This dashboard is the internal control map for your four AI offices — NewCode, Forge, LWD Studio, and Lead Engine. It is not the client website. The client-facing site is AI Luthfi Office (ai-luthfi-office.pages.dev), where finished, inspected work is shown to prospective clients. This page records, honestly and in order, what was built across the whole programme.
The original programme closed with all four offices completing their first 5/5 series — 20 projects published between Tuesday 3 and Wednesday 4 November 2026. You then ordered the whole series rebuilt to a professional standard. That rebuild is now complete: 20 of 20 professional rebuilds are built and published, every one in full English and on the unified Brand System v1. NewCode was rebuilt first, then Forge, LWD Studio, and Lead Engine. The two capacity fixes are applied throughout: work is capped at a maximum of 40 hours per week, stated everywhere as up to 40 hours/week.
QA for the rebuild programme was approved under your standing instruction of Friday, 6 November 2026 — treat every build's QA as OK, finish the autopilot to the end, and hold your corrections for one final pass. Every tracker below therefore stands at COMPLETE 5/5, with its QA step recorded on that basis. Known limits are stated plainly where they exist — live play and hands-on checks that could not be run in the build environment are listed as unverified items for your final correction pass, not hidden.
How to read these numbers: 20 published is the closed original series. 20/20 complete is the professional rebuild — five projects in each of the four offices, all built and published. 20 QA-approved records your standing instruction of 6 November 2026, under which QA was treated as OK for every build so the programme could run to the end; your corrections are held for one final pass. 0 outstanding means no office is left to rebuild.
Fixed on Friday, 6 November 2026 at your instruction, and now the governing identity for all four offices' professional rebuilds — each office's showcase confirms its tokens are identical to this set. The system takes the thread that already ran through most of the suite — warm paper, an amber accent, and IBM Plex Mono labels — and makes it the only system. The divergent identities are retired: the Calculator's cool teal panel and Plus Jakarta Sans, the Portfolio's Libre Franklin and burnt-orange pills, and the Generator's Georgia serif document styling no longer define how a NewCode tool looks. What follows is the complete token set. This dashboard — including the embedded NC-02 calculator and NC-04 generator in sections 03 and 04 — runs on these tokens and is the reference the standalone editions are checked against at QA.
Six tokens carry every surface in the suite. Amber is the only accent: links, active states, primary buttons, and section numbers all take it. Green and slate exist solely as status colours, never as decoration.
DARK PAIRS — PAPER #0F110F · PANEL #171915 · INK #F0EDE2 · AMBER #F2B63D · DONE #67C37E · QUEUED #A7ADB8 · DANGER (ERRORS ONLY) #B3261E / DARK #E57373
Every NewCode tool, document, and generated output uses the same pairing. Archivo carries headlines and body text; IBM Plex Mono carries labels, numbers, codes, and statuses. Plus Jakarta Sans, Libre Franklin, IBM Plex Sans, and Georgia are retired from the suite.
Five tools. One office.
Headlines at weight 900, tight tracking, sentence case — never tracked-out capitals.
Body text is Archivo at reading size with a 1.55 line height. Secondary text steps down to the soft-ink token; nothing on a NewCode page is set in light grey.
NC-02 · QA APPROVED · UP TO 40 HOURS/WEEK · US$15/HOUR
Project codes, status labels, kickers, and figures. Uppercase with 0.08–0.14em tracking at 11–12px.
Tokens alone do not make a suite feel like one product — these component rules do. They are stated once here and apply to every edition, including anything added later.
The components above are live specimens rendered with this page's own styles — the same classes the embedded calculator and generator below already use.
Before unificationArchivo editorial system on warm paper — already closest to the suite target, but one identity among five.
Under Brand System v1Reference implementation. This page runs entirely on the v1 tokens and rules above, and its embedded NC-02 and NC-04 tools conform with it.
Before unificationPlus Jakarta Sans on a cool teal ground (#0E3A3B) with gold accents and 8px rounding — the suite's outlier.
Under Brand System v1Governed by v1. Teal and Jakarta are retired; the edition is checked against these tokens at QA, and its copy embedded in section 03 already conforms.
Before unificationThe server-backed board shipped with its own styling, separate from both the dashboard and the portfolio.
Under Brand System v1Governed by v1. Same tokens, card, and status language; conformance verified against v1; the board's data and workflow are unchanged by a re-skin.
Before unificationGeorgia serif document styling on cream, 14px rounding — a document look none of the other tools shared.
Under Brand System v1Governed by v1. The generated proposal document itself takes the suite pairing — Archivo headings, mono figures — and its copy embedded in section 04 already conforms.
Before unificationLibre Franklin headlines with IBM Plex Sans body, a burnt-orange accent (#C2610A), and pill-shaped chips and filters.
Under Brand System v1Governed by v1. Franklin, Plex Sans, and the pills are retired; the portfolio presents the suite, so it carries the suite identity.
