https://dive.vladyslavpodoliako.com/changelog/ Edition history. What shipped, when, and the receipts behind each. en Wed, 20 May 2026 00:00:00 +0000 https://dive.vladyslavpodoliako.com/changelog/#edition-9-3 Edition 9.3|2026-05-20 Wed, 20 May 2026 00:00:00 +0000 Every page one click from the homepage. Tile grid grows to 3 rows. Shipped: - Homepage tile grid expanded from 8 → 12 tiles (3 rows of 4). Row 3 added: Swarms · GitHub Playbook · Setup showcase (62 skills · 32 agents · 12 plugins) · Research notes (what shifted this week). The pages I shipped this week are now first-class tiles, not secondary-row text links. - Hanging-pages cleanup — secondary text row rewritten to surface the remaining 10 previously-hanging pages: Build vs Buy · For your CFO · CLAUDE.md rules · Cowork setup · Vault starter · Weekend builds · Starter skills · 30-day plan · Sections · About. Every standalone page in the Playbook is now one click from the homepage. - Reachability audit complete — 27 user-facing pages audited (plus /the-bill which is intentionally sitemap-excluded). Zero orphans, zero hanging-from-homepage pages remain. Receipts: - Homepage tiles: 8 → 12 (+4 third-row tiles) - Pages reachable from homepage: 13 → 27 (every page surfaced) - Hanging pages (after audit): 0 (was 9 pre-audit) - Orphans: 1 — /the-bill, intentional sitemap exclusion https://dive.vladyslavpodoliako.com/changelog/#edition-9-2 Edition 9.2|2026-05-20 Wed, 20 May 2026 00:00:00 +0000 Forty-one chapters. The HTML-ization thesis gets its own chapter slot. Shipped: - New Chapter 41 — "Send the Link, Not the File" (~1,800 words, 9-min read). Chapter-length narrative companion to /html-first. The story version of the HTML-ization thesis: every report, pitch, audit, deck, model in the portfolio ships as a live interactive link in a private repo, not a PDF or slide attachment. Same recipe across the portfolio; the chapter tells the story, /html-first is the deep reference with the embedded case studies, the recipe, the gallery, and the 12 public examples in the wild. - Ch 41 placed in Part V (Building Site) — sits next to Ch 19 (Saturday builds), Ch 31 (six stages), Ch 38 (run until done). Also in SECTIONS "building" topical group. - Cmd-K — Ch 41 chapter entry + synonyms (html htmlization living-link artifact deploy private repo afc folderly qbr board update) - Hanging-pages audit + homepage secondary text row expanded — surfaced /swarms, /github-playbook, /showcase, /resources, /research-notes alongside /about. Five previously-hanging pages now one click from the homepage. - Stale "39 chapters" → "41 chapters" across 11 user-facing sites: homepage hero stat ("39 chapters"), homepage section header ("Thirty-Nine Chapters" → "Forty-One Chapters"), Now banner copy, launch-tile description, /journey description meta + body, /sections description meta + body, /launch description meta + body, /github-playbook example card. Historical changelog entries kept original count (accurate at the time). Receipts: - New chapter: Ch 41 · 41-send-the-link · Part V · ~1,800 words - Chapter count: 39 → 41 (+2 this session: Ch 40 Prompting + Ch 41 HTML-ization) - Stale-stat sites corrected: 11 (across 4 pages) - Hanging pages surfaced on homepage: 5 (/swarms, /github-playbook, /showcase, /resources, /research-notes) https://dive.vladyslavpodoliako.com/changelog/#edition-9-1 Edition 9.1|2026-05-20 Wed, 20 May 2026 00:00:00 +0000 Swarms — the deep version, with the skills and the prompts and the receipts. Shipped: - New /swarms page (~3,800 words) — the operator's deep dive into multi-agent orchestration. Embeds the SwarmVisualizer widget (4 patterns animated: fan-out, pipeline, map-reduce, adversarial). The 5 × 4 wave pattern explained