Inspiration links are the genre sources each direction descends from — worth visiting before judging the exploration.
01 · The Terminal
Inspirations: Berkeley Graphics (engineering-tool austerity) · gwern.net (credibility through density) · The Monospace Web · classic MUD/BBS interfaces. Thought process: Your buyers increasingly include engineers and AI-literate founders; a designer who presents like a shell says "I live where you live." The working prompt is the differentiator — most terminal-themed sites are costumes; this one accepts commands and returns honest errors, which is your whole brand in one interaction. Will it work for clients? Works hard for technical founders and platform teams; filters OUT non-technical recruiters instantly (a dark monospace wall reads as "developer," and they'll misfile you). Verdict: killer as a hidden easter egg or /terminal route, wrong as the front door.
02 · The Broadsheet
Inspirations: NYT front page grammar · [The Deck-era editorial sites] · Reuters Graphics · David Hillman's Guardian redesign (print grid discipline). Thought process: Case studies are literally stories with outcomes — the newspaper genre converts claims into headlines and forces deck-quality summarization. It also flatters your copy system: claims-not-labels IS headline writing. Will it work for clients? Strong for brand-side and content-heavy orgs; hiring managers get instant scannability. Risks: reads "concept project" to conservative enterprise buyers, and the genre steals attention from the work itself. Verdict: viable front door with restraint; best borrowed as an annual-review artifact ("Year in Untangling").
03 · The Gallery
Inspirations: David Zwirner / Gagosian exhibition pages · museum wall-label conventions · Cargo portfolio culture. Thought process: You now own a 21-piece illustration system — the gallery direction asks "what if the images ARE the argument?" One plate per viewport imposes the slowest reading pace of all 15; wall labels force one-line claims. Will it work for clients? Beautiful for design-director audiences who judge taste in 10 seconds. Weak for recruiters who need evidence fast — outcomes are a click deeper than they'll go. Verdict: the pacing and label discipline should be stolen (the main site's dossier could end in one full-bleed plate); as a whole site it's style over proof.
04 · The Ledger
Inspirations: Brutalist Websites · craigslist (density as honesty) · [Frank Chimero's writing on tables] · annual-report typography. Thought process: The inverse bet from 03: kill all art direction, let volume speak. A sortable table says "I have enough work that it needs sorting" — quantity as the qualitative claim. Expandable receipts keep the evidence rule. Will it work for clients? Excellent for skeptical engineering leaders and procurement-minded enterprise buyers; the AI-native sort column quietly tells your strategic story. Risky for design-led shops — they may read "no craft." Verdict: closest to shippable of the wild ones; also the best print/PDF version of you.
05 · The Map
Inspirations: [Compas itself] (your thesis, applied reflexively) · Muse's spatial canvas · Figma/FigJam boards · open-world game maps · poolsuite.net (play as brand). Thought process: If your bet is "until we see everything on a map, nothing makes sense," the portfolio should be able to prove it. Islands = projects, regions = domains, you at the center. Drag-to-travel makes the visitor DO the untangling. Will it work for clients? The most memorable of all 15 — and memorability is the scarcest resource in a hiring pipeline. Fails keyboard users, fails hurried recruiters, fails mobile halfheartedly. Verdict: never the only door; extraordinary as a "view work as a map" toggle on the main site, where it doubles as a live Compas demo.
06 · The Slides
Inspirations: conference-talk culture (speakerdeck) · Stripe's early pitch pages · [Spotify Wrapped] pacing (one fact per screen). Thought process: Hiring conversations end in a panel room anyway — this direction skips ahead. One claim per slide enforces your front-loading rule ruthlessly; the progress bar creates completion pressure a scrolling page never does. Will it work for clients? Recruiters finish it — that alone is rare. But it hides depth: no skimming, no jumping, and design leaders may want the craft between the claims. Verdict: wrong as the site, PERFECT as steveux.com/deck for the "send us something before the panel" email.
