𝑘𝑒𝑟𝑛𝑒𝑙𝑡𝑟𝑖𝑐𝑘

jensen huang is having a third affair with my Fidelity account right now

We managed to find a way to backdoor the next release of OpenAI Codex 🔥 This is the best exploit chain I’ve ever witnessed depi.security/blog/20261111-…

Across three ProgramBench rebuilds, MiMo V2.6 Pro shows three moods. It waves off pixel-matching on tui-journal (“I’ll stop pixel-fucking”), loses its mind over an uncapitalized “q” on srgn, and writes “FUCK FUCK FUCK” over inconsistent results on html-to-markdown.

Weight space programming will replace fine tuning. You guys asked about multiple skills. Here you go, edited Qwen3.5-0.8B's weights directly to add decision skills from our vega-U engine. No new dataset and no training: the engine answered the training rows of public benchmarks, and its judgement was written straight into Qwen's weights with a closed-form solve that took 46 milliseconds, on two free Kaggle T4 GPUs. The edited model now makes better decisions on its own, with no catastrophic memory loss: it still knows what it knew before. Notebook: kaggle.com/code/nandakishorc…

In 2014, I sold my first company Loom to Dropbox. Today, I’m proud to announce Roma 2.0 (YC F26): the first to-do list for AI agents. Everyone uses a different agent: Claude, Grok Bot, Muse… Until now, none of them could work together. Roma automatically creates tasks and hands them to the right agent. Allowing you and your agents to work together, with shared context. Try it here: roma.app

This Diwali, a little more light from India's AI research community onto the global AI landscape. Introducing the latest benchmark results from Alpie Seed, our 9B multimodal reasoning model developed at @169Pi_ai Research. We believe the next generation of AI will be shaped not only by models that scale bigger, but by the research and engineering that make smaller models increasingly capable. We put Alpie Seed through a range of demanding evaluations across reasoning, scientific knowledge, mathematics, coding and software engineering. Here are some of the results: → 82.54% on MMLU-Pro → 84.32% on GPQA Diamond → 72.99% on LiveCodeBench v6 → 61.20% on SWE-Bench Verified The comparison highlights some interesting results. Alpie Seed records the highest GPQA Diamond score among the models shown. On MMLU-Pro, it performs competitively with larger models, while its LiveCodeBench result is within striking distance of substantially larger systems. Its SWE-Bench Verified result of 61.20% is another important milestone, reflecting its ability to work through real-world software engineering tasks. This evaluation used mini-SWE-agent without retries. Beyond benchmark accuracy, we also examined inference efficiency. In a separate profiling evaluation, Alpie Seed recorded: → 36.66 ms time to first token → 41.37 tokens/s generation speed → 9,796.22 tokens/s prompt processing speed These measurements come from a separate profiling setup and should not be confused with the comparative benchmark results. What makes this work meaningful to us is the process behind it: developing a compact model through supervised fine-tuning and reinforcement learning, evaluating it across diverse tasks, and making the resulting weights accessible to the wider community. The comparison includes published results from other model developers. These evaluations were not rerun under a common harness, so differences in methodology and configuration should be considered when interpreting the numbers. Built in India. Developed for a global research and developer community. This is Alpie Seed. And this is one step in a much larger journey for 169Pi Research. Explore the model: Hugging Face: huggingface.co/169Pi/Alpie_s… GGUF: huggingface.co/169Pi/Alpie_S… Ollama: ollama.com/169pi/alpie-seed Kaggle: kaggle.com/models/alpie/alpi… Happy Diwali from everyone at 169Pi.

At this point, I'm working under heightened assumption of "If you've ever used Claude Code/Codex, even once, with internet access, your computer's likely compromised" In a few months/year, if new cybersecurity best practices emerge, I'll wipe my laptop + start over

To prevent honeypotting, we must turn to an ancient solution: sensitive information should only be handled by eunuchs

I think people that fear loss of meaning from automation are less than animals. Human field oxen. Do you not have any friends? Family? Have you not once felt the burden of being too busy for someone that needed help? No hobbies or dreams? Is the only thing that gives you joy in life a sense of superiority over your peers? striverism delenda est

Introducing Morph Ultra Unlimited Kimi and DeepSeek for $200/month 1,000 subscriptions available Details below morphllm.com/products/ultra

This Diwali, a little more light from India's AI research community onto the global AI landscape. Introducing the latest benchmark results from Alpie Seed, our 9B multimodal reasoning model developed at @169Pi_ai Research. We believe the next generation of AI will be shaped not only by models that scale bigger, but by the research and engineering that make smaller models increasingly capable. We put Alpie Seed through a range of demanding evaluations across reasoning, scientific knowledge, mathematics, coding and software engineering. Here are some of the results: → 82.54% on MMLU-Pro → 84.32% on GPQA Diamond → 72.99% on LiveCodeBench v6 → 61.20% on SWE-Bench Verified The comparison highlights some interesting results. Alpie Seed records the highest GPQA Diamond score among the models shown. On MMLU-Pro, it performs competitively with larger models, while its LiveCodeBench result is within striking distance of substantially larger systems. Its SWE-Bench Verified result of 61.20% is another important milestone, reflecting its ability to work through real-world software engineering tasks. This evaluation used mini-SWE-agent without retries. Beyond benchmark accuracy, we also examined inference efficiency. In a separate profiling evaluation, Alpie Seed recorded: → 36.66 ms time to first token → 41.37 tokens/s generation speed → 9,796.22 tokens/s prompt processing speed These measurements come from a separate profiling setup and should not be confused with the comparative benchmark results. What makes this work meaningful to us is the process behind it: developing a compact model through supervised fine-tuning and reinforcement learning, evaluating it across diverse tasks, and making the resulting weights accessible to the wider community. The comparison includes published results from other model developers. These evaluations were not rerun under a common harness, so differences in methodology and configuration should be considered when interpreting the numbers. Built in India. Developed for a global research and developer community. This is Alpie Seed. And this is one step in a much larger journey for 169Pi Research. Explore the model: Hugging Face: huggingface.co/169Pi/Alpie_s… GGUF: huggingface.co/169Pi/Alpie_S… Ollama: ollama.com/169pi/alpie-seed Kaggle: kaggle.com/models/alpie/alpi… Happy Diwali from everyone at 169Pi.

I'm the Founder/CEO of @therealquippy, a @ycombinator backed company, and we are expanding our headcount! Our mission is to help people improve their social lives and build social confidence through daily practice. You can check us out on the app store! If you want to work on a product that actually helps people and also pays competitively, look no further. We’re currently hiring for the following roles: •Product Engineer •Product Designer •Marketing Lead •iOS Software Engineer More details regarding each role can also be found on our website. You don't need the perfect resume to apply. We care more about ambition, speed, and being exceptional in some way (I.E. being the world's fastest coca cola drinker, or world's foremost expert on pinecone anatomy). To be considered for a position: 1) Follow @therealquippy 2) Like this post 3) Comment your email We offer competitive pay, comprehensive benefits, and flexible PTO. If you have any questions, feel free to send me a DM.

Sandbox devs, if all you need from the host OS are cgroups, seccomp, and namespaces, and KVM today, which is a very small subset of the Linux kernel, Why do you even need Linux on the Host?

Pervert oomf who is actually really knowledgeable