Notes from the work

Field Notes

Three first-hand notes about decisions that can be inspected in this portfolio, create-mohsen-app, and the Project Fit flow. Every note carries its context, constraint, rejected options, decision, learning, and evidence.

This replaces the generic seed-article surface: it does not publish unsupported metrics, demo projects as proof, or implied client experience.

Reference notes

Decisions with their evidence attached

Open a note for the complete decision trail and the artifacts that make it reviewable.

Distribution assets

Six useful atoms from every published case

Each atom keeps the full-case link and evidence boundary beside the reusable copy, so it can travel without turning delivery into an exaggerated result.

Case-study atom mapMohsen’s portfolioAn evidence-led evaluation product that separates audience jobs, publishable proof, bounded decision support, and the next conversation.

One-line summary

One-line summary

This bilingual Next.js portfolio routes clients, recruiters, delivery partners, enterprise evaluators, and technical reviewers from a relevant promise to registered evidence and a focused next step.

Evidence boundary

Mohsen’s portfolio is a delivered independent bilingual product with public source and three registered real route captures. Conversion, adoption, lead, recruiter-response, and hiring outcomes are not claimed.

Read the full portfolio self-case
English desktop home page of Mohsen's portfolio

Visual slide

Visual slide

Slide caption: the registered `/en` desktop home capture shows the delivered promise-and-proof hierarchy; separate Persian mobile and Work captures verify RTL responsiveness and the evidence index.

Evidence boundary

Mohsen’s portfolio is a delivered independent bilingual product with public source and three registered real route captures. Conversion, adoption, lead, recruiter-response, and hiring outcomes are not claimed.

Read the full portfolio self-case

Decision note

Decision: publish proof through a positive allowlist

Instead of rendering every database record or trusting persuasive copy, the portfolio registers evidence status, source, asset usage, and project slugs before a case can become public. The extra editorial step buys a reviewable claim boundary.

Evidence boundary

Mohsen’s portfolio is a delivered independent bilingual product with public source and three registered real route captures. Conversion, adoption, lead, recruiter-response, and hiring outcomes are not claimed.

Read the full portfolio self-case

Technical snippet

Technical trace

Visitor intent → locale-aware route → typed audience message → verified Evidence Bank entry → admitted case or tool → intent-preserving contact handoff; unsupported project media never resolves as proof.

Evidence boundary

Mohsen’s portfolio is a delivered independent bilingual product with public source and three registered real route captures. Conversion, adoption, lead, recruiter-response, and hiring outcomes are not claimed.

Read the full portfolio self-case

Interview answer

Interview answer

I treated the portfolio as an evaluation product, not a gallery. I separated audience jobs, moved claim truth into typed evidence registries, kept Navigator and Decision Lab deterministic, and let the cosmic layer fall back before it competes with reading or constrained devices.

Evidence boundary

Mohsen’s portfolio is a delivered independent bilingual product with public source and three registered real route captures. Conversion, adoption, lead, recruiter-response, and hiring outcomes are not claimed.

Read the full portfolio self-case

CTA

CTA

Open the full self-case to inspect the audience-routing, evidence-publication, deterministic-assistance, bilingual-responsive, and static-first visual decisions beside the real screens and public source.

Evidence boundary

Mohsen’s portfolio is a delivered independent bilingual product with public source and three registered real route captures. Conversion, adoption, lead, recruiter-response, and hiring outcomes are not claimed.

Read the full portfolio self-case
Case-study atom mapQuaizA document-centred learning product that keeps generated quiz content behind a product-owned data contract.

One-line summary

One-line summary

Quaiz connects study documents, AI-assisted generation, structured quizzes, and persistent practice in a public Next.js learning product.

Evidence boundary

Quaiz is a delivered public product with inspectable source and one registered real screen. Usage, learning impact, and production scale are not claimed.

Read the full Quaiz case
Quaiz application screen

Visual slide

Visual slide

Slide caption: a real Quaiz interface anchors the path from a learner’s study context to inspectable quiz practice.

Evidence boundary

Quaiz is a delivered public product with inspectable source and one registered real screen. Usage, learning impact, and production scale are not claimed.

Read the full Quaiz case

Decision note

Decision: validate before persistence

Generated JSON can be syntactically valid and still miss required quiz structure. The chosen boundary parses a named Zod contract before the Prisma mutation instead of trusting the payload or validating during rendering.

Evidence boundary

Quaiz is a delivered public product with inspectable source and one registered real screen. Usage, learning impact, and production scale are not claimed.

Read the full Quaiz case

Technical snippet

Technical trace

Learner UI → Next.js boundary → structured AI payload → Zod validation → nested Prisma write for the quiz, questions, options, ownership, and result relationships.

Evidence boundary

Quaiz is a delivered public product with inspectable source and one registered real screen. Usage, learning impact, and production scale are not claimed.