07 · The Interview (chat)
Inspirations: AI chat UIs (the genre your buyers now live in) · [poke.com]-style conversational sites · Interactive fiction (choice-driven pacing). Thought process: Every visitor arrives with questions; this direction admits it and answers them in order of their choosing. Pre-written honesty ("replies are canned, obviously") turns the gimmick into a trust move. Quick-reply chips are the depth system reborn as conversation. Will it work for clients? Charming for founders; demonstrates conversational-UX literacy natively. Dangerous ambiguity: visitors may think it's a live model and probe it — the "pre-written" label must stay loud. Verdict: strong candidate for the About page's alter ego; as the whole portfolio it caps how much evidence you can show.
08 · The Annotated One-Pager
Inspirations: gwern.net sidenotes · Butterick's Practical Typography · legal redlines / track changes · director's-commentary tracks. Thought process: Every hiring loop ends at the résumé anyway — so make the résumé the site, and let the margins carry the voice. The two-register trick (formal document + handwritten commentary) shows judgment AND personality without mixing them. Will it work for clients? Recruiters feel instantly at home (it's the artifact they wanted), and the annotations reward the ones who linger. The handwriting font risks cuteness; margin notes vanish poorly on mobile. Verdict: highest hire-signal per pixel of all 15 — the strongest real challenger to the current site for a job-search season.
09 · The Mixtape
Inspirations: poolsuite.net · teenage.engineering (hardware-as-interface) · Winamp/cassette skeuomorph nostalgia. Thought process: Pure charm play: projects as tracks, sides A/B mapping enterprise/AI-native, a fake progress bar making people WAIT with a project title on screen (enforced attention, disguised as play). Will it work for clients? Founders and brand teams smile; enterprise hiring managers wonder if you're serious. Zero evidence surface — everything is a link out. Verdict: won't work as the portfolio; would work brilliantly as a Side Quest tile or a conference-bio link. Keep the A/B side conceit — it's a smart taxonomy.
10 · The Blueprint
Inspirations: Berkeley Graphics · real engineering title blocks · patent drawings · NASA schematic culture. Thought process: Reframes design as ENGINEERING — parts, tolerances, revision blocks, "checked by: the taste file." For orgs that respect rigor over vibes, the genre itself argues you're a systems person. The inverted-image filter proves the illustration system survives any ground. Will it work for clients? Resonates hard with platform/infrastructure companies (your actual target). Downsides: text-dense at small sizes, and the blue monoculture flattens the work's variety. Verdict: niche front door, superb case-study TEMPLATE — imagine the API study drawn as DET.01…DET.06.
11 · The Pinboard
Inspirations: detective evidence walls (True Detective, It's Always Sunny meme) · are.na · Milanote · studio crit walls. Thought process: Research culture made visible: the portfolio as an investigation into complexity, with you as the person of interest. Draggable exhibits + red string that follows = the one interaction people will screen-record and share. Will it work for clients? Design teams (your peers-as-gatekeepers) love it; it looks like how work is done. Recruiters get lost — no order, no hierarchy of importance. The meme association cuts both ways. Verdict: not the front door; ideal as the Compas or research-methods page, where "evidence wall" is literal.
12 · One Sentence
Inspirations: HTML Energy · lynnandtonic.com (radical smallness) · [justfuckingusehtml]-genre minimalism, done politely · print poster typography. Thought process: The maximal restraint test: if your positioning sentence is truly load-bearing, it should function as the entire site — six phrases, six doors. It stress-tests copy the way the terminal stress-tests credibility. Will it work for clients? Confidence reads instantly; design directors respect the nerve. But it demands curiosity — a tired recruiter sees one sentence and leaves. Verdict: the best LANDING state of all 15. Serious option: this AS the homepage hero moment before the bento reveals on first scroll.
13 · StevenOS (desktop)
Inspirations: [poolsuite.net] · windows93.net · Susan Kare-era Mac nostalgia · portfolio-as-OS microgenre (kernel.style etc.). Thought process: You literally run a "personal OS" — the metaphor is already in your positioning, so the desktop makes it visible. Draggable windows show interaction chops with zero framework; the Trash is the best joke on any of the 15 ("gradients — do not restore"). Will it work for clients? Memorable and shareable; proves front-end fluency. But genre-saturated (many indie portfolios do OS bits) and the playfulness can undercut "enterprise-grade." Verdict: won't beat the main site for hiring; WOULD make a legendary 404 page or /os easter egg. The Trash window should be stolen immediately.