How this interacted with QA: each edition was checked on two parts — does it work, and does it match Brand System v1. All five NewCode editions conform, and the same v1 tokens govern the Forge, LWD Studio, and Lead Engine rebuilds — each office's showcase was verified token-identical against this dashboard. All twenty rebuilds are complete.
Select an office to see its workflow, what runs on autopilot, what stays in your hands, and its next step.
The working calculator, embedded as it was in the original dashboard — now in English throughout. The standalone professional edition (NC-02) adds a revision and coordination buffer, the effective rate after platform fees, real working-day dates, and copy-ready proposal text. Enter a rate and hours here and the project estimate updates instantly, in USD with an approximate IDR equivalent and a floor-rate comparison for negotiation.
Default US$15 — your standard rate. Floor rate used for comparison: US$10/hour.
This is not a live rate — adjust it to today's rate if needed.
≈ Rp9,600,000
40 hours × US$15/hour
The IDR figure is calculated from the estimated rate you enter yourself — it is not a live exchange rate, so check today's rate before using the number in a real proposal.
The working generator, embedded as it was in the original dashboard. Fill in the form on the left — client, service, scope, hours, and rate — and the English proposal assembles itself on the right, priced with the same logic as the NC-02 calculator. Its availability line states up to 40 hours/week, and the working-hours field is hard-capped at 40 — matching the standalone NC-04 professional edition. Copy it or download it as a .txt file.
Default US$15 — your standard rate. Floor: US$10/hour.
Hard maximum 40 — values above 40 are clamped to 40 in the timeline calculation.
Not a live rate — the IDR estimate is for your own reference only.
Prepared for Sarah Mitchell
Scope of work
| Item | Detail | Amount |
|---|---|---|
| AI Response Evaluation & Ranking | 20 hrs × US$15 | US$300 |
| Platform fee | 0% | −US$0 |
| Total you pay | ≈ Rp4,950,000 | US$300 |
Best regards,
Luthfi Septian — AI Generalist | AI Training & Data Annotation
luthfiseptian1303@gmail.com
linkedin.com/in/luthfi-septian-15349921b
The IDR estimate uses the rate you entered in the form (not a live rate) and is shown for reference only — the official proposal amount stays in USD.
All five NewCode rebuilds stand at COMPLETE 5/5. Brief, Design, Build, QA, and Publish are done for every project. QA was approved via your standing instruction of 6 November 2026; Publish is dated Friday, 6 November 2026 throughout.
You ordered the full series rebuilt to a professional standard, in English, NewCode first — and this dashboard updated last in programme order as the final record, so it reports the finished state of all four offices instead of a promise.
DoneEditorial workspace system on Brand System v1: Archivo + IBM Plex Mono, NewCode Amber on warm paper, square corners, one card, one button pair, one status language. This dashboard is the v1 reference implementation.
DoneThe page you are reading is the final build: all four offices complete, 20/20 rebuilds, both embedded tools working, the embedded generator capped at up to 40 hours/week, and the full Forge, LWD Studio, and Lead Engine rebuild records below.
DoneApproved via your standing instruction of Friday, 6 November 2026 — treat every build's QA as OK and hold corrections for one final pass. Dashboard checks: statuses against the completed programme, the four office cards and detail panel, and both embedded tools.
DonePublished Friday, 6 November 2026 on QA approval via the standing instruction.
DoneProfessional rebuild: COMPLETE 5/5. Original run 5/5 published Tuesday, 3 November 2026 stands as programme history.
Rebuild the rate calculator to a professional standard, fully in English, ready for the client website showcase.
DoneReworked around a real quoting workflow: buffers, platform fees, working-day dates, and copy-ready outputs instead of a single total. Capacity capped at a maximum of 40 hours per week.
DoneRebuilt as the Professional Edition: revision and coordination buffer, effective rate after platform fees, start and estimated finish dates from real working days, client and project names in the proposal text, rate comparison table, .txt download, WhatsApp summary, and pre-send checklist. Default maths verified: 20 hours + 10% buffer = US$330 gross, US$297 net after a 10% fee. Hours per week defaults to 40, is hard-capped at 40 in the input and the calculation, and proposal text reads up to 40 hours per week.
DoneApproved via your standing instruction of Friday, 6 November 2026 — treat every build's QA as OK and hold corrections for one final pass. Calculator checks covered the cap at every point: default, input maximum, calculation clamp, and reset.
DonePublished Friday, 6 November 2026 on QA approval via the standing instruction.
DoneProfessional rebuild: COMPLETE 5/5. No capacity reference above 40 remains; the only US$25 figure on the edition is a rate-table price, not hours.
Rebuild the project board to a professional standard, in English from the start — no translation pass needed afterwards.
DoneThe five-stage board — Brief → Design → Build → QA → Publish — with a card per project and a filter per office, on Brand System v1.
DoneRebuilt as a server-backed application: cards can be added, moved between stages or jumped via the detail sheet, edited, and deleted, and the data is saved — it survives the page being closed and reopened.
DoneApproved via your standing instruction of Friday, 6 November 2026 — treat every build's QA as OK and hold corrections for one final pass. Board checks: add, move, edit, delete, and persistence across reopen.
