A phase-safe map of the roles Mohammad Rahimi personally carried, compressed, or coordinated across MZN: Phase 1 founder-led/team-built Mazzaneh execution, Phase 2 bounded solo AI-native asset formation, and Phase 3 professional context signal ahead.
Phase 1 (2020–2024): Mazzaneh MVP/company operation, founder-funded, with a 27-person execution team. Phase 2 (2024–2025/early 2026): bounded solo AI-native asset formation. Phase 3: independent context signal, legal/IP review, technical assessment, pilots, partnerships, and commercialization.
Decided not to compete with e-commerce giants head-on. Invented request-first commerce. When the first version failed, rebuilt from scratch. When 12,000 businesses pre-registered, didn't rush — waited until ready.
Closed the office, released 27-person team, continued alone. Not cost-cutting — a strategic bet that one person with AI could outproduce a team if the person knew what to build. Documented output volume increased, subject to evidence review.
The Phase 1 to Phase 2 transition itself. This appears to be an unusual and reviewable transition of a founder deliberately dissolving a functioning team to test a bounded solo AI-native formation model.
Losing the team looked like collapse. It became a strong review signal: if documented output volume increased, subject to evidence review after the team left, the architecture was always in the founder's mind.
22 integrated modules: Radar, Begir, Board, Pulino, Gram, Auto Chat, Style Finder, My Closet, My Size (95% accuracy), Analytics, VIP Pages. Each distinct yet interconnected — like organs in a body.
HUAI (9 layers, 51+ deep-dives, 92 tests), GPU Sentinel (120+ metrics), Tokenizer (12+ nodes), ISBP (4-stage), 5 LLM frameworks, BioCode (5 disciplines), HDTP (12 candidate claim areas), Zoyan (4 personalities).
Mazzaneh was reverse-engineered from Zoyan. Started with "what would a 24/7 AI assistant need?" and worked backwards. Architectural thinking most companies never do.
No developer team meant no one saying "technically impossible." Architecture unconstrained by legacy code.
BioCode (5 disciplines, 10 candidate claim areas). HUAI (L0-L8 complete). DCA, UIOP, Multi-Brain, Suprompt, OFRP. Modeled efficiency potential; subject to technical and commercial validation. Tokenizer: 12+ nodes, 6+ patent-grade candidate areas.
A mechanical engineer documented architecture-level AI research candidates across frontier-lab-adjacent problem spaces. Apparent convergence points with major companies require timestamp, similarity, and independent review.
Outsider perspective. No inherited assumptions. Approached LLMs as systems to diagnose, not magic to accept. Found what insiders miss.
No AI background seemed disqualifying. Became the advantage: fresh eyes see patterns that expertise normalizes into invisibility.
ISBP (4-stage intent-security). ZOE (20+ layers, 380+ components). GPU Sentinel (120+ metrics, 4 algorithms). 254 security assets. 23 protocols. vulnerability analyses, PoC candidates, and patch proposals subject to independent security review.
ISBP is an intent-security architecture candidate. GPU Sentinel is a dedicated GPU security / monitoring platform candidate. No obvious public commercial equivalent has been identified in the same category, pending market and security review.
168K organic users in 7 months. One city. Zero ads. Pitched as "shopping mall under construction." 12,000 businesses pre-registered. Designed Board: gamified ads, high-intent engagement, reduced waste.
Crunchbase signal, dated May 22, 2026: #2 in People across all categories; #1 outside the United States; #1 in Machine Learning and Cyber Security filters. Rankings may change over time and are not official endorsement, valuation, technical validation, or IP validation. Never attended one event.
Embassy closed before Lisbon. Internet cut before Qatar. War started. External signals were pursued primarily through documented materials rather than in-person networking.
27 people — developers, designers, support, operations. Iran's hyperinflationary economy. Sanctions. Full HR from recruitment to payroll.
Released entire team. Not because they were bad — next phase required different model. Documented output volume increased, subject to evidence review. The innovation: knowing when a team helps vs hinders.
Web Summit ALPHA, ongoing with VP of Startups. Slush selected. WSA nominated. TechCrunch applied. VivaTech engaged. Multiple AI companies evaluated.
Every email might be the last before blackout. Every communication maximally effective. No follow-up chains. One shot per connection.
Across 2024–2026, MZN documented a set of architectural design patterns that appear to overlap with capabilities later visible in major AI products or research directions. These should be treated as apparent convergence points, subject to timestamp, similarity, and independent technical review.
This is not presented as a claim of copying, priority ownership, or technical validation. It is a review signal: serious builders often converge on similar solutions when they face similar structural problems. The relevant question is whether MZN’s timestamped artifacts show meaningful independent architectural alignment with frontier-adjacent problem spaces.
Full timestamp documentation, idea-by-idea similarity analysis, and corresponding source files should be reviewed under coordinated diligence before any strong conclusion is drawn.
| Constraint | Expected Result | Actual Result |
|---|---|---|
| No team | Lower output | Higher output — zero coordination overhead |
| No budget | Cannot build | Built more efficiently — reduced waste |
| 1% internet | Cannot communicate | Every interaction maximally impactful |
| No events attended | Cannot network | Digital evidence became more important than physical presence |
| No CS education | Conventional objection: cannot map AI systems | Outsider perspective may have surfaced overlooked patterns |
| No conventional coding role | Cannot build tech | Designed architecture for later implementation review |
| International sanctions | Cannot access tools | Built with tools available to everyone |
| Active military conflict | Cannot focus | Urgency contributed to timeline compression |
| English as second language | Poor documentation | 3,000+ pages unusually extensive for a second-language documentation effort |
The constrained Phase 2 output is not final proof. It is the reason to test the case under better Phase 3 conditions: expert review, legal/IP counsel, technical validation, pilots, partners, and selective team formation.
The next step is independent review, not automatic belief.
15 role surfaces. 330+ mapped assets. 8 domains. 22+ patent-grade candidates pending professional review. Phase 1, Phase 2, and Phase 3 must be read separately.