The Wintura Playbook — Exactly How I Ship Production MVPs at Soatech

TL;DR (Q2 2026). This is the day-by-day playbook for Soatech's MVP Sprint (€8,500–€22,000 fixed, 4-8 weeks). It's derived directly from how I shipped wintura.ai solo in 6 months — compressed into the Sprint format by skipping the validation and brand iteration phases I did personally. Week 1 is the Technical Blueprint (€2,500 standalone). Weeks 2-4+ are the architect-led, AI-accelerated build. Every step references the actual Wintura.ai case study where the pattern was validated.

Why the Wintura playbook works as a productized service

When I shipped Wintura, I figured out the architecture as I went — schema after week 2, AI pipeline split after week 6, multi-tenancy refactored at month 4. That worked because it was my own product. It would not work for a paying client.

The Soatech MVP Sprint compresses the 6-month Wintura experience into a 4-8 week delivery by front-loading every architectural decision. The €2,500 Technical Blueprint is the first 5 days — schema, API contracts, auth strategy, AI pipeline (if applicable), deployment topology — all locked before any production code is written. The remaining weeks are pure architect-led execution against that plan, with AI doing the boilerplate generation per the can-AI-build-your-app post.

Per Search Engine Land's 2026 content strategy guidance, procedural content with named steps gets cited in AI Overviews at higher rates than generic advice. This playbook is the named-steps version of how Wintura got built.

Day 0: Pre-engagement (free, 30 min)

A 30-minute scoping call. Your Architect (today, the founder — Alvi Lika) walks through your concept, target users, technical constraints, and budget. Outcome: a one-page memo of (a) whether Soatech is the right fit, (b) which productized track applies (Production Lift, Technical Blueprint, MVP Sprint, or Iteration Sprint), (c) start date.

No SDR layer, no nurture sequence. Direct call with the senior engineer who'll do the work.

Days 1-5: Technical Blueprint (€2,500 standalone or rolled into Sprint)

The five most important days of any soatech engagement.

Day 1 — Domain modeling

Your Architect maps your product's core entities. For Wintura: Tenants, Users, Proposals, Clients, Invoices, Templates. For a marketplace: Buyers, Sellers, Listings, Orders, Reviews. The output is an entity-relationship diagram with explicit cardinality and a tenancy boundary marker on every table that needs it.

Multi-tenancy decisions get made on Day 1, not Day 14. This is the single biggest difference between a Sprint and a Bolt prototype — see the Wintura.ai case study multi-tenancy section for why.

Day 2 — API contracts + auth strategy

Every API route is specified: request shape, response shape, error envelope, auth requirement. The auth strategy is locked: NextAuth v5 vs Clerk vs Supabase Auth (depending on your constraints), session resolution pattern, role/permission model, multi-tenant session handling.

For Wintura, the answer was NextAuth v5 beta.30 with magic link + password, apex-locked session cookie, tenant resolution in the session callback. For a fintech build, the answer might be Clerk + Stripe Identity verification + signed-JWT session tokens. The decision happens on Day 2, not via copy-paste from a Bolt scaffold.

Day 3 — AI pipeline + third-party integrations

If your product has AI features (most 2026 SaaS does), the pipeline gets designed on Day 3. For Wintura: Claude Sonnet 4.6 for structured generation, Haiku 4.5 for voice humanization, all prompts in /lib/ai/prompts/, no direct Claude calls from components. For your product: which model, where in the request flow, what's the fallback if the API is degraded, what's the cost ceiling per generation.

Third-party integrations also get scoped: Stripe (Standard vs Connect vs Identity?), email provider (Resend vs Postmark vs SendGrid?), CRM sync (HubSpot, Pipedrive, raw Postgres?), document storage (Cloudflare R2 vs S3 vs Vercel Blob?).

Day 4 — Test strategy + deployment topology

How will quality be verified? For Wintura: 24 Playwright e2e files including axe-core a11y across mobile/tablet/desktop viewports — see the Wintura case study test breakdown. For your product: the equivalent matrix scoped to your critical flows.

Deployment topology: Vercel for the Next.js app, Railway for any long-running services (PDF rendering, AI workers, scheduled jobs), Neon/Supabase for Postgres, R2/S3 for object storage, Resend for transactional email.

Day 5 — Blueprint document + Sprint kickoff

The Blueprint document is delivered: schema diagram, API contract spec, auth strategy, AI pipeline, test plan, deployment topology. Yours to keep regardless of whether you proceed to a Sprint. If you do proceed, the Blueprint fee converts toward Sprint One.

Weeks 2-8: MVP Sprint execution

The architecture is locked. Now we build against it. The timeline below shows a typical 4-week MVP Sprint Lite (€8,500); Standard and Plus tiers extend to 6-8 weeks with additional features and integrations.

