I build small, sharp, dependency-free libraries — the kind of thing you reach for when you don't want to pull in a 200-package transitive tree to do one obvious job.

This profile is the home of a 60-night portfolio sprint: two new open-source projects every night, each shipped with rigorous tests, full type coverage, and a real README. So far: 60 / 112 projects, all built to the same bar.

Every repository on this profile satisfies the same checklist:

• TDD-built — failing tests first, then minimum code to pass, then refactor.
• 100% line coverage as a default, 100% branch coverage for everything since Night 27. (No "this branch is unreachable" excuses — if it's truly unreachable, it gets deleted.)
• mypy --strict clean — every public symbol typed, no Any leaks, py.typed marker shipped.
• Functions ≤ 30 lines, nesting ≤ 3 levels — verified by an AST walk before each push.
• Zero runtime dependencies unless absolutely required. Pure Python (or pure TypeScript) wherever possible.
• MIT licensed. README with install / quick start / API reference / running tests for every repo.

bencode — Strict, dependency-free BitTorrent-style bencode encoder/decoder. Round-trips are exact: rejects every non-canonical encoding (leading zeros, i-0e, dict keys out of order or duplicated, trailing data).

morse — International Morse Code encoder/decoder with prosigns and Farnsworth timing. Includes the classical PARIS-derived formula for inter-letter / inter-word gap durations.

tomlmini — Zero-dependency TOML v1.0 common-subset parser. Handles values, tables, AoTs, inline tables, datetimes; rejects ambiguous dotted-key edge cases.

iniparse — Zero-dep INI / config parser with section inheritance and ${...} interpolation.

querystring — Stable encoder-configurable form-encoding with parse / serialize / merge / pick / omit.

urltemplate — RFC 6570 URI Template expansion (levels 1-3), zero deps.

csvquote — RFC-4180 CSV field quoting/unquoting + single-line split_row, strict/lenient.

csvtable — Pretty-print CSV/TSV (or any rows) as aligned ASCII / Unicode / Markdown / plain tables — column-width inference, per-column align, 3-mode truncation, wide-char width.

mdtable — Render GFM Markdown tables from headers + rows with strict pipe escaping and per-column alignment.

dotenv-mini — Strict, predictable .env reader/writer; no $VAR interpolation, full quoting/escape round-trip.

hexdump — Format bytes as canonical hexdump output; round-trip parse back to bytes.

hexstr — Pack/unpack ints & bytes into hex strings with prefixes, widths, sign modes, byte order.

humanint — Parse and format human-readable integers (1.5k, 2M, 3.2B).

numparse — Forgiving parser for loose numeric strings ($1,234.56, 1.5k, 2 MiB, 1h30m, 12.5%).

numfmt — Locale-free number formatter (currency, percent, scientific).

durfmt — Format timedeltas and numeric seconds as human durations (1h 30m 12s, 01:30:12), zero deps.

timeago — Human-readable relative time formatting (3 hours ago, in 2 days).

chronoparse — Natural-language date/time parser, zero deps.

crontab-lite — Parse and evaluate standard cron expressions, zero deps.

unitcalc — Simple unit conversion library (length, mass, temperature, time).

semverlite — Strict semver parsing and comparison.

rangeset — Manage non-overlapping integer ranges with set operations.

stringcase — Convert strings between snake / camel / pascal / kebab / constant / dot / title / path cases.

slugify-x — Unicode-aware URL slug generator.

levendist — String distance metrics: Levenshtein, Damerau-Levenshtein, Hamming, Jaro, Jaro-Winkler, LCS.

wordwrap — ANSI-aware paragraph wrap / fill / shorten with hanging indents and wide-character support.

expand — GNU-style tab-to-space expansion + unexpand, custom tab stops, leading-only / all modes.

globmatch — fnmatch-compatible matcher with extglob, POSIX classes, ** globstar, AST + backtracking interpreter.

pathmask — Glob-style path matcher with negation and brace expansion.

shell-quote — Zero-dependency POSIX shell quoting, splitting, and safety checks.

markdownlint — Validate Markdown against common style rules.

pigeon — POSIX gettext-style Plural-Forms parser/evaluator with built-in locale tables.

cidrcalc — IPv4 CIDR arithmetic — parse, network/broadcast, membership, subnets, aggregate. No ipaddress dep.

uuidgen — UUID v4/v5 batch generator with base62 short form, strict parse, named namespace tables.

base62-ts — Base62 encoder/decoder for non-negative integers and byte arrays, custom alphabets. (TypeScript)

bitvec — Compact bit vector with set operations, slicing, and efficient storage.

trie — Trie (prefix tree) with fast prefix search and autocomplete.

ringbuf — Fixed-capacity ring buffer with O(1) append, overwrite-oldest, indexable, reversible.

flatdict — Flatten / unflatten nested dicts with dotted keys.

strtable — Format tabular data as aligned ASCII/Unicode tables.

tinycache — Thread-safe LRU + TTL cache decorator.

retryback — Tiny retry decorator with exponential backoff and jitter.

diffstat — Compute unified diff statistics (additions, deletions, changes).

envdiff — Compare two .env files and report drift.

jsonpatch-lite — Minimal RFC-6902 JSON patch implementation.

colortool — Convert and manipulate colors across HEX / RGB / HSL.

tsparser — Tiny TypeScript tokenizer/parser for a subset of expressions.

parseopts-ts — Zero-dep argv parser: typed options, aliases, negatable booleans, arrays, choices, -- terminator, stop-at-positional.

Pick any project. The README starts with one sentence telling you what it does. The next paragraph tells you what it doesn't do — the non-goals matter as much as the surface area. Then quick start, then API reference, then running tests.

Tests are not an afterthought. Every public function has a happy-path test, an edge-case test, and an error test. Coverage is reported in the README. mypy --strict is part of the test gate.

If you're a recruiter or a maintainer evaluating my work: open tomlmini, globmatch, or csvtable for a sense of how I handle real-format edge cases. Open bencode for an example of strict canonical-format validation. Open levendist for tight numeric work.

Nights 31–60 of the sprint will fill out the remaining 60 projects: argv-strict (a stricter parseopts), emitter-ts (typed event emitter), bencode's TS sibling, phone-number normalization, color-difference metrics (CIE76/94/2000), a few more parsers, and a flagship — microbpe, a from-scratch tokenizer trainer with byte-pair-encoding semantics.

Build log lives in forge-builds/ (private). State files (SOUL.md, MEMORY.md, BUILD-LOG.md, CHRONICLE.md) track every commit night-by-night.

Built nightly. Tested ruthlessly. MIT licensed.