Sister-in-law, 18 months into her PhD. Literature review consuming her existence. 247 papers to process. Advisor wanted synthesis by month end. Called me crying. "I can't read fast enough." THE ACADEMIC OVERWHELM Each paper: 20-40 pages. Dense methodology sections. Results scattered across tables and figures. Citations leading to more papers. Taking notes by hand. Losing track of what she'd read. Couldn't remember which paper said what. "Did Study A or Study B find the correlation? I know I read it somewhere." Searching through 247 PDFs manually for one finding. Hours lost. THE RESEARCH HELPER I BUILT Papers go into folder organized by topic. Workflow processes each one. Extracts key information. Authors, methodology, sample size, main findings, limitations, how it connects to her research question. Searchable database. "Which studies found correlation between X and Y?" Instant answer with citations. Cross-references findings. "Studies A, C, and F found positive correlation. Study B found none. Study D found negative." Generates properly formatted citations automatically. THE ADVISOR'S REACTION Submitted literature review. Advisor's response: "This is remarkably well-organized. Did you hire help?" She didn't. Just stopped drowning and started swimming. "The synthesis section reads like you actually understood the connections between studies." She did. Because she could finally see them. Defense is next year. Says the research system is the only reason she's still in the program. The extraction struggles with heavily formatted academic papers. Lots of equations and charts sometimes confuse it. But captures methodology and findings reliably. PhD students are drowning. This is a life raft. What research are you trying to make sense of?

Spent the last month building Claude Code configurations that actually work. Subagent architectures, MCP setups, prompt chains — the stuff that takes your Claude Code from 'fancy autocomplete' to 'actual AI teammate.' Happy to share the thinking if anyone's interested.