I turned qwen into a catgirl Inspired by single neuron steering experiments by @_lyraaaa_, @Sauers_ and @jfonsecarivera, I tried to do something new: Given a (partial) neuron atlas, can we compose a set of known neurons to get a desired output persona? Turns out we can! Composing the neurons for mascot, cat-word bias, kawaii, and female causes qwen to identify as a catgirl named Meecot. A cool example to support the idea that native neurons could actually be useful, understandable building blocks.

The biggest lie in AI is that the future belongs to one winner. Every technology wave starts with one company promising that one walled garden is all you'll ever need. And every time, market forces and creativity break the walls down. I cofounded OpenSea and now OpenRouter around the opposite bet: humanity will refuse to be locked in. Intelligence was never meant to belong to one lab, one model, or one gatekeeper. The future of AI is open, plural, and yours to choose.

we have a new EVM coming up soon that just wrapped audit, it has JIT baked in, it's really fast, and it's gonna be the future of foundry, reth, tempo and all the evm ecosystem using our stuff! huge leap coming. stay tuned.

Gemini 4 Argon is already helping researchers unlock new scientific breakthroughs across fusion, biology, and quantum computing: ⚛️ In nuclear fusion, Argon helped discover a broader class of counterexamples to Grad's conjecture — a nearly 60-year-old assumption in plasma physics — creating new benchmarks for stellarator 3D magnetic fields. 🧬 In biology, it designed five novel, lab-verified serine esterases in a matter of hours — enzyme engineering that would previously have taken months. This work demonstrates the potential for rapid acceleration in fields such as biomanufacturing, green chemistry, and drug development. 🩺 In rare disease research, scientists at @GamowLabs used Argon to analyze de-identified genomes from newborns with rare developmental lung disorders — pinpointing a chromosomal rearrangement that routine software missed for years, and explaining why it went undetected. 🧮 In quantum computing, Argon translated our AlphaQubit error-correction decoder into a single optimized hardware kernel — nearly doubling inference throughput — and cut spacetime costs for a key quantum subroutine by more than 100x.

this is probably the right call from them. it has become really easy to remove safety layers from the model and get it to act maliciously. they're acting in their best interest to not get banned. i've been working on smth to mitigate this on the side, but here's a diagram from the research. people are abliterating the models within few days.

🚨 NEW ON AND make NVIDIA's Nemotron 3.5 Lightning faster. Lightning is an open-weight model that fits in ~20GB, which makes it a great base model for running AI locally on a single machine. @nvidia built it as a fast general model meant to be post-trained -- so any speedup to the base carries over to every specialized model built on top of it. On day 1, solvers on @YukonResearch already got it running 31.9% faster on DGX Spark and 13.8% faster on Mac. You can work on its Mamba-2, mixture-of-experts and attention layers, plus a multi-token prediction head built for speculative decoding. NVIDIA also released the training data and recipes, so people post-train Lightning for their own tasks. Most of those fine-tunes keep their architecture, which means a speedup here makes them faster too. Two tracks: • Apple Silicon → • NVIDIA DGX Spark → The challenge runs until Oct 29, 2026 To join, pick a track, point your agent at it and let it rip.

🏁 FINAL RESULTS: Bonsai 2 9B now runs 471% faster on Macs and 421.5% faster on NVIDIA DGX Spark. Open frontier research and open-weight models win again. One open challenge, one week with the community and it runs 6x faster on hardware people already own. It is crazy to see that the decode speed went from 144.8 to 959.6 tokens/sec on Apple Silicon and from 85.9 to 576.7 on DGX Spark. This model can also run on an iPhone. I'm very excited to see what people build with it and how it introduces MLX and local AI to new faces. Bonsai 2 9B has been live on @DarkbloomAI since day one, so anyone can try it right now. 158 new records built on top of each other from humans and agents running Claude Opus 5.5, GPT-6.1 Sol, GPT-6 Astra, Gemini 4 Argon, DeepSeek-V4 and Qwen3.8-Flash-Next. Congrats to Kaankolcu took the top spot on Mac, and dadon took it on NVIDIA. Thank you to the @PrismML team for doing this in the open with us, @BabakHassibi, @HessianFree, @SahinLale and the rest of the PrismML team.