with the real five-wave breakdown from /swarm-strategic-plan. - The skill shelf — 10 swarm skills listed with shape + use case + file path: /swarm-strategic-plan, /debug-swarm, /audit, /agent-wave-verify, /folderly-external-audit, /cross-trio-audit, /bughunter, /ultrareview, /gstack-qa, /gstack-review - Seven swarm patterns from memory — CANON-lock + red-team / design-variant / perspectival editorial / multi-model bake-off / script-not-swarm collector / data-injection-slot / demand-test ≠ build. Each one earned its place by failing first; each gets one paragraph with the receipt. - Three orchestration prompts to steal — the universal subagent dispatch envelope (with the inline-fallback contract), the /audit 6-agent parallel dispatch shape, the BRIEF.md template with HARD CONSTRAINTS section that load-bears across 20 agents - Three things that quietly break a swarm — the 600s watchdog stall, filesystem contention with 5+ agents, load-bearing false premises in the BRIEF. Each one with the fix. - 3 ScreenshotPlaceholder slots — tmux running multi-agent dispatch, /swarm-strategic-plan output folder, /agent-wave-verify catching a stub. Placeholders mark intent; screenshots get captured live. - Ch 06 callout — added a tip box at the top of "The Swarm" chapter pointing to /swarms as the deep reference (chapter = story, page = playbook) - Ch 40 cross-link upgraded — the "swarms tease" section in Ch 40 now links to /swarms (was text-only when /swarms didn't exist yet) - Cmd-K — /swarms page entry + 10 section anchors searchable Receipts: - New page: /swarms · ~3,800 words · 10 sections · reuses SwarmVisualizer - Skills documented: 10 (with paths in ~/.claude/skills/) - Swarm patterns codified: 7 (each with a failure-receipt origin) - Orchestration prompts published: 3 (dispatch envelope, /audit shape, BRIEF template) https://dive.vladyslavpodoliako.com/changelog/#edition-9 Edition 9|2026-05-20 Wed, 20 May 2026 00:00:00 +0000 Prompting is basic now. The leverage moved up the ladder. Shipped: - New Chapter 40 — "Prompting, or the Knob You Probably Shouldn't Tune" (~3,200 words, 12-min read). The contrarian take: prompting is a basic skill now, the real ladder is skills → swarms → memory → data layer. Opens with the verbatim "rigor enforcer" prompt (works, but the wrong leverage) and contrasts it side-by-side with PROMPT_EOD from /resources (the operator-grade shape). Closes with a Monday move: pick the prompt you've tweaked 5+ times and promote it to a skill, schedule, or swarm. - New page /github-playbook — "GitHub for non-developers" (~3,400 words). GH-is-the-new-LinkedIn reframe, 5 non-code uses (private repo as delivery surface, README as living doc, Issues as ops queue, Discussions as forum, Releases as launch posts), the 8 gh commands you actually need, what you can safely ignore (branches, rebasing, merge conflicts, 80% of git), six examples in the wild (Vlad's Playbook + AFC + Karpathy/nanoGPT + Simon Willison/TIL + Willison/blog source + GitHub profile README pattern), and the Saturday move. - Ch 40 placed in Part II (Working Memory) — sits next to the vault, skills, build-a-skill, and connectors-MCP chapters. Also added to SECTIONS "claude" topical group. - Cmd-K wired — Ch 40 chapter entry + synonyms (prompting, prompt engineering, chain of thought, ladder, rigor enforcer, etc.), /github-playbook page entry + 8 section anchors searchable - Ch 19 cross-link callout — added a tip box near the gh repo create snippet pointing non-developers to /github-playbook - Fact-checked: nanoGPT marked deprecated by Karpathy in Nov 2025 (still live; example notes the move to nanochat). Originally proposed Patrick Collison reading-list example swapped for Simon Willison blog source (Collison's reading list is NOT on GitHub — fact-verified). Anthropic prompting docs URL host updated (docs.anthropic.com → platform.claude.com). Receipts: - New chapter: Ch 40 · 40-prompting-knob · Part II · ~3,200 words - New page: /github-playbook · 6 examples · 8 gh commands · ~3,400 words - Slop catches by research agent: 3 (Collison/nanoGPT/gh-pages-flag — all corrected before push) - Cross-links wired: Ch 19 → /github-playbook, Ch 40 → Ch 04/05/06/12/37, /github-playbook → /html-first + /sovereign-stack https://dive.vladyslavpodoliako.com/changelog/#edition-8-2 Edition 8.2|2026-05-20 Wed, 20 May 2026 00:00:00 +0000 HTML-ization, expanded — the recipe, the gallery, the receipts, the edges. Shipped: - /html-first expanded from 4 sections to 9: kept the existing case (AFC + Folderly audit + Do-this-Monday) and added five new ones - New section "More across the portfolio" — the order of operations (pitches first, audits next, internal deliverables third, communication fourth), the one-to-two gap that flips the format from curiosity to default, the Playbook itself as the maximal version of the thesis - New section "The recipe — do this Saturday" — exact verbatim prompt template (12-line constraint block), the four-step ladder (shape → prompt → three iterations → deploy), and the gh-create-repo deploy snippet - New section "Applications gallery" — 12 deliverable shapes that work (audit, pitch, QBR, board, mentee tracker, weekly KPI, financial model, brief, strategy memo, newsletter, conference companion, retro) rendered as a 3-col grid - New section "Not for everything — where this fails" — four honest edges where htmlization should not be used (legal documents, e-signatures, link-averse readers, deploy overhead vs reward) - New section "The pattern in the wild" — 12 public examples sourced (Ciechanowski Moon, Stripe Annual Letter, AI 2027, Anthropic attribution graphs, NYT needle, Ramp benchmarks, a16z market map, Pudding, Bloomberg year-in-graphics, Linear changelog, OWID, Nicky Case) rendered as a clickable 3-col grid with deep links, plus four emerging-pattern paragraphs - SectionNav updated 4 items → 9 items; Cmd-K HTMLFIRST_SECTIONS updated to match for discoverability Receipts: - Page sections: 4 → 9 (+5) - New gallery cards (operator applications): 12 shapes - New "in the wild" citations: 12 sourced public examples, 2024–2026 - New page weight: +5,500 words approx · still single page, no nav split https://dive.vladyslavpodoliako.com/changelog/#edition-8-1 Edition 8.1|2026-05-20 Wed, 20 May 2026 00:00:00 +0000 The launch as its own experiment in the thesis. Shipped: - New /launch-week page — the launch is its own forwardable artifact. Numbers panel (live GitHub stars, newsletter open rate, subs delta, page-views), a day-by-day timeline of every distribution surface, a "what I'd do differently" section that populates from real data not Day-0 guesses, a meta-callout explaining why a PDF version would be obsolete the second it was exported - src/lib/launch-week.ts — single source of truth for the timeline + numbers + retrospective; append-only daily edits (one-line additions, no schema breaking) - Day-0 distribution drafts authored locally (notes/launch/, gitignored): 8-tweet X thread (each tweet stands alone), LinkedIn longform (founder-realization frame), Hacker News Show HN (title + first-comment), Day-7 Numbers newsletter template with TBD slots - Homepage tile grid restored to 4-up across multiple rows — 8 tiles: Sovereign Stack · HTML-ization · The launch · Launch week · Vault library · Cheat sheet · Glossary · Tier list. About moved to the small text-link below. - Cmd-K indexes /launch-week as a discoverable destination Receipts: - New page: /launch-week - Distribution surfaces drafted: 4 (Newsletter · X · LinkedIn · HN) - Homepage tiles: 4 → 8 (2 rows of 4) - Day-7 follow-up scheduled: Numbers post — receipts of launch week, sent regardless of how it goes https://dive.vladyslavpodoliako.com/changelog/#edition-8 Edition 8|2026-05-20 Wed, 20 May 2026 00:00:00 +0000 Two stacks, not one. The sovereign tier is built. Shipped: - New /sovereign-stack page — the open-weights answer to the deprecation calendar. The Onyx Open-Source LLM Leaderboard rebuilt as a tier list (Overall / Coding / Math / Chat / Reasoning), Ollama vs LM Studio vs vLLM, five hardware tiers ($2k MacBook → $40k+ deploy) with what-breaks-first per tier, quantization + KV-cache math, open-vs-closed cost-per-task table (30× ratio on repeated evals), the heretic question (abliteration framed as sovereignty, not jailbreaking), the Mythos lesson, Karpathy nano-gpt as a Saturday curriculum, the 6-month watch list - New OnyxLeaderboard.tsx widget — interactive S/A/B/C/D tier list with category tabs, size filters (Small/Medium/Large by parameter count), and per-lab filters showing the Chinese-lab dominance of the open-weights frontier (6/6 S-tier slots) - New glossary terms — Sovereign stack · Heretic model · Mythos · Abliteration — all cross-linked - Ch 24 (tier-list) callout pointing to /sovereign-stack — the closed-model tier list and the open-weights tier list now talk to each other - Homepage tile grid rebalanced — Sovereign Stack takes flagship-first position (4 across), Glossary + tier-list + launch + about demoted to a single text row of secondary links below the grid Receipts: - New page: /sovereign-stack - Leaderboard snapshot: Onyx Open-Source LLM Leaderboard · 2026-03-24 · 19 models across 5 categories - Sourced citations: 60+ inline links — Arditi 2024, modded-nanogpt, Onyx, Ollama MLX, all model releases - Open-weights S-tier (May 2026): GLM-5 · Kimi K2.5 · MiniMax M2.5 · DeepSeek V3.2 · Step-3.5-Flash · Qwen 3.5 https://dive.vladyslavpodoliako.com/changelog/#edition-7 Edition 7|2026-05-20 Wed, 20 May 2026 00:00:00 +0000 The book is open. The source is the recipe. Shipped: - Repo flipped to public after a full git-filter-repo history scrub — 71 commits rewritten, 0 leaks across blobs and commit messages from an independent fresh-clone verification, backup ref preserved server-side - New /launch page — the launch artifact IS the demo: typewriter hero, animated stat odometer (chapters/widgets/glossary/embeds/editions/stars), the 3 live embedded case studies, all 39 chapters as a click-anywhere mosaic, what-it-took timeline, post-credit AFC tease - Homepage "Now" banner above the tile grid promoting the launch through launch week - README rewrite to current reality (Edition 6/7, real counts, dive.vladyslavpodoliako.com as primary CTA), repo metadata (description, homepage, 12 topics, Discussions on), custom 1280x640 GitHub social-preview image - SECURITY.md + CONTRIBUTING.md refreshed for the public posture; private-vulnerability-reporting path documented - Ch 02 contradiction fixed (the Fathom-in-the-don't-use-bin error caught externally) — reframed as "picked Fathom, killed the rest" with the discipline of one transcript surface Receipts: - Repo state: PUBLIC · github.com/Belkins/ai-dive-deep - History leaks (post-scrub, fresh-clone verified): 0 / 71 commits - New page: /launch - Edition 1 → public: 13 days, 7 editions https://dive.vladyslavpodoliako.com/changelog/#edition-6 Edition 6|2026-05-19 Tue, 19 May 2026 00:00:00 +0000 The method, embedded. Click the artifacts. Shipped: - New /html-first page — the htmlization thesis (every deliverable ships as a live interactive artifact, not a dead file) with two REAL artifacts embedded and clickable - AFC case: the dinner-table idea whose investment deck + robot-stable annex got spun up as interactive HTML before the next meeting — Vlad's own venture, embedded as-is - Folderly case: a real ~90-domain / ~5K-mailbox external deliverability audit, swarm-produced on the Folderly methodology — fully de-identified (client, all 90 domains, SPF/infra fingerprints, WHOIS name) and embedded as a sanitized sample - New ArtifactEmbed.astro — sandboxed, lazy, render-on-click iframe modal (no src until opened; referrerpolicy=no-referrer) - Ch 19 callout + glossary "Htmlization" + Cmd-K (page + 4 section anchors) wired Receipts: - Live artifacts embedded: 3 (AFC ×2, audit ×1) - Folderly identifiers redacted: 90 domains + name + 2 SPF + WHOIS → 0 residual - New page / component: /html-first + ArtifactEmbed - External calls in embeds: 0 (fully self-contained) https://dive.vladyslavpodoliako.com/changelog/#edition-5 Edition 5|2026-05-19 Tue, 19 May 2026 00:00:00 +0000 The bench moved. The reports stopped being files. Shipped: - Gemini 3.5 Flash + Claude-for-the-legal-industry logged as dated research notes — signal-vs-receipt discipline, not a leaderboard edit - Ch 35: corrected the stale "Gemini 3 Pro" reference + added the Flash-beats-last-gen-Pro signal, scoped as a Ch 29 cost question rather than a re-tiering - Ch 29: new section — "The price of a model is not the price of a task" + a cost-per-task test loop you run on your own traffic - Ch 24: May-2026 addendum extended; the live LMArena widget stays the source of truth — no slide-driven tier edits - New thesis across Ch 19 / Ch 26 / About — every report ships as a living link on a private repo, not a dead file; unsanctioned copying as the truest adoption metric Receipts: - Research notes added: 2 - Chapters touched: 19, 24, 26, 29, 35 - Live bench edits: 0 (signal, not receipt) - New components: 0 https://dive.vladyslavpodoliako.com/changelog/#edition-4 Edition 4|2026-05-11 Mon, 11 May 2026 00:00:00 +0000 Retitled for navigation. Voice survived. Shipped: - All 36 chapter titles rewritten — topical-primary, with question form on 9 chapters where intent shape beats subject shape - Old hype titles preserved as subtitles ("The Day I Killed My Tabs" now sits below "AI as an Operating System") - Cmd-K resolves "cron" → Scheduled Tasks, "browser" → Browser Agents with Playwright, "permissions" → When to Skip Permissions - scripts/apply-titles.py — atomic re-runnable rewrite of all 36 frontmatters + chapters.ts. Re-run any time the map changes. Receipts: - Chapters retitled: 36 - Time to scan TOC: ~6s → ~2s - Voice lines lost: 0 https://dive.vladyslavpodoliako.com/changelog/#edition-3-5 Edition 3.5|2026-05-09 Sat, 09 May 2026 00:00:00 +0000 Repo flipped private. Site stayed public. Shipped: - Repo visibility flipped to PRIVATE (Belkins on GH Pro tier) - Issues, Wiki, Projects, Discussions disabled — attack surface reduced - SECURITY.md committed with vulnerability reporting + ops rules - .gitignore hardened — .pem, .key, .aws/, .kube/, secrets/, .vercel - Vercel fallback config committed — dual-target astro.config.mjs reads DEPLOY_TARGET env - Vercel security headers baked (HSTS, X-Frame-Options DENY, X-Content-Type-Options, Permissions-Policy) Receipts: - Repo visibility: PUBLIC → PRIVATE - Site downtime during flip: ~3 min (Pages re-enable required) - Surfaces disabled: 4 https://dive.vladyslavpodoliako.com/changelog/#edition-3 Edition 3|2026-05-07 Thu, 07 May 2026 00:00:00 +0000 Closed the original brief. Six new chapters, two new pages. Shipped: - Ch 31 — The Stages: Ideation → Foundation → Creation → Polishing → Security → Deploy - Ch 32 — How to Build Rick (OpenClaw, NemoClaw, Hermes archetypes) - Ch 33 — Browser Agents with Playwright (login, click, scrape, post) - Ch 34 — Persona Agents and the Four NEVERs - Ch 35 — Codex × Claude Code (day shift, night shift) - Ch 36 — Beyond Claude Code (CrewAI, LangGraph, SDK) - StagesFlow widget (Ch 31) — six clickable stages with artifact + ready test + failure mode - ArchetypePicker widget (Ch 32) — 3-question intake → recommended Rick archetype + install command - /day-zero — literal first 30 minutes with 12 localStorage-persisted checkboxes - /sections — chapters grouped by General / Claude / Security / AI Agents / Building / Team+Tier Receipts: - Chapters added: +6 (30 → 36) - Widgets added: +2 (12 → 14) - New pages: +2 (/day-zero, /sections) https://dive.vladyslavpodoliako.com/changelog/#storytelling-layer Storytelling layer|2026-05-08 Fri, 08 May 2026 00:00:00 +0000 36 islands became one six-act journey. Shipped: - /how-to-read — ~1,100-word prologue. Who this is for, the reframe, the journey, three ways to use the book. - /journey — 6-part narrative arc (Reframe → Memory → Workshop → Discipline → Building → Frontier) - /questions — 18 questions Vlad answers most weeks, each with short answer + chapter pointers - Part pill on every chapter hero — readers know where they are in the arc - Chapter footer transitions detect part boundaries: "Next: Part III — The Workshop →" - /showcase — auto-extracted from ~/.claude/: 62 skills + 32 custom agents + 12 plugins, categorized + searchable - /cowork-setup — sanitized: 12 connector categories + 8 scheduled-task patterns + day-shape timeline. Zero local-data scan. Receipts: - Reader-facing pages added: +5 - Skills surfaced: 62 - Sensitive paths read: 0 (Cowork showcase derived from published chapters) https://dive.vladyslavpodoliako.com/changelog/#edition-2 Edition 2|2026-05-07 Thu, 07 May 2026 00:00:00 +0000 Fixed the muscles. Wrote the spine. Shipped: - 6 new chapters (Ch 25-30): Evals, Team Adoption, Voice Agents, Failure Receipts, Cost Economics, Anthropic SDK Direct - 4 new widgets: TokenBurnCalculator (Ch 2/29), TempAgencyLoop (Ch 3), VaultGraphPreview (Ch 4), HookEventTimeline (Ch 16) - 12 a11y + UX patches: glossary popovers via Radix (no more navigate-away mid-read), focus rings, skip-to-content, paper contrast fix, 17px body on ≥640px, anchor links on H2/H3, print stylesheet - 10 new operator prompts added to /resources: deal post-mortem, hire screen, model migration, board update, mentee prep, RFP triage, kill decision, customer-call synthesis, writing-filter, Tuesday-9am triage - ResumeReading pill on landing (localStorage-driven) - Substack subscribe iframe in footer - Sharper "Most readers go here next" copy in chapter footer Receipts: - Chapters: 24 → 30 - Widgets: 8 → 12 - Pages: 32 → 38 https://dive.vladyslavpodoliako.com/changelog/#edition-1 Edition 1|2026-05-07 Thu, 07 May 2026 00:00:00 +0000 The book became an artifact. Shipped: - 24 chapters migrated from .docx into MDX (4 parallel agents, partitioned by chapter range) - 8 interactive widgets: StackSelector, SwarmVisualizer, CronBuilder, ModePicker, SkillComposer, ConnectorMap, PermissionSimulator, TierListBuilder - Cmd-K command palette + dark theme + view transitions + reading progress bar - Resources page with copy-paste vault library: CLAUDE.md skeleton, .mcp.json examples, hook scripts, SKILL templates, 5 reusable prompts - Printable cheat sheet (Ch 14, @media print styled) - 30-day plan generator (3 intake questions → custom 30-day calendar, .ics + markdown export) - Drag-and-drop tier list builder with shareable URL hash - GitHub Pages deployment via Actions; live in ~50s per push Receipts: - Source words migrated: ~43,000 - Parallel agents in chapter migration: 4 - Wall-clock for full Edition 1 build: ~3 hours - Dist size: 1.4 MB / 32 pages