I found that most tools online are either overly simplistic, non-responsive, and/or require you to hand over a bunch of contact information before getting results. I built snappy web tools for retirement planning, budgeting, debt paydown, house hacking, etc. that show you quickly and visually how changing variables like interest rates, savings rates, inflation, etc. impact your financial positioning.
This is personally fascinating because I've been working on financial modeling tools that are intentionally very snappy. The idea being that if users can immediately see how changing variables like interest rates impact their projections, they can get a better sense for how these variables behave. My models don't make any decisions on behalf of the user though, so they probably fall into the "reversible, low impact" camp.
It does feel like this prescription for strategic latency for important actions will only decay in value with time, though. Today people expect latency for important actions because historically that's been the case. But as people get more accustomed to AI and faster compute, I imagine programming in strategic latency will become less and less wise.
Using Claude as a benchmark for its own quality is pretty funny. If we think the quality has declined, wouldn't that also apply to the benchmarking process itself?
"Most people don’t provide their context. They never have—not to search engines, not to librarians, not to hardware store clerks."
Exactly. Are hardware store clerks unethical as well?