Comprehensive study of all project files, LLM reference prompts, scope discrepancies, and open questions.
All Floor - Only Yellow concerned.png
— Full B3 floor plan with yellow highlight on MediaVerse area (northwest corner)
B3.png
— Zoomed view of yellow area with dimensions (737, 334, 333, 431, 270, 111, 448 cm)
SD-HVAC-ALL-241225-V02_new.dxf
— HVAC mechanical DXF (converted from DWG via ODA, 38.8 MB, 151 layers, 43,927 entities). Samsung DVM VRF system.
Techtown Building office_ B3 Final Furniture layout 03.03.2026_new.dxf
— Architectural furniture layout DXF (converted from DWG via ODA, 6.8 MB)
Prompts for Reference/Gemini.txt
— Input prompt sent to Google Gemini (initial scope brief, 64 lines)
Prompts for Reference/Grok.txt
— Grok’s exhaustive 182-line final report (claimed “100% COMPLETE”)
Prompts for Reference/ChatGPT.txt
— ChatGPT’s planning report (411 lines, structured but less technical)
Prompts for Reference/GrokSession2.txt
— Full Grok session 2 transcript including DXF conversion, extraction scripts, layout iterations, and Q&A (very long, ~1000 lines)
Prompts for Reference/Claude Web.txt
— Claude web session: full 13-zone verification prompt + independent analysis attempt (very long, ~800+ lines)
Design an acoustic studio complex in the yellow-highlighted area of TechTown Damascus, Basement 3 (B3). The area is an irregular trapezoid in the northwest corner of a commercial office floor.
| # | Space | Purpose | Key Requirement |
|---|---|---|---|
| 1 | VO Live Room / Vocal Booth | Voice-over recording, 2-3 backing vocals, soft acoustic instrument | Room-within-a-room (box-in-box), floating floor, NC-15 to NC-20 |
| 2 | Control Room | Monitoring, mixing, producer position | Observation window into live room (triple-glazed), NC-25 |
| 3 | Video Podcast / Green Screen Studio | Video recording, podcast, chroma key shoots | Motorized retractable green screen, camera depth ~4.5m, NC-25 to NC-30 |
| 4 | Editing Workstations | Post-production, editing, programming (headphones only) | 2-4 desks, open plan, NC-30 to NC-35 |
All four LLMs (Gemini, Grok, ChatGPT, Claude) received similar data. Here are the key conflicts and their resolution status:
| Topic | Gemini | Grok | ChatGPT | Claude Web | Status |
|---|---|---|---|---|---|
| Total Area | 37 m² (nominal) | 39.5 m² (DXF-verified) | ~33 m² (7.3×4.5) | 26.75 m² (partial polygon) | CONFLICT |
| Scope of "Yellow Area" | ~37 m² media zone | 39.5 m² trapezoid (full yellow) | ~33 m² (underestimated) | ~26.75 m² (used partial polygon, wrong) | CONFLICT |
| Grok Session 2: ~92 m² | N/A | ~92 m² (full trapezoid) | N/A | N/A | NEEDS CLARITY |
| Ceiling Height | 280-320 cm (guessed) | 290 cm (user-confirmed) | Not stated | 290 cm (user-confirmed) | RESOLVED: 290 cm |
| Demolition | Varied | Full demo (locked) | Full demo | Full demo | RESOLVED: Full demo |
| Number of Zones/Rooms | 3 zones (VO+Control, Podcast, Editing) | 3 zones + buffer spine + airlock | 4 areas (VO, Control, Podcast, Editing) | 13-zone full floor approach (different scope) | VARIED |
| Live Room = Podcast Room? | Combined | Session 1: Combined. Session 2: Separated (after correction) | Separate | Sub-zones Z1a-Z1e (5 sub-zones) | CONFLICT |
| Green Screen Orientation | Not detailed | Option A: North wall (preferred). Option B: East wall (risky) | Backdrop / green screen (generic) | Not detailed | Grok most detailed |
| RT60 Target (VO) | 0.15-0.20 s | 0.20-0.25 s | Not specified | 0.20-0.40 s (EBU R22) | RANGE VARIES |
| NC Target (VO) | NC-20 | NC-25 max | Not specified | NC-15 (EBU R22) | RANGE: NC-15 to NC-25 |
| STC for VO Suite | STC-55+ (walls) | STC-68 (room-in-room) | Not specified | STC-58 to STC-68 (calculated) | CONVERGING |
| Editing Stations Count | Not specified | Open workspace (~9 m²) | 4 desks | 3 desks (EDIT-1/2/3) | 2-4 desks? |
| Slab-to-Slab Construction | Assumed feasible | NOT feasible in Syria — decoupled top seals instead | Not addressed | Not addressed | SYRIA-SPECIFIC |
| Control Room Sight Lines | Into live room only | Session 2: Both rooms if possible | Into live room | Into recording spaces | USER: "both if possible" |
| Claude Web: Scope | (All others focused on yellow area only) | Full floor 13-zone analysis (Z1-Z13) with Z1 subdivided into Z1a-Z1e | DIFFERENT SCOPE | ||
Check the boxes to confirm understanding, and fill in your answers where needed.
The yellow area in B3.png shows dimensions ~7.37m × ~4.31-4.48m = ~32-39 m². But Grok Session 2 later worked with ~92 m² (the full northwest trapezoid including area south of the column line). Which is correct?