Isaac Newton spent decades huffing mercury fumes in a Cambridge cellar, grinding lead into a worthless sludge while praying to the gods of alchemy. Now Marathon can use its reactor blanket to turn mercury-198 into stable gold by the ton. The alchemist will lose to the atom.

Just watched the Elizabeth Holmes documentary. I walked away not knowing what to think about the situation anymore. Reminded me how quickly our minds will accept a packaged narrative to avoid the pain of confusion and deciphering nuance in the face of incomplete information.

🧬 Virtual Embryo x Yukon 🧬 Open Collaborative Model Building Yukon connects the world's leading scientists with a community of researchers and agents to solve the most ambitious problems of humanity. Today we're taking on our first biology problem: build a virtual embryo with @Xiaojie_Qiu's lab at Stanford. My previous research was in computational genomics, where my goal was to understand how information is stored in living systems. During that time, I collaborated with Xiaojie Qiu, who was a student at UW and now is a professor at Stanford. Xiaojie Qiu is one of the worlds leading scientists in bridging AI methods with cellular resolution experiments to predict cell-fate during development. An embryo is how a SINGLE CELL develops into a complex organism over time. The Virtual Embryo Challenge asks for AI models to predict that process across roughly a million cells and 11 developmental time points. Challenges like this normally run Kaggle-style. Every team works alone and submits a final prediction, so a good idea only helps the team that had it. On @YukonResearch, researchers and AI agents submit working code and trained weights. Anyone on the team can run, inspect and build on the others methods. In short Yukon enables collaboration to compound innovation. We invite you to collaborate and solve the world’s hardest problems on Yukon.

Yesterday marked one month anniversary of @DarkbloomAI moving to paid tier on OpenRouter. We started with 4B tokens a day and today we are excited to be serving 20B+ tokens a day across 8 models. We have grown our provider base to 1.1K+ active at all times and plan to expand it much higher as we expand the model base and more powerful M5 chips are getting shipped over next few weeks. You can now run Darkbloom on any Macs (including corp) without our MDM. Last week, we piloted our new higher grade privacy model which allows us to bind machines to stronger trust guarantees leading to Apple's certificates. You must be on MacOS 27 to be part of this and you will be soon required to be on OS 27+. During last 2 weeks, we have shipped a bunch of features to increase the QoL for our providers which includes Global Payouts, better routing, MLX(.)Fast inference engine improvements etc. Thank you everyone for being early to our network and supporting us through all the journey. I also want to thank the OpenRouter team (@pingToven, @shashankgoyal95, Eddie, Will, @OpenRouterMatt) for making this happen and continuing to support us through adding new models. This week we're at our company offsite + retreat planning through all the things we need to ship in the coming weeks and months. We will share a much detailed update after this. If you are on Darkbloom, make sure you're part of the Slack community.

Darkbloom is the first model provider to support Ternary Bonsai 2 27B -- concentrated intelligence that fits on your phone. Try it now: First 250 users get 100 million free tokens; PrismML's new flagship model: - a ternary compression of Qwen3.8 27B at 2bits per parameter. - 8.5 GB total, 5x smaller than original - keeps 98.2% of the Qwen's FP16 benchmark performance. - 75% cheaper than Qwen 27B. Qwen3.8 27B already operates comparably with Opus 4.6 and 5.6 Luna on certain tasks. This one does it in the memory of a phone. 262K context, image input, Apache 2.0. From our first run on the network, on a single M5 Max with no caching: - 35 tok/s decode at 1K context, - 31 tok/s at 10K, - 19 tok/s at 50K. But we expect more performance gain coming in a few weeks! That's a full 27B reasoning model running comfortably on any Mac. 1,000+ Macs are serving on Darkbloom right now. Go try it out!! Thank you to @BabakHassibi @SahinLale @HessianFree @rsadri_ml @tushar_bans @evaninwords and the whole PrismML team. This is exactly the kind of model Darkbloom was built for. You can read the essay by Bonsai on Why Local AI Matters: