- You can use AI for writing help (emails, resumés, poetry), knowledge retrieval (e.g., caffeine content), coding and debugging, personal recommendations (like trips to Zermatt), internet searches, detailed topic research (on browser privacy or life extension), data analysis and visualisation, code editing integrations (“vibe coding”), voice interactions, image interpretation (nutrition labels, memes), generating visuals (YouTube thumbnails), customising conversations with memory and specific instructions, automating tasks with saved prompts, and enhancing comprehension by summarising books and research papers https://www.youtube.com/watch?v=EWvNQjAaOHw
- Every line of code you write introduces a potential bug, so only write necessary code. Adding unnecessary complexity—even something as small as an extra subtraction for optimisation—can create subtle and unexpected errors, like the harmful wait-time bug in the example. Avoid speculative routines, unneeded abstraction layers, and premature optimisations, as they’ll likely lead to harder-to-maintain code and mysterious failures in the future https://www.teamten.com/lawrence/writings/every_line_is_a_potential_bug.html
- Economists generally agree that raising tariffs on imports isn’t likely to reduce a nation’s trade deficit and can even worsen it. This is because trade balances mainly depend on other factors—like savings habits, government spending, economic growth, wealth, and demographics—not tariffs. Although currency manipulation can affect trade, it’s not currently a major issue with countries like China, Canada, Mexico, or the EU. Even if tariffs reduce imports by making them more expensive, they also strengthen the domestic currency, making exports harder to sell, which cancels out much of the intended benefit https://archive.is/sZRs9
- The attention mechanism helps language models understand words based on context. Each word becomes a high-dimensional vector—a query, keys, and values (e.g., GPT-3 uses 12,288 dimensions). For example, in the sentence “Look it up in the dictionary,” the query vector asks, “What does ‘it’ refer to?”, keys represent context words like “look” or “dictionary.” Values carry possible meanings, such as “‘it’ refers to a word.” The model selects relevant values by comparing queries with keys, clarifying that “it” means something you can look up — a word https://bartoszmilewski.com/2025/03/06/understanding-attention-in-llms
- [Re-read] Learn to like everything more and building the skill of enjoyment: appreciate overlooked details, admire creators, build vocabulary for nuance, imagine perspectives from other times and places, explore context, find unique ways to enjoy, give things time, and embrace flaws as part of the whole. https://sashachapin.substack.com/p/how-to-like-everything-more
What I’ve been reading (W10, 2025)
You can use AI for writing help (emails · Every line of code you write introduces a potential bug