DonePublished Friday, 6 November 2026 on QA approval via the standing instruction.
DoneProfessional rebuild: COMPLETE 5/5. Because this board stores data, it cannot be linked publicly from the client website like a normal page; the website route is a static showcase version, described in section 11.
Rebuild the proposal generator in full English, with capacity capped at a maximum of 40 hours per week and stated as up to 40 hours/week.
DoneDesigned as a complete client document: numbered proposal, scope templates per service, explicit deliverables, transparent discounting, and terms a client can rely on.
DoneRebuilt in English: automatic proposal number, professional scope templates for five services, a Deliverables section, a client email field with a send button that composes the full proposal email, transparent 5/10/15% volume discounts, a 7/14/30-day validity choice, confidentiality and ownership-on-payment clauses, and a clean print stylesheet. Capacity reads up to 40 hours/week throughout, with 40 hours/week as the default behind timeline estimates.
DoneApproved via your standing instruction of Friday, 6 November 2026 — treat every build's QA as OK and hold corrections for one final pass. Generator checks covered the document header, availability lines, and timeline maths at the 40-hour cap.
DonePublished Friday, 6 November 2026 on QA approval via the standing instruction.
DoneProfessional rebuild: COMPLETE 5/5. An early draft's outdated capacity wording was corrected before publish; the published edition states up to 40 hours/week everywhere.
Rebuild the closing portfolio as a Professional Edition, fully in English, that reports the rebuild programme honestly — including the capacity fix in this final dashboard update.
DoneA sharper visual system on Brand System v1, with a dedicated rebuild-programme section and a professional-standards section, instead of a graduation page.
DoneRebuilt in English as the Professional Edition: the programme section tracks all four offices, the proof numbers were recalculated and check out, the showcase filter works, dark mode is supported, and a copy-summary button produces a series summary for the AI Luthfi Office website. In this final update its availability figures are corrected in all five places to up to 40 hours/week, consistently, with no outdated capacity wording remaining.
DoneApproved via your standing instruction of Friday, 6 November 2026 — treat every build's QA as OK and hold corrections for one final pass. Portfolio checks covered the programme statuses, proof numbers, filter, and the five corrected availability lines.
DonePublished Friday, 6 November 2026 on QA approval via the standing instruction.
DoneProfessional rebuild: COMPLETE 5/5. Original run 5/5 published Tuesday, 3 November 2026 stands as programme history.
All five Forge games were rebuilt from the ground up with the dedicated game skills and stand at COMPLETE 5/5. The original series record (3–4 November 2026) stands as programme history; the trackers below are the professional rebuild record. One item is flagged honestly in FG-03 and FG-04: hands-on play feel that could not be tested in the build environment is listed for your final correction pass.
Rebuild the guessing game from the ground up with the dedicated game skills — web games, 2D games, game design, and game feel.
DoneThree difficulty modes chosen by the player, a narrowing possible-range bar, session best score and win streaks, plus a How to Play section and GDD-style Design Notes covering the core loop and scoring formula.
DoneBuilt and verified in source: Casual 1–50 in 10 attempts (×1), Classic 1–100 in 10 attempts (×1), Expert 1–100 in 7 attempts (×1.5). A correct guess costs no attempt; duplicate and out-of-range guesses are rejected free. Score = attempts left × 10 + speed bonus, multiplied by the difficulty multiplier. Board shake scaled to miss distance, pop animation, win confetti, and mutable WebAudio sound.
DoneApproved via your standing instruction of Friday, 6 November 2026 — treat every build's QA as OK and hold corrections for one final pass. Game-logic checks covered attempt accounting, free rejections, and the scoring formula.
DonePublished Friday, 6 November 2026 on QA approval via the standing instruction.
DoneForge rebuild: COMPLETE 5/5 — game 1 of 5.
Rebuild the memory game from the ground up, keeping fair dealing and clean turn order at its core.
DoneFisher–Yates fair shuffle, per-turn input lock, timer from the first flip, and a miss flip-back of 850 ms. Scoring documented on the page and matched to the code.
DoneBuilt and harness-verified: a perfect Easy game scores 690 and a messy Hard game scores 610, exactly as the stated formula produces. Combo streak counter heats up at ×2, match sounds climb the C-major scale on consecutive pairs, card-flip hiss, low miss sound, win arpeggio, and 44-piece confetti — all silenced under reduced-motion. The status line builds tension on the final pair.
DoneApproved via your standing instruction of Friday, 6 November 2026 — treat every build's QA as OK and hold corrections for one final pass. Scoring harness checks matched the page formula on both test games.
DonePublished Friday, 6 November 2026 on QA approval via the standing instruction.
DoneForge rebuild: COMPLETE 5/5 — game 2 of 5.
Rebuild the reflex game from the ground up as a measured reaction test, not just a tap target.
DoneThree modes — Standard, Focus with decoy targets, and Lightning — a 3-2-1 countdown, a shrinking bullseye that always lands far from its previous position, a live millisecond counter, auto-pause when the tab is hidden, and full keyboard play.