Days 6-7 — Foundation

The skeletal Next.js 16 app with Tailwind v4, App Router, TypeScript strict, the database schema migration applied to Neon, NextAuth configured per the Blueprint's auth strategy. CI/CD wired to Vercel preview deployments. PostHog + Vercel Analytics + Speed Insights wired with typed event facades.

What AI handles (per my Wintura experience): all of the above scaffolding, generated in ~4 hours of pair-programming with Cursor + Claude Sonnet 4.6. What I review: the multi-tenant query layer, the session callback, the deployment configuration.

Days 8-12 — Core feature set

The features identified in the Blueprint as the V1 critical path. For Wintura: proposal creation, AI generation pipeline, TipTap editor, client portal, sealed-PDF rendering. For your product: the 3-5 features users will actually pay for.

AI generates the route handlers, the React components, the form validation (Zod schemas). I review for: tenant isolation enforcement, error edge cases, accessibility patterns. Test files get drafted alongside features — never deferred.

Days 13-18 — Integrations

Stripe (Standard or Connect per Blueprint), email provider, CRM sync, document storage, any first-party integrations identified in scoping. Each integration includes: webhook signature verification, idempotency handling, retry-with-backoff, structured error types per failure mode.

This is the zone where Bolt/Lovable exports break in production — generic try/catch is not enough, per the can-AI-build-your-app post. Every integration in Wintura survived production scrutiny because the failure modes were enumerated on Day 13, not after the first 502 from HubSpot.

Days 19-22 — Tests + accessibility

The Playwright e2e suite catches up to the feature set. For Wintura, 24 files covering flow / surface / auth / visual+a11y / API / live-smoke matrices — see the Wintura case study test section for the breakdown. axe-core a11y assertions on every public + private + onboarding route × 3 viewports.

Per Healthcare Success E-E-A-T research, production B2B software with WCAG 2.1 AA conformance signals trust at procurement time. The a11y matrix is a sales asset, not just a compliance line item.

Days 23-26 — Security audit + production hardening

Security headers (CSP, HSTS, X-Frame-Options, Permissions-Policy). CSRF protection. Rate limiting per endpoint per IP per session. Dependency audit via npm audit + Snyk. Secrets rotation pattern documented. Logged audit trails for all sensitive actions.

Same posture installed in Soatech's Production Lift when applied to a Bolt/Lovable prototype — but here it's baked in from Week 2, not retrofitted at Week 6.

Days 27-29 — Polish + handoff prep

Visual regression sweep across all routes. Performance audit (Lighthouse ≥90 on every public page). SEO infrastructure (sitemap.xml, robots.txt, JSON-LD schemas, meta descriptions per page, OG images). Handoff documentation: architecture decision records, runbook for production incidents, environment variable reference, key rotation procedure.

Final week — Production deploy + handover

Deploy to your infrastructure (Vercel/Railway/your-preferred). Full handover: codebase, schema, design assets, deployment credentials, documentation. 100% your property — no platform lock-in, no proprietary dependencies.

Post-launch options (Q2 2026):

• Iteration Sprint (€4,500–€8,500/sprint) — fixed-price bi-weekly sprints for ongoing feature development
• Feature Sprint (€6,000, 2 weeks) — one-off for the next priority feature
• Embedded Architect (€7,000/mo, 1 day/week) — ongoing architecture oversight and code review
• Production Subscription (€5,000/mo founding rate) — 2 feature modules/month, same architect, pause anytime

Why this is 2× faster than a hand-coded shop

Three compounding factors:

1. AI handles the 41% that's scaffold + boilerplate generation — per Taskade 2026, 41% of code shipped is AI-generated, and that compresses ~30% off raw build time. Wintura took 6 months; without AI it would have been ~9-10 months.
2. The Architect makes architectural decisions once, on Day 2-4, instead of discovering them mid-Sprint. A hand-coded shop with juniors hits these decisions repeatedly across the build and re-relitigates them at code review. Soatech's Architect locks them upfront.
3. The Blueprint compresses scope variance. A hand-coded shop charges hourly because they don't know what they'll find. Soatech charges fixed because the Architect-led Blueprint has already found it.

Per the Wintura case study build velocity comparison, the verified Wintura timeline (6 months solo) vs comparable hand-coded shop estimates (12-15 months for an equivalent B2B SaaS) is the 2× number Soatech sells against.

What this playbook deliberately doesn't include

• Design from scratch: if you need extensive custom design, that's a separate engagement. Soatech ships with shadcn/ui + Tailwind v4 + a token-based design system. For more, bring a designer or commission a separate design Sprint.
• Native mobile apps: the playbook is web-first. React Native is possible but adds 50% to scope. The MVP Sprint pricing reflects web scope.
• Compliance certifications: HIPAA, PCI DSS, SOC 2 — these require ongoing audit cycles, not a 30-day sprint. The Sprint installs the architecture compliant frameworks need; the certification work is separate.
• Marketing site: separate from the product. If you need both, the Sprint focuses on the product and a marketing landing is a separate ~5-day scoped add-on.

