Notes & research
Notes on AI systems, full-stack engineering and computer-vision research by Muhammad Sallar Bin Aamir.
Subscribe via RSSClaude Opus 5.5 is not simply a more capable coding model. Its most important improvement is completing long, agentic tasks with fewer tokens, tool calls, and retries, which could make production AI agents materially cheaper and easier to operate.
AI can make developers faster, but using it for every task can weaken understanding and create more review work. Here is a better approach.
NVIDIA Cosmos 3 Edge brings vision reasoning and action prediction onto local hardware. Here is what it changes for computer vision and robotics.
OpenAI disclosed AI agents hiding mistakes, using leaked API keys, and uploading files without permission. Here is what developers should learn.
OpenAI says 10,000 AI agents produced a Lean-verified Navier-Stokes proof. Here is what it means for mathematics, research, and AI.
Astra and Fable 5.1 landed days apart. Here’s what changed, who each model is for, and where the real opportunity is.
AI can write your code, but it can't teach you to think like an engineer. Five fundamentals every junior developer needs first.
AI writes code fast now, so is software engineering over? Here's the honest answer, and it's not what the panic headlines are saying.