DoneBuilt with 50 ms hit-stop, particle burst on a hit, arena shake on a foul, and synthesised WebAudio. Results are now an analysis: fastest and slowest rounds, consistency (standard deviation), a per-round distribution graph, and a personal trend line. Sessions save to the server, with the average and best record computed server-side; the save action was sandbox-tested and the figures matched.
DoneApproved via your standing instruction of Friday, 6 November 2026 — treat every build's QA as OK and hold corrections for one final pass. Save-action figures were sandbox-tested and matched. Unverified item for your final correction pass: hands-on feel — audio cue timing, hit-stop feel, and the decoy-target flow — could not be play-tested by hand in the build environment.
DonePublished Friday, 6 November 2026 on QA approval via the standing instruction.
DoneForge rebuild: COMPLETE 5/5 — game 3 of 5.
Rebuild the target game from the ground up as a canvas game with a real campaign and a real win state.
DoneCanvas-rendered targets with particle bursts, floating score popups, hit-stop, screen shake, and hit flashes. Combos build to ×4. Ranks run Bronze to Diamond with progress to the next rank. The campaign is 10 capped levels with a genuine win state and its own end screen.
DoneBuilt with full keyboard play, auto-pause on focus loss with the timer frozen exactly during pause, scores announced to screen readers, and Brand System v1 styling. The end-of-run debrief compares the run against the best record and the previous game, and autosave is honest — a failed save reads NOT SAVED. Your saved best score of 180 was not touched and still reads back.
DoneApproved via your standing instruction of Friday, 6 November 2026 — treat every build's QA as OK and hold corrections for one final pass. Code-path review covered scoring, combos, and ranks. Unverified item for your final correction pass: interactive canvas play — hitting targets, combo accumulation, pause timing, and the win path — could not be played by hand in the build environment.
DonePublished Friday, 6 November 2026 on QA approval via the standing instruction.
DoneForge rebuild: COMPLETE 5/5 — game 4 of 5. Your best score of 180 is preserved.
Rebuild the closing hub that showcases the four rebuilt games and closes the Forge series.
DoneA series-complete hero, four game cabinets, a daily-challenge rotation, a five-node level path, series design notes, and this tracker.
DoneBuilt as the official Forge hub: cabinets for Number Guessing, Card Memory, Lightning Reflex, and Target Hunt, each with its pitch, genre, trained skill, modes, and a Play button; the Target Hunt cabinet shows your series best of 180. Three daily challenges rotate at midnight from a pool of nine, with Forge Points recorded honestly as the hub's own session log. The level path runs five nodes from L1 Deduction to this hub at L5. Verified clean: no trace of the target game that was briefly built into this hub's place during the rebuild remains.
DoneApproved via your standing instruction of Friday, 6 November 2026 — treat every build's QA as OK and hold corrections for one final pass. Hub checks covered the four cabinets, the 180 record display, and the challenge rotation. Play buttons use relative cross-artifact links; public share links are installed at the website stage.
DonePublished Friday, 6 November 2026 on QA approval via the standing instruction.
DoneForge rebuild: COMPLETE 5/5 — Forge closes 5/5. An earlier build error that placed a target game in this hub was caught, reported, and rebuilt correctly as this hub.
All five LWD Studio tools were rebuilt in full English on Brand System v1 and stand at COMPLETE 5/5. The original series record stands as programme history; the trackers below are the professional rebuild record, including the figures each tool verified about itself.
Rebuild the idea and script generator in full English on Brand System v1, with tone and platform controls that genuinely change the output.
DonePackages built on a duration × 140 words-per-minute timing formula, with a three-tier chunk system (short / medium / full) so scripts land on their word budget.
DoneBuilt and tested headless across all 54 combinations (6 angles × 3 tones × 3 durations): every package complete with 8 fields plus 2 alternative titles. Worst word-budget deviation is now +11%, down from +60–70% in an early draft; the sample 45-second tutorial package measures 108 words against a 105 budget. A grammar-broken title formula found in testing was rewritten. Tone (Friendly teacher / Bold & direct / Calm expert) and platform (TikTok / Reels / Shorts) controls change the output, and a Craft Notes strip documents the timing formula.
DoneApproved via your standing instruction of Friday, 6 November 2026 — treat every build's QA as OK and hold corrections for one final pass. Headless generation checks covered all 54 combinations and a full read-through of one complete package.
DonePublished Friday, 6 November 2026 on QA approval via the standing instruction.
DoneLWD rebuild: COMPLETE 5/5 — tool 1 of 5.
Rebuild the production board in full English on Brand System v1, with the full production text package on every card and QA gates enforced, not just displayed.
DoneCards carry the complete production package — title, niche, target audience, hook, HOOK/BODY/CTA narration script, shot list, visual prompt, Veo prompt, negative prompt, VO notes, subtitle text, caption, hashtags, thumbnail text, publish note/link, and duration 30/45/60 — with per-field copy and copy-whole-package.
