The Last One Percent (OUTLINE)

productivity

education

my recent thoughts after bringing some projects to completion.

Author

Amiri Hayes

Published

May 21, 2026

The Pareto Principle (The 80/20 Rule): Introduce the idea that the first 80% of a project takes 20% of the time, while the final 20%—and truly, the last 1%—demands an asymmetric amount of energy, discipline, and attention to detail.
The Nature of Sprints: Frame the post around the concept of the “final sprint” across three massive milestones that hit all at once this month: closing out a national lab internship, submitting a year-long research project to top-tier AI conferences, and graduating from college.

Structure and Management: Contrast the structured, consistent 40-hour work week at BNL with the more chaotic nature of academic research. Detail the hands-on management style and the value of co-managing a project divided into three distinct phases.
The Phenomenon to the Minute: Discuss the realization that even when working with massive climate datasets, the quality of downstream ML tasks relies entirely on understanding the phenomena at the lowest, most granular level.
Perfecting the Pipeline: Explain how the last 5% of the internship was spent obsessively perfecting the first two stages to guarantee the success of the third:
- Stage 1: Cold front identification via sliding window statistical measures.
- Stage 2: Visualization and understanding through composite plus correlation plots (allowing for precise, small-scale timing adjustments).
- Stage 3: Anomaly detection utilizing softmax regression from ground truth.

The Horizon of a Long-Scale Project: Look back at a project with a much wider timescale (stretching since June of last year), marked by months of floating ideas, pivoting research questions, and rebuilding prototypes before the final submission to the ICML Mechanistic Interpretability workshop and NeurIPS.
The Pareto Rebuild: Tie back to the opening principle. Explain how months of varied approaches left a fragmented codebase. The final 1% required completely rebuilding the core library from scratch using a single, unified strategy for program synthesis (strict prompt structures, refinement loops, and curated examples).
The Final Gating Item: Detail the pressure of having to wait for this foundational code rebuild to finish before executing the ultimate experiments: analyzing the downstream effects of replacing real LLM internal activations with the synthesized Python program outputs.

Looking Back at the Foundation: Connect the current graduation milestone to the 8-semester sprint in high school that resulted in three Associate Degrees (Physics, Computer Science, and Mathematics)—a relentless, max-credit pace through summers and winters that left you exhausted.
Moving Past the Finish Line: Address the perspective of family and friends who suggest celebrating accomplishments more. Pivot this into a personal defense of the trait: not dwelling on wins means you don’t dwell on failures either.
The 1% Left: Describe the shift as the final weeks turned to days, and the days turned to the night before the ceremony. Instead of a technical sprint, the final 1% was intentionally spent resting, building a custom fixed-gear bicycle, skating, and spending time with loved ones.
The Shift from Learning to Creating: Introduce the word inflection—the transition from a desire to simply learn to a desire to create as the BS/MS program comes to an end. Contrast the sprints of four years ago (focused on passing tests) with the sprints of today (focused on building interesting, original things).
Conclusion: Note that while these experiences look different on the surface, they differ only in structure, not in their fundamental nature. Express excitement for the next technical sprints ahead this summer at IBM Research and while wrapping up the Master’s degree next semester.