14 · The Wiki
Inspirations: Wikipedia's neutral-voice authority · [gwern.net] citations · fan-wiki culture. Thought process: The most subversive bet: write about yourself in the third-person neutral voice, and let the CITATION apparatus carry the evidence rule — every [n] links to a real case study. Familiar genre = zero learning curve, instant scannability (infobox = your facts row). Will it work for clients? Recruiters parse it faster than any custom design (they read Wikipedia daily). The self-aware jokes ("citation cheekily needed") keep it from feeling delusional. Risks: borrowed authority can read as borrowed authority. Verdict: sneaky-strong for a about/press page; the infobox pattern deserves a place on the real About.
15 · End Credits
Inspirations: Art of the Title · film end-crawl conventions · [Semi-Permanent] title sequences · the "no modals were harmed" footer, expanded to a genre. Thought process: A career as filmography: roles instead of pitches, collaborators credited (including the AI pair and the dog), auto-motion doing the pacing. Hover-to-speed-up hands the projector to the visitor. Will it work for clients? Emotionally the warmest of all 15 — it's the one that makes people finish and email. Terrible for skimming, invisible to SEO, useless on a second visit. Verdict: not a portfolio; a perfect FINALE — the last section of the main site's About page, or the page you send with a resignation letter.
The meta-read (what 15 explorations taught us)
- The strongest single challengers to the current site: 08 (annotated one-pager) for a job-search season, 04 (ledger) for enterprise credibility, 12 (one sentence) as an opening state.
- Steal, don't switch: 03's closing plate, 05 as a work-map toggle, 06 as /deck, 13's Trash joke as the 404, 14's infobox on About, 15 as the About finale.
- The pattern in the verdicts: directions fail when they hide EVIDENCE behind charm. The current site wins because depth/evidence/honesty are structural — every exploration that kept those (04, 08, 14) got a "shippable" verdict; every one that traded them for delight (09, 13, 15) got "easter egg."
- Client lens, condensed: recruiters need scannable proof in 30s (04, 08, 14 deliver); design directors need taste signals (03, 12); founders need memorability (05, 07, 13). The standing site is the only artifact serving all three — which is itself the argument for keeping it as the front door.
Rebrand pass (2026-07-06, second edition)
Every exploration now carries a distinct visual identity — no two share a palette, and none default to the house system: 01 phosphor-CRT green/amber on near-black · 02 newsprint white + oxblood (Playfair) · 03 white-cube + terracotta (Cormorant) · 04 banker's green on ivory, typewriter mono · 05 parchment atlas + compass red (Fraunces, sepia plates) · 06 keynote black + signal yellow (Space Grotesk) · 07 iOS messenger grays + system blue · 08 manila folder + blue/red pens · 09 vaporwave violet/pink/cyan with glow · 10 graphite chalkboard + safety orange · 11 true corkboard + white cards · 12 vermillion poster + cream (Archivo 800) · 13 platinum-Mac grayscale + cobalt · 14 Wikipedia dark mode · 15 black & gold (Cormorant). The lesson available here: branding is a variable, not a constant — the content survived fifteen identities without a single copy change, which is the strongest evidence yet that the copy system is doing the heavy lifting.
The matrix (2026-07-06, third edition)
Every page now carries a skin switcher (bottom-left, "◐ Skin"): any of the 15 layouts can wear any of the 15 brand identities — 225 combinations, choice persists as you move between explorations. This works because every layout was built on CSS custom properties from the start: one 4KB engine (skins.js) remaps ~20 token names and the entire identity swaps live, no reload. Native restores each page's own brand. Some combinations clash on purpose — a matrix that only produced tasteful output would be lying about the search space. Point demonstrated: separating structure (layout), identity (tokens), and content (copy) means iteration cost approaches zero on any single axis.