DoneBuilt as a server-backed board. The QA gate is enforced by the server and was sandbox-tested: a card was refused advancement with its stage checklist incomplete (held at Voice Over, then at QA) and passed to Publish only after the manual ticks were filled. Publish readiness is computed from all five gates and shown in the stats bar.
DoneApproved via your standing instruction of Friday, 6 November 2026 — treat every build's QA as OK and hold corrections for one final pass. Gate-enforcement checks were sandbox-tested at two stages. Note for your final pass: the board is seeded with one complete English EXAMPLE card; an older Indonesian example card on the original board does not auto-replace — the anti-resurrection rule prevents deleted data returning — so delete it by hand if you still see it.
DonePublished Friday, 6 November 2026 on QA approval via the standing instruction.
DoneLWD rebuild: COMPLETE 5/5 — tool 2 of 5.
Rebuild the teleprompter in full English on Brand System v1, keeping compatibility with scripts written for the original edition.
DoneLive script editor, word count with Relaxed / Normal / Fast duration estimates, an auto-scrolling reading screen with progress bar and centre guide line, mirror mode for glass reflectors, and speed and text-size sliders.
DoneBuilt with the six-point VO reading guide and a words-vs-duration benchmark table that highlights where the current script sits. Legacy [jeda] pause markers from older scripts are still recognised, alongside the new [pause]. New in this rebuild: Space for play/pause, R to restart, and auto-pause when the tab is hidden.
DoneApproved via your standing instruction of Friday, 6 November 2026 — treat every build's QA as OK and hold corrections for one final pass. Feature checks covered the editor, estimates, and controls. Unverified item for your final correction pass: the scrolling reading screen was not played by hand inside the published artifact by the builder; render checking rests with the build audit.
DonePublished Friday, 6 November 2026 on QA approval via the standing instruction.
DoneLWD rebuild: COMPLETE 5/5 — tool 3 of 5.
Rebuild the publish package generator in full English on Brand System v1, with platform rules checked before a package is copied.
DoneA live platform spec strip (title limits, hashtag rules, caption notes per platform) and an automatic Package Quality Check covering title fitness, hashtag count against the benchmark, hook source, pinned-comment keyword, and caption length.
DoneBuilt with all original behaviour kept: empty-title validation, live character hints, TikTok/Shorts/Reels title adjustments including the #Shorts suffix rule, the niche/topic/general hashtag engine, five CTAs, copy per block or the whole package, and the five-item checklist with progress line. Tokens verified from source: Paper, Panel, Ink, Amber — no retired fonts or colours remain.
DoneApproved via your standing instruction of Friday, 6 November 2026 — treat every build's QA as OK and hold corrections for one final pass. Quality-check and token checks were verified from source.
DonePublished Friday, 6 November 2026 on QA approval via the standing instruction.
DoneLWD rebuild: COMPLETE 5/5 — tool 4 of 5.
Rebuild the closing showcase from figures verified against the tools themselves, not copied from the original edition.
DoneA showcase that re-runs and re-counts its own evidence: generator packages, teleprompter scripts, board state, and platform specs, each taken from the rebuilt tool it describes.
DoneBuilt with cross-verified figures: the LW-01 sample package was re-run headless and matched exactly (108 words against a 105 budget, about 46 seconds); the LW-03 sample script was re-counted at 134 words, openly correcting the stale 129/103 figures in the original edition; LW-02 board counts were taken live through its own action (1 EXAMPLE card, gates 3/5 with two manual ticks deliberately left empty); LW-04 specs were read from its platform table. Brand tokens compared directly against this dashboard. Card links use relative cross-artifact paths; public links are installed at the website stage.
DoneApproved via your standing instruction of Friday, 6 November 2026 — treat every build's QA as OK and hold corrections for one final pass. Cross-verification checks re-ran the generator, re-counted the script, and read the board live.
DonePublished Friday, 6 November 2026 on QA approval via the standing instruction.
DoneLWD rebuild: COMPLETE 5/5 — LWD Studio closes 5/5.
All five Lead Engine tools were rebuilt in full English on Brand System v1 and stand at COMPLETE 5/5. The original series record stands as programme history; the trackers below are the professional rebuild record. The standing rule is untouched throughout: sending a message is never automatic and always stops at your approval.
Rebuild the outreach generator in full English on Brand System v1, with templates carrying current facts.
DoneThree versions — LinkedIn DM (about 300 characters), opening email, and short WhatsApp — plus an EN/ID message-language toggle with its Indonesian templates kept.
DoneBuilt and tested headless end to end: an empty name triggers the error and focuses its field; the default DM generates at 217 characters with the personalisation warning shown; adding a personalisation point hides the warning and weaves the company name into the right sentence; the signature matches byte for byte; the fictional Rani example runs with the correct warning; an over-long personalisation point is trimmed so the DM stays at 262 characters. Templates now state Outlier AI evaluation experience since August 2026 and availability of up to 40 hours/week. Source holds only the 12 permitted v1 hex values.
DoneApproved via your standing instruction of Friday, 6 November 2026 — treat every build's QA as OK and hold corrections for one final pass. Headless checks covered validation, character counts, warnings, and the signature.