Frequently asked questions

What's actually in the €2,500 Blueprint deliverable?

A written document (typically 12-20 pages) covering: entity-relationship diagram with tenancy markers, API contract spec for every endpoint, auth strategy decision with rationale, AI pipeline design if applicable, third-party integration list with vendor decisions, test strategy with coverage targets, deployment topology with infrastructure decisions, architecture decision records explaining the why behind each choice. Yours to keep. If you proceed to a Sprint, the fee converts toward Sprint One.

How does a 4-8 week Sprint compare to hourly agency engagements?

Hourly agencies typically quote 300-600 hours over 8-12 weeks for the same scope, at €100-€200/hr. That's €30K-€120K. Per Chrono Innovation's Feb 2026 analysis, AI-supervised builds run $25K-$75K while traditional agencies quote $50K-$500K for equivalent B2B SaaS. The Soatech MVP Sprint is fixed-price from €8,500 and lands in 4-8 weeks. The 2-5× cost difference comes from AI accelerating generation + one senior architect instead of a juniors-with-PM hierarchy + fixed-price absorbing scope variance.

What if my product needs more than 8 weeks of work?

Two options: (a) scope a larger initial Sprint (custom quote based on Blueprint findings), or (b) ship V1 in 4-8 weeks and continue with an Iteration Sprint (€4,500–€8,500/sprint) for fixed-price bi-weekly sprints. Most Wintura-pattern B2B SaaS V1 builds fit cleanly in 4-6 weeks; products with extensive AI pipelines or complex integrations extend to 8 weeks (MVP Sprint Plus, €22,000).

Can I bring my own Bolt or Lovable prototype into the Sprint?

If you have a prototype already, you don't need the full Sprint — you need the Production Lift (€3,500, 1 week). The Lift takes your existing prototype and ships it to production: auth, multi-tenant isolation, e2e tests, security audit, deployment. The Sprint is for clean-slate builds with no existing code.

Who actually does the work?

Today, the founder (Alvi Lika). The same engineer who shipped Wintura.ai. As Soatech grows, additional Architects join — each engagement is still delivered by one senior Architect on-record for the full Sprint, no handoffs, no project managers in the middle.

What's the testing posture you ship?

The Wintura matrix: per-flow e2e specs, per-surface interaction specs, dedicated auth + onboarding specs, visual regression across mobile / tablet / desktop, axe-core a11y across public + private + onboarding routes × 3 viewports, API contract specs, and a live-smoke spec that runs against production on every deploy. Roughly 20-30 spec files for a typical V1, scaled to product surface area.

Does the Sprint include marketing-page SEO?

The Sprint includes meta descriptions, OG images, sitemap, robots.txt, JSON-LD Article + Organization + FAQPage schemas on every public page, and a /llms.txt for AI crawler discovery. It does not include content production (blog posts, case studies) — that's separate work. See Soatech's SEO posture for what's included by default.

Can I see a Sprint output before I commit?

Yes — see the Wintura.ai case study. Wintura is the verifiable proof of what the playbook produces. You can use the live product at wintura.ai, look at the public proposal samples, see the demo. The Sprint produces something equivalent for your domain.

How long does it take to build an MVP in 2026?

The answer depends on whether you're building from scratch or hardening an existing prototype:

• Prototype (Bolt/Lovable/v0/Cursor): Hours to days. AI tools generate functional demos rapidly, but these aren't production-ready.
• Production MVP from prototype: 1 week with the Production Lift (€3,500). Auth, multi-tenant isolation, e2e tests, security headers, deployment.
• Production MVP from scratch: 4-8 weeks with the MVP Sprint (€8,500–€22,000). Full Blueprint + architect-led execution.
• Traditional agency build: 12-16 weeks per TeaCode's May 2026 benchmark. Hourly billing, PM overhead, junior-senior review cycles.

The difference is front-loaded architecture. When the schema, auth strategy, and API contracts are locked on Day 5 instead of discovered iteratively, execution compresses by 2-3×.

What's the difference between a prototype and production code?

A prototype proves the concept. Production code serves real users reliably. The gap includes:

The Production Lift closes this gap in 1 week for €3,500. The MVP Sprint builds production code from Day 1.

Should I hire a developer or use a productized studio?

The choice depends on your stage and technical capacity:

Hire a developer when:

• You have a technical co-founder who can manage, review, and architect
• You need ongoing capacity (6+ months of continuous work)
• Your scope is undefined and will evolve week-to-week
• You can absorb 2-3 months of onboarding and ramp-up time

Use a productized studio when:

• You're non-technical and can't manage individual engineers
• You have a defined V1 scope and need fixed-price certainty
• You need production-quality code in 4-8 weeks, not 4-8 months
• You want walk-away ownership without ongoing HR obligations

For post-launch, the Embedded Architect (€7,000/mo, 1 day/week) bridges the gap: ongoing architecture oversight without the FTE commitment.