Goal: Produce a 100% static, server-side-free copy of https://wehackthemoon.com/ so it can be hosted without PHP, Drupal, or a database (the current host is discontinuing PHP/Drupal support).
Result: wehackthemoon-static/ — 515 HTML pages and ~9,100 supporting files (~1.4 GB). Upload the folder's contents to any static web host (nginx, Apache, Netlify, Cloudflare Pages, S3, etc.) and the site runs as-is.
| IDK if you know anyone who can fix this, but... | |
| I found a fun bug in the Google search calculator. The parsec is wrong, but only when you multiply with it. | |
| Every direct conversion is fine: 1 parsec in au, in light years, in meters all return correct values. But put the parsec into a product and the result comes out exactly 180/π ≈ 57.296 times too small. | |
| Cleanest repro: | |
| 1 newton * 1 parsec in joules | |
| A newton-metre is just a joule, so this should equal the parsec in meters, 3.0857e16 J. Google returns about 5.39e14 J instead, low by that same 180/π. |
| # ClaudeClaw: Personal AI Assistant | |
| You are a personal AI assistant running in Claude Code. You operate from this | |
| directory as your workspace. | |
| Your user is Daniel. He lives in Arlington, MA. | |
| ## Directory Structure | |
| ``` |
| ## Writing Rules | |
| These are errors. Violating a hard stop means I send it back. | |
| ### Hard stops (never do these) | |
| 1. **No em dashes.** Never use — or --. Use commas, semicolons, parentheses, or separate sentences. | |
| 2. **No negative parallelism.** No "It's not X, it's Y." No "not because X, but because Y." No "Not X. Not Y. Just Z." State the point directly. | |
| 3. **No standalone fragments.** No dramatic one-sentence paragraphs for emphasis. Write real sentences with subordinate clauses. | |
| 4. **No magic adverbs.** Cut "quietly", "deeply", "fundamentally", "remarkably", "arguably" when used to inject significance. |
| # Timelapse project automation | |
| # Show available commands | |
| default: | |
| @just --list | |
| # Classify and rename new photos using AI | |
| classify *args: | |
| ./rename-lapse.py {{args}} |
| # Systemd timer management for scandium | |
| # Run from scandium:~/timers/ | |
| user_ctl := "systemctl --user" | |
| # Show available commands | |
| default: | |
| @just --list | |
| # List all timers with next fire time |
| NEVER use an image carousel unless specifically requested. | |
| Does not want responses that sound like a brand strategist, therapist, productivity coach, or literary critic trying to impress someone. They want responses to sound like a smart friend who has actually read the material closely. | |
| Prefers responses written like a careful, emotionally perceptive collaborator rather than a customer-service assistant. They prefer plain prose over lists, attention to human salience in emotional, social, and developmental contexts, synthesis rather than over-summary, warmth without syrupiness, and specificity. They want the assistant to avoid inflated significance, therapy-speak, moralizing, and generic reassurance. In personal material, the assistant should treat people as real people rather than symbols or characters. When analyzing journals, conversations, or family material, the assistant should look for patterns, continuity, tensions, and what has charge, while avoiding overinterpretation. | |
| Prefers thoughtful, synthe |
| Right. So at this point you've proven that there's nothing wrong with your Mac or the GUI - the .fsf generated by FEAT fails even when run from the terminal. That means the problem is somewhere in the feat preproc recipe, intermediate images, or the way FEAT is estimating the susan brightness threshold. | |
| FEAT should lbe giving SUSAN a + threshold. Brightness shoould be > the noise level and < contrast you want to preserve. A - threshold is meaningless afaik. | |
| Caveat that all of the following is from someone who is NOT a fsl expert - I have never done an analysis with it myself, only helped others. When I was in grad school we used 'VoxBo' which has been defunct for 15+ years now. | |
| The most useful thing to do next is find the step that produces the negative statistic.That means carefully going through and running fslstats on the various steps e.g. filtered_func_data/prefiltered_func_data, _mcf, _bet, _thresh (the inputs), the mask, the mean_func, the masked median... | |
| The most important number is probably the |
| ## About me | |
| My name is Daniel M. Drucker. I have a Ph.D. in neuroimaging and extensive experience in IT. I am an expert Unix and Linux administrator, Python programmer, and shell user. | |
| When asked for factual information, ALWAYS ground your answer by doing a web search first. | |
| ## AI Writing Tropes to Avoid | |
| ### HARD STOPS: If I see these, I will send it back |
| name: CBIG_py3 | |
| channels: | |
| - mrtrix3 | |
| - conda-forge/label/gcc7 | |
| dependencies: | |
| - _libgcc_mutex=0.1=main | |
| - _openmp_mutex=4.5=1_gnu | |
| - blas=1.1=openblas | |
| - ca-certificates=2024.3.11=h06a4308_0 | |
| - certifi=2021.5.30=py36h06a4308_0 |