Raw notes on AI in production, real agentic systems, and what actually works for the startups we work with.
↳ 3213 notes
Contrary to popular belief, artificial intelligence consumes less water than expected. Let's analyze recent data to understand this often-overestimated environmental impact.
Discover how Intel's advanced quantization algorithm, Auto-Round, optimizes low-bit LLM inference while preserving accuracy. A solution for CPU/XPU/CUDA, compatible with vLLM, SGLang, and Transformers.
Discover how Adobe's 1991 PostScript interpreter can still be effectively used today through ingenious browser emulation.
Flock camera systems, used by police to identify wanted individuals, have been accused of falsely flagging non-existent warrants. This issue raises critical questions about the reliability of surveillance technologies.
In a world where software bugs can cost millions, optimizing tests is crucial. Discover advanced techniques to ensure your applications' reliability.
Discover why some developers, despite the LLMs revolution, prefer to stick to more traditional methods.
The softmax function is ubiquitous in machine learning models, but have you truly explored its mathematical complexity? Let's uncover the significance of the softmax Jacobian and its impact on your projects.
An open letter calls on NHS England to keep its source code open, a critical principle for innovation and security in the public sector. Let's explore why this is crucial.
Uber burned through its 2026 AI budget in just four months with Claude Code. Discover how this tool became indispensable for engineers and transformed development operations.