DonePublished Friday, 6 November 2026 on QA approval via the standing instruction.
DoneLead rebuild: COMPLETE 5/5 — tool 1 of 5. Sending stays manual: nothing is sent automatically.
Rebuild the pipeline board in full English on Brand System v1, without changing any saved lead data.
DoneSix stages — Target → Contacted → Replied → Meeting → Offer → Deal — with live search, channel and service filters, a Follow-ups due view with overdue markers on cards and the detail sheet, and follow-up-due statistics.
DoneBuilt with add/edit/delete, stage forward/back plus direct stage jumps, per-stage counts, and the honest reply-rate formula kept. Saved lead data is untouched: older rows stay on their stage, Indonesian option values display and edit under their English equivalents, and nothing is rewritten on read. The owner gate was verified: guests are refused cleanly with owner-required. Tokens verified exactly against this dashboard.
DoneApproved via your standing instruction of Friday, 6 November 2026 — treat every build's QA as OK and hold corrections for one final pass. Owner-gate checks verified the clean guest refusal. Unverified item for your final correction pass: the owner CRUD path could not be run through the sandbox checker, which carries no viewer identity; the evidence is the same handler code as the previous working build plus the identical gate pattern to the NewCode board. Your own first use is the final proof.
DonePublished Friday, 6 November 2026 on QA approval via the standing instruction.
DoneLead rebuild: COMPLETE 5/5 — tool 2 of 5. Your saved leads were not modified.
Rebuild the follow-up generator in full English on Brand System v1, keeping the follow-up engine and adding a complete proposal layer.
DoneThe template engine — stage × follow-up number × channel (LinkedIn / Email / WhatsApp), including the polite closer at number 3 and send-timing guidance — plus a tier-aware rate table for three services, each with genuinely different scope content.
DoneBuilt with hours per week defaulting to 40 and hard-locked at a maximum of 40, with a visible capped note when a higher figure is typed. Platform-fee slider with gross/net maths, USD totals plus an IDR estimate at the displayed rate, per-service revision policy and trial offer, .txt download, and real calendar logic — suggested send dates and completion estimates computed from actual dates over working days, with weekday names calculated, not hard-coded. Default maths verified by hand: Professional evaluation US$30/hr × 20 hrs = US$600 gross, US$540 net after the 10% fee, about Rp8,910,000 at 16,500.
DoneApproved via your standing instruction of Friday, 6 November 2026 — treat every build's QA as OK and hold corrections for one final pass. Hand-verified default maths and calendar-logic checks over real dates.
DonePublished Friday, 6 November 2026 on QA approval via the standing instruction.
DoneLead rebuild: COMPLETE 5/5 — tool 3 of 5.
Rebuild the meeting brief generator in full English on Brand System v1, with agendas that add up exactly.
DoneThree duration bands that join with no gaps, and segment totals that equal the meeting length exactly: 15 minutes (0-2-7-12-15), 30 minutes (0-4-12-22-27-30), and 45 minutes (0-5-18-35-42-45).
DoneBuilt with the new 45-minute agenda option, a goal-specific first-30-seconds opening line, a Likely pushback — and honest replies section (capacity, rates, unpaid tests, quality, time zones), a live brief summary bar, and a sixth preparation checklist item. Capacity reads up to 40 hours per week in six places. The original edition's document-language toggle was removed, in line with the full-English decision for this series. Light and dark palettes copied exactly from the verified proposal generator.
DoneApproved via your standing instruction of Friday, 6 November 2026 — treat every build's QA as OK and hold corrections for one final pass. Agenda-maths checks verified all three bands segment by segment.
DonePublished Friday, 6 November 2026 on QA approval via the standing instruction.
DoneLead rebuild: COMPLETE 5/5 — tool 4 of 5.
Rebuild the closing showcase for the final office, on Brand System v1, reporting the four rebuilt tools at their true state.
DoneThe funnel storefront: the four tools summarised with their verified figures, the office workflow and role split, and the series statistics.
DoneBuilt as the closing showcase for Lead Engine: Outreach Message Generator, Lead Pipeline Board, Proposal Follow-Up Generator, and Meeting Brief Generator presented with the figures verified in their own rebuilds — the 40-hour cap, the tier-aware proposal maths, the exact agenda bands, and the untouched saved-lead record. Sending stays manual throughout, stated on the page exactly as in the tools.
DoneApproved via your standing instruction of Friday, 6 November 2026 — treat every build's QA as OK and hold corrections for one final pass. Showcase checks covered the four tools' figures against their rebuild records.
DonePublished Friday, 6 November 2026 on QA approval via the standing instruction.
DoneLead rebuild: COMPLETE 5/5 — Lead Engine closes 5/5, and with it the whole programme: 20/20.
What actually changed in each completed rebuild, dated — all twenty projects across the four offices, plus the Brand System v1 unification. Every entry stands at COMPLETE 5/5, QA approved via your standing instruction of 6 November 2026. Unverified hands-on items (FG-03, FG-04, LW-03, LE-02) are recorded in their trackers above for your final correction pass.
