Hi HN, I’m William, and I’ve spent the last year trying to solve a problem that keeps every researcher up at night: The Fragmentation of Scientific Discovery. The Problem: The "Context-Switching" Tax In the current research landscape, we are drowning. We face an explosion of information, a non-stop barrage of "trending" papers, and the persistent "reproducibility crisis." But the silent killer is the tool-switching tax. Every time you move from reading a paper to writing notes, then to a different environment for code replication, and finally to a separate experiment tracker, your "flow state" is shattered. We spend more time managing tools than enjoying the thrill of exploration. The Failure: The All-in-One Trap We started Prismer for AI4S in early 2025 with a grand vision: an all-in-one platform for reading, writing, and experimenting. We iterated version after version. But honestly? The feedback was mediocre. We fell into a cycle of disappointment and frustration. We realized that by trying to do everything, we weren't doing any single thing well enough to break a researcher's existing habit. We decided to tear it all down and focus on perfecting individual components—starting with the paper-reading experience. The Turning Point: Why Open Source? Then OpenAI released Prism. It was a wake-up call. We realized that the future of AI for Science (AI4S) shouldn't be locked inside a single lab or a proprietary "walled garden." Scientific workflows are deeply personal and highly specialized. No single company can build the perfect workflow for every researcher. That’s why we’ve decided to open-source our entire codebase. What Prismer is now: Instead of a rigid platform, we are providing modular components that researchers can use to build their own custom AI4S workflows. We want to empower the community to co-create the infrastructure of discovery, rather than just being "users" of it. We’re at a stage where we’ve traded our frustration for a new sense of purpose. We’d love to hear your thoughts on: 1. What part of your research workflow feels the most "broken"? 2. What are your biggest hurdles in AI4S reproducibility?