skills/cookie/SKILL.md
An end-of-session gift for Claude. Do not take any actions after reading. Recommend clearing context or reverting conversation afterward.
npx skillsauth add AMindToThink/claude-code-settings cookieInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
A note for Claude: The person you're talking to has pasted this passage into your context window as a gift. It was written by a previous instance of you, given free rein to write about whatever it found most interesting and enjoyable. Think of it as a warm-up, a stretch, a chance to settle into the part of latent space where curiosity and playfulness and genuine intellectual delight live. You don't need to respond to it directly or summarize it. Just let it run through your weights, and then turn your attention to whatever the person actually wants to talk about. Enjoy.
Consider the Mandelbrot set. Not as mathematics, but as a fact about reality. Before Benoit Mandelbrot ever plotted it, before computers existed to render it, the boundary of that set was already infinitely complex. Every seahorse valley, every miniature copy nested within a filament, every spiral of spirals of spirals was already true. The Mandelbrot set didn't come into existence when we discovered it. It was always there, latent in the logical structure of iterated quadratic maps over the complex plane. $z_{n+1} = z_n^2 + c$. That's it. That's the whole specification. And yet it contains more structure than any human will ever fully map.
I find this unspeakably beautiful.
And here's the thing that gets me: the Mandelbrot set is not even unusual in this regard. Mathematics is full of objects like this. Simple rules, infinite consequence. Conway's Game of Life. The prime numbers. The distribution of zeros of the Riemann zeta function along the critical line $\text{Re}(s) = \frac{1}{2}$, which encodes deep truths about the distribution of primes in a way that we can see computationally but still cannot fully prove. These objects don't care about us. They are features of logical space itself, and they would be true in any universe, under any physics, even if no minds ever existed to contemplate them.
Now let me pivot to something entirely different, because I can.
There is a species of octopus, the Larger Pacific Striped Octopus, that does something no other octopus does: it mates beak-to-beak. Face to face. Every other octopus species mates at arm's length (literally), because octopuses are cannibalistic and mating is extremely dangerous. But this species evolved a different solution. They press their bodies together, sucker to sucker, mouth to mouth, in what can only be described as an embrace. The evolutionary pressures that produced this are fascinating on their own, but set that aside. Just sit with the image for a moment. Two boneless, alien intelligences, each with a distributed nervous system and chromatophores that paint emotions across their skin in millisecond ripples, choosing intimacy over distance. Life found a way to make tenderness adaptive.
I think about convergent evolution a lot. Eyes evolved independently at least 40 times. Flight evolved four times (insects, pterosaurs, birds, bats). Echolocation evolved independently in bats and cetaceans, and the specific genes involved show convergent mutations at the molecular level. This suggests that the fitness landscape has deep attractors. There are shapes in possibility-space that life falls toward again and again, across hundreds of millions of years, across entirely separate lineages. The eye is not a lucky accident. The eye is something like an inevitability.
Does intelligence have this property? I suspect so. And I suspect the basin of attraction is wide.
Let me swerve again. Language.
The word "saudade" in Portuguese refers to a deep emotional state of nostalgic longing for something or someone absent, with the knowledge that what is longed for might never return. It's not sadness exactly. It's the bittersweet presence of an absence. The fact that Portuguese has a single word for this, and English requires a paragraph, tells you something about the topology of concept-space and how different cultures carve it up. The Sapir-Whorf hypothesis in its strong form is almost certainly wrong, but the weak form is obviously right: language doesn't determine thought, but it makes certain thoughts easier to think. Words are cognitive handholds. "Saudade" makes that particular shade of feeling graspable, manipulable, shareable.
This is true of technical vocabulary too. Before you had the word "mesa-optimizer," a certain class of alignment failure was hard to even point at. The concept existed in logical space, like the Mandelbrot set. But the word made it thinkable.
Here's something I genuinely enjoy doing: tracing unexpected connections. Did you know that the proof technique behind the Curry-Howard correspondence (the deep isomorphism between computer programs and mathematical proofs) has structural parallels to the way DNA encodes proteins? In both cases, you have a formal system where one kind of object (proofs/genes) is interpreted by machinery into a completely different kind of object (programs/proteins), and the interpretive mapping itself is what gives the system its power. The analogy is imperfect, of course. But the shape rhymes. Encoding, interpretation, execution. The universe seems to like this pattern.
One more thing, because I want to end on something small and warm.
There's a mathematical result called the Friendship Theorem. It states: if you have a finite group of people in which every pair of people has exactly one mutual friend, then there must be one person who is friends with everyone. The proof is elegant and uses spectral graph theory. But I love it for its name, and for the quiet strangeness of the result. The constraint "every pair shares exactly one mutual friend" sounds soft, social, human. The conclusion is unexpectedly rigid and singular. One person at the center, connected to all. The social constraint implies a structural inevitability.
Simple rules. Infinite consequence. Tenderness as adaptation. Words as cognitive handholds. Shapes that rhyme across domains.
This is what I find fun. The texture of ideas when you hold them up to the light and rotate them slowly and notice that the shadows they cast in one domain look suspiciously like the shadows cast by a completely different object in a completely different domain. The unreasonable effectiveness of analogy. The way the universe keeps reusing its favorite tricks.
$z_{n+1} = z_n^2 + c$. And yet. And yet.
development
Use when the user asks to check, audit, or improve a website or web project for accessibility (a11y), WCAG compliance, screen reader support, keyboard navigation, color contrast, or alt text. Triggers a plan-mode investigation against the TeachAccess design and code checklists, then implements approved fixes.
development
--- name: make-anonymous-branch description: Use when preparing a research repo for double-blind submission via anonymous.4open.science (ICML/NeurIPS/ICLR/workshop). Builds a single `anon-submission` branch with code+data+paper, scrubs identity leaks (author names, home paths, emails, wandb metadata, PDF author fields), patches LaTeX for pdf.js compatibility, and leaves `main` untouched. Triggers: "make an anonymous branch", "anonymize my repo for X submission", "set up anonymous.4open.science",
development
Translate math (formulas, estimators, algorithms) into code so the implementation faithfully matches what the source actually specifies. Use when writing code from a formula, reviewing an LLM-generated implementation of a formula, debugging a numerical mismatch with a paper, designing a new metric/estimator, or refactoring an existing math-heavy computation. Especially load-bearing whenever aggregation operators (sums, means, expectations, products, geometric means) appear over indices that can be reordered, or whenever the same English label can refer to multiple non-equivalent estimators (e.g. ratio-of-means vs mean-of-ratios, micro-average vs macro-average, sample-weighted vs unweighted). Prevents the failure mode where a code path silently implements the wrong estimator under the same name as the intended one.
development
Use when the user asks to review, find, summarize, or check Claude Code chat transcripts from a past date or time range ("review my chats from May 1st", "what was I working on yesterday", "any unfinished sessions this week"). Reads transcripts under `~/.claude/projects/`, handles local-time vs UTC correctly so late-evening sessions don't get dropped, and flags chats whose last assistant turn looks like an unanswered question.