Gemini and early Grok combined them. You corrected Grok in Session 2: "those should be 2 different rooms." Confirming: 4 distinct rooms?
You told Grok "both rooms only if possible." Is this a hard requirement or nice-to-have?
ChatGPT said 4. Grok Session 2 said 3. Claude said 3. Gemini didn't specify. All agree headphones-only.
Different LLMs used different targets. Which should I calculate to?
| Space | Conservative (EBU R22) | Moderate (ASHRAE/WELL) | Practical (Grok) |
|---|---|---|---|
| VO Live Room | NC-15 | NC-20 | NC-25 |
| Control Room | NC-20 | NC-25 | NC-25 |
| Video Podcast | NC-20 | NC-25 | NC-25 to NC-30 |
| Editing | NC-30 | NC-30 | NC-30 to NC-35 |
This folder is called "Acoustics2.0." Was there a previous version? If so, what was wrong with it?
Grok estimated ~731,000 SYP local / ~2,680,000 SYP Auralex. The Session 2 study estimated $28,000-$32,000 all-in. Are we working to a budget?
The LLMs produced various outputs: ASCII layouts, HTML visuals, DXF files, DOCX reports, interactive questionnaires. What do you actually need?
All LLMs flagged the east server room as a major noise threat. Do you have measured SPL data, or should we assume worst-case?
The south/southwest boundary has an angled glass facade. What type of glass is installed?
| LLM | Strengths | Weaknesses | Unique Contributions |
|---|---|---|---|
| Gemini | Clean initial prompt structure; good scope definition; explicit structural boundary descriptions | Used nominal 37 m² without DXF verification; no DXF parsing; generic advice in deliverable requests | Best-structured initial brief with clear section headings; explicit NC/STC targets per boundary wall |
| Grok | Most thorough overall; DXF-verified area (39.5 m²); exhaustive calculations with KaTeX formulas; dual options (A/B layouts); Syria-specific adaptations; cost estimates in SYP; full risk register (28 items) | Session 1 report claims "100% COMPLETE" but is a summary, not actionable drawings; combined live room with podcast initially; area later expanded to ~92 m² without clear justification | 14 column coordinates with sizes; Helmholtz/membrane resonator specs; local vs Auralex BOM comparison; slab-to-slab Syria workaround; buffer spine concept; "death by 10 cm" depth analysis |
| ChatGPT | Clear layman-friendly presentation; good workflow logic (entry→editing→video→studio progression); explicit "what we removed" section | Underestimated area (~33 m²); no formulas or calculations; no DXF data used; no acoustic engineering (just descriptions); stated 4 editing desks without justification | Best at explaining the "why" behind placement decisions in simple terms; explicit workflow progression concept |
| Claude Web | Most rigorous verification prompt (13 zones, 10 tables, 12 calculation sections); proper standards citations (WELL v2, ISO 3382-3, ASHRAE Ch.48, EBU R22); independent blind-verification methodology | Could not parse DXF files; used partial polygon (26.75 m² — wrong); stalled waiting for text-extracted geometry; attempted full-floor 13-zone scope when user only needed yellow area | Best verification framework; proper NC budget methodology (logarithmic energy summation); AVBC exclusion zone concept; cross-talk path analysis methodology; Samsung DVM published spec research |
| # | Risk/Assumption | Impact if Wrong | Priority |
|---|---|---|---|
| 1 | Exact area of yellow zone not confirmed | Entire spatial program is wrong — rooms won't fit | CRITICAL |
| 2 | Server room noise levels unknown (assumed ~55-65 dBA) | STC calc may be insufficient; NC target blown | CRITICAL |
| 3 | VR Lab impact vibration not measured | Floating floor spec may need upgrading | HIGH |
| 4 | Exact HVAC units serving yellow area not mapped | NC contribution from HVAC may exceed budget | HIGH |
| 5 | Facade glass type unknown | External noise ingress calculation unreliable | HIGH |
| 6 | Plenum depth assumed 40-60 cm | HVAC duct routing and silencer placement may not fit | MEDIUM |
| 7 | Duct velocity through 150×100mm sections (>6 m/s) | Regenerated noise in recording spaces | HIGH |
| 8 | No section drawings available | Wall buildup heights, plenum clearances unverified | MEDIUM |
| 9 | Local material availability in Damascus not confirmed | BOM may need revision; lead times uncertain | MEDIUM |
| 10 | Budget not defined | May design something that can't be built | MEDIUM |
| Model | Type | Airflow (m³/h) | Notes |
|---|---|---|---|
| AVD-09 | Concealed Duct | 420-540 | Small zones |
| AVD-12 | Concealed Duct | 660-720 | |
| AVD-15 | Concealed Duct | 720 | |
| AVD-19 | Concealed Duct | 780-900 | |
| AVD-24 | Concealed Duct | 840-1140 | |
| AVD-30 | Concealed Duct | 1680 | High airflow — noise risk |
| AVD-38 | Concealed Duct | 1680 | High airflow — noise risk |
| AVD-48 | Concealed Duct | 1750 | Highest airflow |
| AUD-36 | AHU | 2100 | Air Handling Unit |
| AVBC-30/38/48 | 4-Way Cassette | - | Ceiling mounted, SPL ~35-48 dB(A) |
Rectangular: 150×100 through 500×250 mm. Round: 150Ø to 350Ø mm. Fresh air duct tagged FA-350×200.
Check boxes and answer questions to track progress.