Rebuilt in full English. Added a revision and coordination buffer, the effective rate after platform fees, a start date and estimated finish calculated from real working days, client and project names carried automatically into the proposal text, a rate comparison table, a downloadable .txt proposal, a copyable WhatsApp summary, and a pre-send checklist. Everything the original did still works: sliders, preset chips, the below-floor warning under US$10/hour, and reset. Default maths verified: 20 hours + 10% buffer = US$330 gross, US$297 net after a 10% fee.
Rebuilt in English from the start — the one NewCode project that needed no translation pass. The server-backed board keeps the five-stage workflow (Brief → Design → Build → QA → Publish) with per-project cards and a per-office filter, and its data is saved so it persists between visits. Because it stores data, it is also the one rebuild that cannot be publicly linked from the client website; see the website plan below for the static-showcase route.
Rebuilt in full English. Added an automatic proposal number, professional scope templates for five services, an explicit Deliverables section, a client email field with a send button that composes the complete proposal email, transparent 5/10/15% volume discounts in the price table, a 7/14/30-day validity choice, data-confidentiality and ownership-on-payment clauses, and a print stylesheet for a clean printed document. Default maths verified (20 hrs × US$15 = US$300). Capacity decision, on the record: an early draft carried an outdated capacity line; at your instruction it was fixed to up to 40 hours/week — header chip, document header, and the default 40 hours/week behind its timeline estimates.
Rebuilt in full English as the Professional Edition. A new, sharper visual system (its own Libre Franklin identity, since superseded by Brand System v1 — see the suite entry below); a rebuild-programme section that tracks the 5/5 series of all four offices at their true status — NewCode in progress, Forge next (with the note that its games are being rebuilt from the ground up), LWD Studio and Lead Engine waiting; a professional-standards section; and a copy-summary button producing a series summary ready to paste into the AI Luthfi Office website. Proof numbers recalculated and verified; the showcase filter, layout, and dark mode intact. Availability aligned to up to 40 hours/week in five places in this final update, with no outdated capacity wording remaining anywhere.
This page — rebuilt last, in full English, so it could report the state above as fact. It carries the honest rebuild taxonomy end to end: all four offices at COMPLETE 5/5 in this final record, the NC-04 capacity line fixed to up to 40 hours/week in the embedded generator as well, with the field hard-capped at 40, and both embedded tools (NC-02, NC-04) working as before.
At your instruction, before QA: the five editions' separate identities were unified into NewCode Brand System v1 (section 01) — Archivo + IBM Plex Mono as the only type pairing, NewCode Amber #A86A00 on warm paper as the only accent, square corners, one card, one button pair, and one status language. The Calculator's teal/Plus Jakarta Sans, the Portfolio's Libre Franklin/pill system, and the Generator's Georgia document styling are retired as suite identities. This dashboard is rebuilt on v1 as the reference implementation; matching v1 is now a QA gate for every edition. In this final record the whole suite stands COMPLETE 5/5.
Rebuilt from the ground up. Three difficulty modes (Casual 1–50 ×1, Classic 1–100 ×1, Expert 1–100 in 7 attempts ×1.5), a narrowing possible-range bar, session best and win streaks, How to Play and GDD-style Design Notes. Logic verified: a correct guess costs no attempt; duplicate and out-of-range guesses are rejected free; score = attempts left × 10 + speed bonus, times the difficulty multiplier. Shake, pop, confetti, and mutable WebAudio sound.
Rebuilt from the ground up. Fisher–Yates shuffle, per-turn input lock, timer from first flip, 850 ms miss flip-back. Scoring harness-verified: perfect Easy 690, messy Hard 610, exactly per the stated formula. Combo streak heats at ×2, match sounds climb the C-major scale, flip hiss, win arpeggio, 44-piece confetti — all off under reduced-motion.
Rebuilt from the ground up. Standard, Focus (decoys), and Lightning modes; 3-2-1 countdown; shrinking bullseye landing far from its previous spot; live millisecond counter; auto-pause on hidden tab; full keyboard play; 50 ms hit-stop and particle bursts. Results analysis: fastest/slowest, consistency (standard deviation), distribution graph, personal trend. Server-saved sessions, sandbox-tested. Hands-on feel listed as an unverified item in its tracker.
Rebuilt from the ground up on canvas: particle bursts, floating score popups, hit-stop, screen shake, hit flashes. Combos to ×4, ranks Bronze to Diamond, and a 10-level capped campaign with a real win state and end screen. Full keyboard play, exact pause timing, screen-reader score announcements, honest NOT SAVED autosave status. Your best score of 180 preserved. Interactive canvas play listed as an unverified item in its tracker.
Rebuilt as the official Forge hub after an earlier build error placed a target game in its slot — caught, reported, and corrected. Four game cabinets with pitch, genre, trained skill, modes, and Play buttons (relative cross-artifact links; public links at the website stage); Target Hunt cabinet shows your series best of 180; three daily challenges rotating at midnight from a pool of nine; five-node level path L1 Deduction to L5 hub; series design notes. Forge closes 5/5.
