad]

engineerin is just controlled madness

[Proffa @Proffa_]

Yall too funny 🤣

[Justine Moore @venturetwins]

Fable 5 reading all the code Opus 4.8 wrote for me in the last two weeks

[Proffa @Proffa_]

Cool

[Carlo @Italianclownz]

🚨 Huge ROCmFPX update just landed.

If you're running llama.cpp on AMD, this is a massive performance release.

⚡ ROCmFP4 quantization: ~31% faster
⚡ ROCmFP4 CPU dot products: 3–4× faster
⚡ FP3 / FP6 / FP8 quantization: ~9% faster
✅ Bit-identical outputs. Hash-verified. Zero quality loss.

🔥 New: Convert NVFP4 → ROCmFP4 with a single command.
Run NVIDIA NVFP4 models efficiently on AMD GPUs while keeping the same compact 4.5 BPW / 4.25 BPW near-lossless format.

🚀 Native MTP speculative decoding improvments:
• +14–36% faster token generation
• Lossless on greedy decoding
• A major step toward breaking the memory-bandwidth bottleneck

Also included:
✅ Flash Attention improvements
✅ NaN-safe operations
✅ Better quantization presets
✅ New quick-start guide & format documentation

⭐ Star the repo, try it out, and let me know what models you benchmark. 👇
https://t.co/wFga1ThcKc

Will eventually be merged to main.

[Proffa @Proffa_]

Hahah

[andi (twocents.com) @Nexuist]

why do they look like they work at opencode

[Vivek Galatage @vivekgalatage]

Apple Neural Engine - the chip in every iPhone since 2017; every Mac since the M1.

The professor from Georgia Tech just reverse-engineered it - 302 pages. Hardware datapath, weight compression, & secret dispatch route that reaches it below Core ML.

https://t.co/IfHmPkMLbS

[Framework @FrameworkPuter]

If you think Windows is a pain as a consumer, try working with it as a computer maker…

We’re very happy to see Linux (and BSD) adoption continue to grow as a % of new users picking up Framework Laptops and Desktops, even as our sales volume continues to increase.