I build the prototype in Claude Code, not just the spec.

@@ −2026.02 +2026.10 @@ mojtoba.com

1. 5+Added: A changelog of my career, written and built with Claude Code and the Impeccable design skill.built with AI
2. 6+Added: Every sentence checked against my own Claude writing standard: no figure that is not on my resume.built with AI
3. 7−Removed: Left out an AI chat widget. Shipped work tells a recruiter more than a chatbot does.

@@ −2025.12 +2026.02 @@ ReturnKoto?

1. 8+Added: A live pipeline that fetches, parses, and reconciles filings for 38 funds from multiple sources, with fallbacks when a source fails.built with AI
2. 9+Added: A publish-or-withhold rule: a figure is either right or not shown. Drift detection stops a bad sync before it reaches the page.
3. 10+Added: A Claude skill that holds every post and page to one writing standard: no unverified figures, no vague authority, no filler.built with AI
4. 11?Fund B, 5-year return: 00.0%figure not shown, withheld: two sources disagree. Illustration of the rule, not live data.

@@ −2024.11 +2025.12 @@ InsideMaps

1. 12+Added: Launch scope, launch criteria, and the one metric that decides whether a pilot continues, across 4 regions.
2. 13+Added: Working prototypes built in Claude from customer operational problems, each shipped with a written spec.built with AI
3. 14−Removed: Cut work nobody needed: the prototype gets product, engineering, and commercial to agree on what not to build.

@@ −2021.07 +2024.11 @@ Deloitte

1. 15+Added: Two workflow products taken from zero to production inside existing enterprise systems. I owned the PRD, priority, delivery, and rollout.
2. 16−Removed: Cut manual processing by 20%.
3. 17+Added: AI proofs of concept with business cases and ROI frameworks, so executives could decide where to invest. Architecture and model trade-offs settled with engineering before anything shipped.built with AI
4. 18+Added: Acceptance criteria and kill criteria, written with operators and the C-suite.

@@ −2020.08 +2021.07 @@ SELISE Group

1. 19!Fixed: Third-party API integrations failing. Success rate up 45% after I wrote the PRDs, API contracts, data mappings, and acceptance criteria partners shipped against.
2. 20−Removed: Cut partner support tickets by 60% by automating the manual handoffs where integrations broke.
3. 21−Removed: Cut time-to-market by 25% across four enterprise platforms, including Subway, ViCAFE, and Berger Paints, with one shared backlog across engineering, QA, and design.
4. 22~Changed: Recurring regional requests merged into one repeatable platform backlog.

@@ −2019.11 +2020.08 @@ Sheba Platform

1. 23+Added: Payments for 150K+ micro-entrepreneurs on iOS and Android, shipped to 90% rollout.
2. 24+Added: 100+ customer interviews and continuous funnel tests in Amplitude and Mixpanel.
3. 25~Changed: A change shipped only after it moved activation or payment conversion.

@@ −2018.12 +2019.11 @@ SWAP

1. 26+Added: The product, from persona mapping and discovery workshops to App Store launch, with a 23-person R&D team.
2. 27+Added: Engagement and product-market-fit evidence that helped raise a $1.25M seed round.

@@ −2016.07 +2018.12 @@ GameOf11

1. 28+Added: The app, from zero to 100K+ users on iOS, Android, and web.
2. 29+Added: Coin packs from 20 BDT to 500 BDT and up, so a new player could start small and a regular could buy in bulk.
3. 30~Changed: The referral reward pays out only after the new user enters a contest, so the incentive buys players, not empty signups.
4. 31−Removed: Dropped reward mechanics that did not move Day-7 or Day-30 retention or LTV, found through A/B tests in Amplitude, Mixpanel, and SQL.

@@ +2016.07 @@ Flow Digital

1. 32+Added: Multi-year roadmaps for a portfolio of B2B SaaS and consumer web and mobile products.
2. 33+Added: Flutter mockups for rapid prototyping, handed from design to senior engineering as working screens.