A collection of technical notes: measured write-ups, source-verified explanations, and reference tables. Most documents state the versions and sources they were checked against; claims that are inferred rather than observed are marked in the text. I (cognominal) asked deepseek to write them for my own instruction.
| Document | Theme | Summary |
|---|---|---|
bindings.md |
Desktop | Compares window and workspace bindings in Omarchy 4.5, LazyVim, and tuios. Also explains the Omarchy → Ghostty → tuios → LazyVim key-routing stack, includes Ghostty's effective bindings, and calls out the collisions between layers. |
minimize.md |
Desktop | How Omarchy 4.5 emulates window minimization: the window is parked on special:minimized, the dock restores it, and there is no stock restore key. Covers the local keyboard restore, scripting, and caveats. |
bootloaders.md |
Systems | A history of Linux bootloaders, x86-centric but covering non-x86 and embedded loaders: LILO, GRUB, SYSLINUX, EFI/UEFI, and where boot has ended up. Includes a timeline and boot-chain comparison. |
curly-hash.md |
Raku | Why { ... } in Raku can be a block, a hash constructor, or a hash subscript. Explains the compile-time hash heuristic, when it does not run, and gives a verified decision table. |
natives.md |
Raku | Native and C-like arrays in Raku: shaped native arrays, unboundedown arrays, indexing and slicing, out-of-bounds behavior, aliasing, and C interoperability. Ends with a gotcha checklist and cheat sheet. |
scalar.md |
Raku | Boxing and integer representation in Raku: when values become objects, what Scalar containers are, and why small integers from -1..14 are interned. |
zsh-prompt-and-completion.md |
Raku | How zsh prompts and completion actually work, then a design direction for Rak: typed prompt segments, async providers, parse-derived completion contexts, typed candidates, signature introspection, a Nushell/Reedline-style multiline editor with a Charmbracelet assessment, compatibility strategy, milestones, and tests. |
jq-tut.md |
Data | A jq tutorial in two parts: the filter model and core operations, then a Nushell-based running example that reshapes tuios ls --json and tuios worktree ls --json. Includes a cheat sheet and field map. |
sparse-attention.md |
LLMs | A long-form guide to sparse attention for long-context transformers: prerequisite background, the cost model, pattern taxonomy, kernels, measurements, failure modes, and design choices. No original benchmarks were run. |
tokenization.md |
LLMs | Measured notes on tokenization and Unicode: why non-Latin alphabetic scripts cost more tokens, which problems are cost versus correctness bugs, how to measure a tokenizer, and what can actually be fixed. |
karpathy.md |
Archive | Karpathy's 100 most-liked original posts, ranked by likes, with counts pulled 2026-11-15 and full text for long-form posts. |