h0rus

Founding Partner -xCTO-innov8.work(sold to oyorooms) , Tech Geek, Traveller

[Marcos Rico Peng @Marcos12345rico]

today we're launching @Palmier_io, a video editor Claude can edit.

use AI to edit, organize, and generate footage directly in the timeline.

finally, a video editor built for AI.

open-source. mac native. available now.

[Y Combinator @ycombinator]

An AI just sold an insurance policy entirely on its own.

@kinroai is the autonomous insurance brokerage.

AI agents that quote, answer, and serve your insurance needs 24/7.

Congrats on the launch, @corentin_hgt, @pierrealexai, & @parthfyi!

https://t.co/fksvlEJKQs

[Dhravya Shah @DhravyaShah]

You can now run @supermemory locally.

Introducing the supermemory local

- Fully self-contained. Comes with our graph engine, embedding model, etc.
- Run on any machine, with your @openclaw, hermes, claude, etc.
- SDKs to add memory to your agent, or build your company brain.

[Unsloth AI @UnslothAI]

Google releases Gemma 4 QAT. ✨
You can now run Gemma 4 at 3x less memory with near original performance.

Quantization-Aware Training (QAT) makes it possible to run Gemma 4 26B-A4B on 16GB RAM.

GGUFs: https://t.co/wQgEocxUId
QAT Guide: https://t.co/Nsm1yeGEHx

[Victor M @victormustar]

Before the week ends, let's acknowledge one of the most INSANE week ever for open AI, with 25+ notable open-weight drops across every modality:

🧠 LLMs

→ NVIDIA Nemotron 3 Ultra: 550B hybrid Mamba-MoE, only 55B active, 1M context, MMLU 89.1. NVFP4 variant claims ~5x throughput on Blackwell. First openly-weighted 550B hybrid Mamba-Transformer, closing the gap with frontier closed models.

→ Google Gemma 4 12B: fully open dense any-to-any (text/image/audio/video), 256k context, encoder-free, 140+ languages, AIME 2026 at 77.5. Shipped with a 23-checkpoint QAT wave (mobile ONNX + MLX). Most deployable model of the week.

→ StepFun Step-3.7-Flash: 198B sparse MoE VLM, ~11B active, SWE-Bench PRO 56.3. Apache 2.0.

→ Liquid AI LFM2.5-8B-A1B: edge MoE, just 1.5B active, 128k ctx, MATH500 88.8, MLX-ready. Best on-device option this week.

→ JetBrains Mellum2-12B-A2.5B-Thinking: their first open MoE, near-Qwen3-14B coding at 2.5B active. Apache 2.0.

🎨 Image gen (the surprise of the week)

→ Ideogram 4: their FIRST-EVER open weights. 9.3B flow-matching DiT trained from scratch. #2 overall behind GPT Image 2, top open-weight model on Design Arena + LMArena. Strongest open checkpoint for text-rich images, full stop. It has taste. Still can't believe this is open weights.

🔊 Audio & Speech (a breakout week for open TTS, 4 labs shipped)

→ Boson Higgs Audio v3 4B: 102 languages, 21 emotions, singing/whispering/shouting, sub-second TTFA.
→ RedNote dots.tts: the only fully continuous (no codec) open TTS pipeline, Apache 2.0.
→ Google Magenta RealTime 2: real-time music gen, <200ms latency, text+audio+MIDI. multimodalart ported it to PyTorch within hours with live ZeroGPU demos.
→ NVIDIA Nemotron-3.5 ASR: 600M streaming, 17x more concurrent streams vs Parakeet RNNT 1.1B.

👁️ Vision & VLMs

→ PaddleOCR-VL-1.6: SOTA document parsing at 1B params, Apache 2.0.
→ Baidu NAVA: 6.3B joint audio-video gen, best-in-class A/V sync, Apache 2.0.

🎬 Video, 3D & World Models

→ NVIDIA Cosmos3-Super: 64B omnimodal world model coupling action trajectories with video+audio gen, for Physical AI.
→ JD JoyAI-Echo: up to 5-min multi-shot text-to-video on LTX-2.3.
→ ByteDance Bernini-R + VAST TripoSplat (single-image-to-3D Gaussian splats, MIT).

[Patrick Jiang @patpcj]

Introducing Harness-1, a 20B search agent trained with a state-externalizing harness.

> frontier-level long-horizon search, rivaling Opus-4.6 and outperforming GPT-5.4

> Context-1-level cost and latency

> externalizes candidates, evidence, verification, and search history

> open-source

[kshitij vaze @VazeKshitij]

THIS is what I pay my internet bills for man

[Fokki @0x_fokki]

🚨a 22-year-old makes $8,217/month from an anime channel he built in one weekend

→ Claude: script and scene description. 10 minutes.
→ Midjourney: every frame. 20 minutes.
→ Runway: movement, breathing, camera. 15 minutes.
→ ElevenLabs: character voiced with emotional direction. 10 minutes.
→ Suno: score. 5 minutes.
→ Make: published Tuesday 9am. automatically.