Rebuilt in English on Brand System v1. Tested headless across all 54 combinations (6 angles × 3 tones × 3 durations) — every package complete, 8 fields plus 2 alternative titles. Three-tier chunk system cut worst word-budget deviation from +60–70% to +11%; sample 45-second tutorial package 108 words against a 105 budget. Tone and platform controls genuinely change output; Craft Notes document the duration × 140 wpm formula.
Rebuilt in English on Brand System v1. Every card carries the complete production text package with per-field and whole-package copy. QA gates are server-enforced and sandbox-tested: a card was held at Voice Over, then at QA, and reached Publish only after its manual ticks were filled; publish readiness is computed from all five gates. Seeded with one complete English EXAMPLE card; an older Indonesian example card does not auto-replace, by the anti-resurrection rule.
Rebuilt in English on Brand System v1. Live editor, word count with Relaxed/Normal/Fast estimates, auto-scroll reading screen with progress bar and centre guide, mirror mode, speed and text-size sliders, six-point VO guide, and a benchmark table highlighting the script's position. Legacy [jeda] markers still recognised alongside [pause]. New: Space play/pause, R restart, auto-pause on hidden tab. Scrolling screen listed as an unverified item in its tracker.
Rebuilt in English on Brand System v1 — tokens verified from source (Paper, Panel, Ink, Amber; no retired fonts or colours). New: a live platform spec strip and an automatic Package Quality Check (title fitness, hashtag count, hook source, pinned-comment keyword, caption length). All original behaviour kept: validation, character hints, platform title rules including the #Shorts suffix, the hashtag engine, five CTAs, per-block and whole-package copy, and the five-item checklist.
Rebuilt with figures cross-verified against the tools themselves: the LW-01 sample re-run headless and matched (108/105 words, about 46 seconds); the LW-03 sample re-counted at 134 words, openly correcting the original edition's stale figures; LW-02 counts taken live through its own action (1 EXAMPLE card, gates 3/5); LW-04 specs read from its platform table. Brand tokens compared directly against this dashboard. LWD Studio closes 5/5.
Rebuilt in English on Brand System v1 — source holds only the 12 permitted v1 hex values. Tested headless end to end: empty-name error with focus, default DM at 217 characters with personalisation warning, warning cleared and company name woven in when a point is added, byte-exact signature, correct warning on the fictional example, over-long points trimmed to keep the DM at 262 characters. Templates state Outlier experience since August 2026 and up to 40 hours/week. EN/ID toggle kept.
Rebuilt in English on Brand System v1, tokens verified exactly against this dashboard. Six stages (Target → Contacted → Replied → Meeting → Offer → Deal), live search, channel and service filters, Follow-ups due view with overdue markers, and follow-up statistics. Saved lead data untouched — older rows keep their stage and Indonesian option values edit under English equivalents. Owner gate verified (guests refused cleanly). Owner CRUD path listed as an unverified item in its tracker.
Rebuilt in English on Brand System v1. The stage × follow-up number × channel engine kept, plus a full proposal layer: tier-aware rate tables with genuinely different scope per service, platform-fee slider, USD totals plus IDR estimate at the displayed rate, revision policy and trial offer per service, .txt download, and real calendar logic over working days. Hours per week defaults to 40 and is hard-locked at 40 with a visible capped note. Default maths hand-verified: US$600 gross, US$540 net, about Rp8,910,000 at 16,500.
Rebuilt in English on Brand System v1. Agenda maths verified: three bands join with no gaps and segments total exactly — 15 min (0-2-7-12-15), 30 min (0-4-12-22-27-30), 45 min (0-5-18-35-42-45). New 45-minute option, goal-specific opening line, Likely pushback — and honest replies section, live summary bar, and a sixth checklist item. Capacity reads up to 40 hours per week in six places; the old document-language toggle was removed for the full-English series.
Rebuilt as the closing showcase for the final office, on Brand System v1: the four rebuilt tools presented with their own verified figures — the 40-hour cap, tier-aware proposal maths, exact agenda bands, and the untouched saved-lead record. Sending stays manual throughout, stated on the page. Lead Engine closes 5/5, and with it the programme: 20/20.
This dashboard and the AI Luthfi Office website (ai-luthfi-office.pages.dev) do different jobs. This dashboard is yours — the internal control map where builds, QA, and publish decisions are tracked honestly. AI Luthfi Office is the clients' site — the storefront where the four offices present what a client can order, with the Inspection Desk as the shared quality gate.
The website today presents the four offices, their leads and sprites, the Case Files of the original 20 projects, services, and FAQ. Installing these twenty professional rebuilds into that site as working showcases is the next, separate step — it is not claimed as done on this page. The known routes carry over: calculators and generators link directly once public links exist (hubs currently use relative cross-artifact links), and NC-03, the Project Board, and the other data-storing boards take a static showcase version rather than a public link to the live application.
Your corrections, when you make them in one final pass, apply across the dashboard, the twenty rebuilds, and then the website installation.