Read the full Quaiz case

Interview answer

Interview answer

When AI output reaches product data, I do not treat it as trusted application state. In Quaiz, I placed validation at the API boundary and kept the document as the ownership context for related learning records.

Evidence boundary

Quaiz is a delivered public product with inspectable source and one registered real screen. Usage, learning impact, and production scale are not claimed.

Read the full Quaiz case

CTA

CTA

Inspect the full case to review the document-centred model, validation decision, nested write path, live product, and evidence limits.

Evidence boundary

Quaiz is a delivered public product with inspectable source and one registered real screen. Usage, learning impact, and production scale are not claimed.

Read the full Quaiz case
Case-study atom mapcreate-mohsen-appA composable project generator that makes intent and destination safety precede filesystem side effects.

One-line summary

One-line summary

create-mohsen-app turns guided setup choices into a configurable Next.js starter through a published TypeScript CLI.

Evidence boundary

create-mohsen-app is a published independent CLI with public source and a registered real screen. Adoption, downloads, and time saved are not claimed.

Read the full create-mohsen-app case
create-mohsen-app CLI

Visual slide

Visual slide

Slide caption: the real prompt surface collects project intent before template composition and the final optional install.

Evidence boundary

create-mohsen-app is a published independent CLI with public source and a registered real screen. Adoption, downloads, and time saved are not claimed.

Read the full create-mohsen-app case

Decision note

Decision: refuse an unsafe destination

Automatic merging into a non-empty folder risks conflicting with user work. The CLI stops and explains the target conflict, accepting less convenience in exchange for predictable failure.

Evidence boundary

create-mohsen-app is a published independent CLI with public source and a registered real screen. Adoption, downloads, and time saved are not claimed.

Read the full create-mohsen-app case

Technical snippet

Technical trace

Prompt answers → resolved configuration → target guard → base template plus selected extras → generated files → optional dependency installation.

Evidence boundary

create-mohsen-app is a published independent CLI with public source and a registered real screen. Adoption, downloads, and time saved are not claimed.

Read the full create-mohsen-app case

Interview answer

Interview answer

For a code generator, I treat cancellation and filesystem safety as product behavior. I collect required choices first, reject unsafe targets, compose only selected templates, and leave installation until the generated files exist.

Evidence boundary

create-mohsen-app is a published independent CLI with public source and a registered real screen. Adoption, downloads, and time saved are not claimed.

Read the full create-mohsen-app case

CTA

CTA

Open the full case to inspect the CLI source, template tree, package, real screen, trade-offs, and current evidence boundary.

Evidence boundary

create-mohsen-app is a published independent CLI with public source and a registered real screen. Adoption, downloads, and time saved are not claimed.

Read the full create-mohsen-app case
Case-study atom mapThelegroumA real-time messaging experiment that keeps durable conversation state authoritative and live events transient.

One-line summary

One-line summary

Thelegroum combines authenticated conversations, durable message records, live Pusher events, and responsive navigation in a public Next.js product experiment.

Evidence boundary

Thelegroum is a public independent product experiment with a live demo, source, and one registered real screen. Reliability, latency, scale, and active usage are not claimed.

Read the full Thelegroum case
Thelegroum application screen

Visual slide

Visual slide

Slide caption: a real conversation interface connects the write-then-publish architecture to the product surface a user sees.

Evidence boundary

Thelegroum is a public independent product experiment with a live demo, source, and one registered real screen. Reliability, latency, scale, and active usage are not claimed.

Read the full Thelegroum case

Decision note

Decision: persist, then publish

An event-only system is fast but cannot be the conversation history. The chosen path writes the message through Prisma first, then publishes a Pusher event so durable state remains recoverable truth.

Evidence boundary

Thelegroum is a public independent product experiment with a live demo, source, and one registered real screen. Reliability, latency, scale, and active usage are not claimed.

Read the full Thelegroum case

Technical snippet

Technical trace

Authenticated client → Next.js action or route → Prisma conversation write → Pusher channel event → subscribed clients reconcile against durable state.

Evidence boundary

Thelegroum is a public independent product experiment with a live demo, source, and one registered real screen. Reliability, latency, scale, and active usage are not claimed.

Read the full Thelegroum case

Interview answer

Interview answer

In real-time products, I separate authoritative state from delivery signals. For Thelegroum, Prisma owns recoverable conversation history while Pusher improves immediacy; the interface still has to reconcile both paths.

Evidence boundary

Thelegroum is a public independent product experiment with a live demo, source, and one registered real screen. Reliability, latency, scale, and active usage are not claimed.

Read the full Thelegroum case

CTA

CTA

Open the full case to review the write route, Pusher adapter, conversation model, responsive shell, live demo, and stated evidence limits.

Evidence boundary

Thelegroum is a public independent product experiment with a live demo, source, and one registered real screen. Reliability, latency, scale, and active usage are not claimed.

Read the full Thelegroum case
Mohsen
Mohsen
contact@mohsen.info

Tehran, Iran

نسخهٔ فارسی

© 2026 Mohsen.