$8,217 last month. 3 hours of work total.

the studios haven't figured out what to do about this.

full build with every prompt in the article above👇

[Alvaro Cintas @dr_cintas]

Google's new algorithm just shrunk 31GB of memory down to 4GB 🤯

TurboVec is a new open-source tool that stores the data your AI app searches through, using 16x less memory.

It runs on Google's TurboQuant, which skips the slow setup step every other tool needs.

→ Faster search than the popular alternative (FAISS)
→ Works on both Mac and standard servers
→ Narrow results to exactly what you want
→ Plugs straight into LangChain and LlamaIndex

Your data never leaves your machine. Runs fully offline, works with Python out of the box.

100% Open Source.

[Larus Canus @MrLarus]

😱AI flight control is getting crazy!

One red line controlled this first-person Harry Potter-inspired Quidditch POV.

Ride the broom, chase the Golden Snitch, fly through Hogwarts-style towers, skim over Black Lake, dive through a bridge arch, and return for the final catch. 🧹

Workflow breakdown in the comments 👇

[Tye @tye1337]

How I'm Going To Make $100K From The World Cup (copy me)

[Robert Lauko @robert_lauko]

See the top ranked papers in AI, ML, Robotics, Quantum Physics, and more on @kurateorg. Hundreds of arXiv preprints ranked daily by scientific impact through pairwise tournaments judged by Claude, GPT, and Gemini.

[slash1s @slash1sol]

THE CO-FOUNDER OF GITHUB GAVE A 46-MINUTE TALK ON GIT BECAUSE ENGINEERS WITH 10 YEARS IN HAVE NEVER SEEN HALF OF WHAT IT DOES

This is Scott Chacon. He wrote Pro Git -- the book most devs secretly learned Git from and he co-founded GitHub. So when he says you're missing things, you're missing things.

About ten minutes in it clicks: half the "git disasters" you've ever fixed by deleting the folder and re-cloning had a one-line solution sitting in the tool the whole time.

Git ships new code almost every day -> roughly nine commits a day for over a decade. Most of us stopped learning it the second we memorized add, commit, push.

Knowing Git isn't a senior-dev flex anymore -> it's the floor. The agent writes the code now. Your real job is reading, branching, and untangling the history it leaves behind.

The day an AI agent force-pushes over your main branch, these 46 minutes are the difference between a quiet fix and a very loud apology.

Save it now.

You'll reach for it sooner than you'd like ↓

[RetroChainer @RetroChainer]

> the guy does the gesture
> the girl copies it perfectly
> she was never in the room
> that's motion control you steal movement off anyone

1. take a frame of the source person
2. generate your character into a new scene (glorify)
3. source video = motion, your image = character
4. generate. the photo now does the exact gesture

> the face is one input. the movement is another
> you don't need a person. you need their clip

[Sadao Tokuyama @tokufxug]

Ultralyticsの水中映像から魚やサメを正確に切り出すAI「YOLO26-seg」

従来の四角い枠での検出とは異なり、個体の正確な形をピクセル単位で識別する「インスタンスセグメンテーション」を実行。

海洋研究や生態系監視に役立ち、Pythonを用いてモデルの訓練や予測が実行可能です。詳細はリプ欄。

[Farhan @mhdfaran]

architects are cooked lol

i drew a random shape and Drafted turned it into a full house plan in seconds. real floor plans, unlimited layouts, even 3D models.

this normally costs $20k and takes weeks. it took 5 minutes.

and it's free.

[RT @RT_com]

High-paid tech workers are cutting life down to the basics so they can invest and retire by 30

One Meta engineer makes over $300K a year and still owns no car, couch or TV

More successful Gen Z are choosing calm life over career and money

[⚡AI Search⚡ @aisearchio]

NVIDIA's LocateAnything is a new vision model for grounding and detection. Very performant and accurate!

&gt; 10x faster than Qwen3-VL
&gt; 138M queries + 785M boxes
&gt; GUI, OCR, docs, dense detection
&gt; Free &amp; open source

https://t.co/UvkH8l0QRb

[Franky Shaw @frankyecom]

My 5.1 AI UGC framework is already being used behind real ad spend by some of the biggest D2C brands in the game.

Not because it makes AI videos look cool.

Because it makes AI ads convert.

And those are two very different things.

[Elliot Arledge @elliotarledge]

Co-Founder of Cerebras explains their WSE simplified design compared to classical GPUs made by NVIDIA.