- Deep Research is asking LLMs to turn a fuzzy question (that needs interpretation) into a precise answer (that needs accuracy). Since LLMs are great at understanding meaning and bad at pulling exact facts—while computers are the opposite—this leads to inevitable mistakes, making AI a supercharged intern rather than a fully reliable tool and leaving open the question of whether it’ll ever be 100% right or stay a useful but flawed commodity https://www.ben-evans.com/benedictevans/2025/2/17/the-deep-research-problem
- Generative AI evolves rapidly, but evaluating progress is tricky because while LLMs excel at probabilistic tasks like content generation, where mistakes are subjective or easy to correct (e.g., marketing and coding), they fail at deterministic tasks requiring precise, verifiable facts (e.g., database-style retrieval), making them powerful assistants but unreliable sources of truth—raising the question of whether future models will ever eliminate errors or if AI’s true potential lies in entirely new, non-deterministic applications where approximation is a feature, not a bug https://www.ben-evans.com/benedictevans/2025/1/the-problem-with-better-models
- To make the most of Cursor AI, set up a .cursorrules file in your project’s main folder. This file acts like a guidebook, telling Cursor how to handle your code. By including specific instructions—like which coding styles to follow or which tools to use—you help Cursor generate code that fits your project’s needs. For example, you can tell it always to use async/await instead of .then() for handling promises or to avoid certain libraries. This setup ensures that the AI’s suggestions align with your project’s standards, making your coding process smoother and more efficient https://ghuntley.com/stdlib/
- Enterprise software is not about making a coffee machine, it is about putting a coffee machine in every room for 1000s of hotels and maintaining them — Joel Spolsky https://en.wikipedia.org/wiki/Joel_Spolsky
- For a well-paced two-week trip, immediately travel to the farthest, most remote destination upon arrival, then slowly work your way back toward the departure city, giving yourself 10 days of immersive exploration before enjoying the comforts of urban life at the end—this keeps the journey exciting, minimises fatigue, and enhances both rural and city experiences. https://kk.org/thetechnium/50-years-of-travel-tips
- Walking more could add as much as 11 years to your life https://fortune.com/well/article/walking-to-live-longer/
What I’ve been reading (W8, 2025)
Deep Research and the pace of generative AI