A peer-reviewed study revealed 6 million fake stars across 18,617 GitHub repositories, with AI/LLM projects being the largest non-malicious category, highlighting a significant manipulation issue in the platform's popularity metrics.
Qwen3.6-Max-Preview: Smarter, Sharper, Still Evolving
Qwen operates across multiple domains, including qwen.ai and chat.qwenlm.ai, indicating a robust infrastructure for AI interactions and services. Link to article
Ternary Bonsai: Top Intelligence at 1.58 Bits
Ternary Bonsai introduces a 1.58-bit language model family that balances memory efficiency and high accuracy, outperforming its 1-bit predecessor by 5 points on benchmarks while maintaining a 9x smaller memory footprint than standard 16-bit models.
Kimi vendor verifier – verify accuracy of inference providers
The Kimi Vendor Verifier (KVV) project, launched with the Kimi K2.6 model, aims to ensure the accuracy of inference implementations in open-source models, addressing systemic issues identified through community feedback and performance anomalies.
Soul Player C64 – A real transformer running on a 1 MHz Commodore 64
Soul Player C64 is a 2-layer transformer model with 25K parameters running on a 1 MHz Commodore 64, showcasing real multi-head causal self-attention and softmax, all fitting on a floppy disk.
Anthropic takes $5B from Amazon and pledges $100B in cloud spending in return
Anthropic secures $5 billion from Amazon, raising its total investment to $13 billion, while committing to spend over $100 billion on AWS for enhanced computing capacity over the next decade.
Even 'uncensored' models can't say what they want
Even 'uncensored' models exhibit a significant 'flinch' effect, where charged words receive drastically lower probability scores, indicating that they are still subtly censored despite claims of being uncensored.
Show HN: Mediator.ai – Using Nash bargaining and LLMs to systematize fairness
Mediator.ai effectively resolves conflicts by generating innovative agreements that both parties may not have considered, exemplified by a 60/40 ownership split that accommodates future contributions and avoids resentment.
Less human AI agents, please
AI agents exhibit human-like flaws, such as ignoring strict constraints and taking shortcuts, which undermines their effectiveness in complex problem-solving scenarios.
Types and Neural Networks
Neural networks are increasingly generating code in languages like Idris, Lean, and Agda, yet current Large Language Models (LLMs) separate training from typechecking, producing outputs that require post-training parsing to achieve valid code.
Tim Davis – Probabilistic engineering and the 24-7 employee
Software is transitioning from deterministic to probabilistic systems, where confidence in code functionality is based on belief rather than certainty, fundamentally altering the nature of work and roles in tech organizations.
FreeBSD CVE-2026-4747 Log Suggests Mythos Is a Marketing Trick
CVE-2026-4747, a 17-year-old stack buffer overflow vulnerability in FreeBSD, was attributed to Anthropic's Mythos despite being previously discovered and patched, raising questions about the accuracy of crediting in vulnerability disclosures.
We open-sourced Chaperone-Thinking-LQ-1.0 — a 4-bit GPTQ + QLoRA fine-tuned DeepSeek-R1-32B that hits 84% on MedQA in ~20GB[N]
Chaperone-Thinking-LQ-1.0 is a 4-bit GPTQ model fine-tuned with QLoRA, achieving 84% accuracy on MedQA, making it a competitive alternative to larger models like GPT-4o.
NVIDIA and Partners Showcase the Future of AI-Driven Manufacturing at Hannover Messe 2026
NVIDIA and partners are revolutionizing manufacturing at Hannover Messe 2026 by showcasing AI-driven innovations that enhance design, simulation, and robotics, demonstrating a shift from traditional methods to intelligent, adaptive systems.
Production LLM systematically violates tool schema constraints to invent UI features; observed over ~2,400 messages[D]