TurnerNet

If that addict on your street were your own son, what would you do? That is the defining question that guides my 5 step plan to fix the homelessness problem in LA. We *must* end this evil racket of corrupt politicians and NGOs who profit off the misery of these poor souls. They launder money and feed them more drugs, so they can keep their customers locked in this hell on our streets. We have a moral obligation from God to help them and make our city safe and clean for everyone. Karen Bass and Nithya Raman have forsaken this city. Time for real leadership. Time for real compassion.

GOOGLE 🔥: Gemini desktop app will get Gemini Live, Gemini Spark, Gemini Omni, and a new "Stream to Cursor" feature. What we know so far 👀 - "Stream to Cursor" feature will allow Gemini to have something similar to "Magic Pointer" announced last week during Android Show. - Gemini Spark Agent will be able to operate local files from attached folders. - Gimini Omni is referred to as "Veo4 Omni" internally. - Skills will be supported too. - Gemini Live feature is WIP and not functional yet. A short demo from testers ⚡

@vivoplt @karpathy @fchollet @ylecun @AndrewYNg @rasbt @dair_ai @lilianweng @jeremyphoward @simonw @_akhaliq @ID_AA_Carmack @gwern @goodside @drfeifei @demishassabis @TurnerNetTech

15 AI related accounts you should follow on Twitter: 1. @karpathy 2. @fchollet 3. @ylecun 4. @AndrewYNg 5 @rasbt 6. @dair_ai 7. @lilianweng 8. @jeremyphoward 9. @simonw 10. @_akhaliq 11. @ID_AA_Carmack 12. @gwern 13. @goodside 14 @drfeifei 15 @demishassabis Let me know who I missed guys

@msurguy @jessegenet @useTRMNL She is using a 32" display. If I am reading this right, the largest here is 10.3". Is that correct?

@jessegenet Pretty sure you can do the same or better with @useTRMNL too! I have one and it’s amazing

GIANT e-ink display LIVE in my house and actively removing the “mental load” of motherhood 😅 Turns out my household chaos just needed to be tamed by a display my team of @openclaw and @NousResearch Hermes agents manage for me 💅

HE. is. RISEN. ✝️

@iruletheworldmo Enterprise only???

🚨BREAKING FRONTIER MODEL NEWS claude mythos set for release april 16th dario has more leaks than the titanic, here’s some info from anthropic staff. >95 or higher on every single benchmark. except arc agi 3, yet to be tested on. >dramatically outperforms opus 4.6 on coding, reasoning, and cyber >anthropic privately warning government officials about its capabilities >so powerful they’re calling it “unprecedented cybersecurity risk” >already being tested with early access customers >priced at $120/$600 per million tokens >10 million token context window >enterprise use only capybara is here.​​​​​​​​​​​​​​​​ capygpt is agi.

@iruletheworldmo Done

follow my homie and i’ll send you a special strawberry man gift. only counts if your verified. if you ain’t verified GET VERIFIED, it’s lit 🔥

Interface: CLI, Python API, Web UI, Docker. Extensible: plugin system via Python entry points for custom detectors and restorers. Install: pip install artefex[all] MIT licensed. Good first issues labeled. Contributions welcome. Docs: turnert2005.github.io/artefex #OpenSource #Python #ComputerVision #MediaForensics #ONNX #DevTools #ImageProcessing

The problem with every existing restoration tool: they treat all images the same. Got a blurry photo? Upscale it. Got noise? Denoise it. Never mind that the blur came from a 4x re-JPEG, the noise is sensor pattern from a phone camera, and someone also watermarked it twice and ran it through Instagram's compression pipeline. One-size-fits-all filters don't fix that. They make it worse. Artefex runs 13 forensic detectors before it touches a single pixel: Compression: JPEG artifact detection via 8x8 block boundary analysis, plus multi-recompression detection using double quantization and ringing analysis. Resolution: Upscaling/loss detection via high-frequency spectral analysis and autocorrelation. Color: Color shift detection via channel imbalance and clip ratio analysis. Artifacts: Screenshot remnant detection via border uniformity, aspect ratio, and dimension analysis. Noise: Sensor and added noise via Laplacian MAD estimation. Overlay: Watermark detection via tile correlation, histogram peaks, and alpha channel analysis. Metadata: EXIF stripping detection via metadata presence and completeness checks. Provenance: Platform fingerprinting that identifies whether your image was processed by Twitter, Instagram, WhatsApp, Facebook, Telegram, Discord, or Imgur -- from dimension, compression, and EXIF signatures alone. Provenance: AI-generated content detection via frequency spectrum, histogram smoothness, noise uniformity, and patch consistency. Security: Steganography detection via LSB analysis, chi-square test, entropy, and pairs analysis. Provenance: Camera and device ID via sensor noise PRNU analysis (DSLR, smartphone, webcam, scanner). Forgery: Copy-move detection via patch-based feature matching for cloned regions. The output is a ranked degradation chain, graded A-F by severity. Then and only then does restoration begin -- neural (ONNX) models first, plugin restorers second, classical fallbacks third. Each step targeted to what was actually found.

Introducing Artefex -- open-source neural forensic restoration for images. Most tools blindly upscale or denoise. Artefex diagnoses first, then reverses each degradation step specifically. Diagnosis before treatment. Every time. github.com/turnert2005/ar…

@aaalexhl @grok How would I prompt you to build this?

Holy shit guys! A solo dev just vibe coded what Palantir charges governments millions for! By plugging into available public APIs and displaying that public data!

@TheAhmadOsman Please!

Many are asking for the GPU giveaway to be for an RTX PRO 6000 Blackwell How about this, if this tweet gets: > 5k likes > 1k retweets > 200+ replies I will make sure the giveaway announced on Monday will be for the RTX PRO 6000 Blackwell with 96GB of VRAM It’s up to you guys

It is

GPT-5.3-Codex-Spark is now in research preview. You can just build things—faster.

@fmdz387 @trq212 This is awesome!

@trq212 Then generate insights card x.com/fmdz387/status…

Opus 4.6 is a special model, it really feels like a true collaborator you might have got a sneak peek at its work earlier this week- the videos I launched were made completely by Opus 4.6 (see below)

Introducing Kitten TTS, a SOTA tiny text-to-speech model - Just 15M parameters - Runs without a GPU - Model size less than 25 MB - Multiple high-quality voices - Ultra-fast - even runs on low-end edge devices Github and HF links below

This is probably one of THE most important paper of the last few months. Small language models are sufficiently powerful, operationally suitable, and economical Agentic tasks. - Phi-2 matches 30 billion models running 15x faster. - Serving a 7 billion parameter small language model is 10–30x cheaper than larger models. - Agentic applications use only limited language model capabilities, fitting well with specialized small models. - Heterogeneous systems use efficient small models routinely, invoking large models sparingly for general tasks. - A conversion process is recommended that involves logging agent interactions, clustering tasks, selecting small models, and fine-tuning them on task-specific data. SLM fine-tuning aligns behavior precisely for structured agent interactions like tool calls. Heterogeneous systems blend SLM efficiency for routine tasks with LLM power for complex steps. On-device SLM inference delivers low latency and enhanced data privacy for agent users. --- Paper - arxiv. org/abs/2506.02153 Paper Title: "Small Language Models are the Future of Agentic AI"