Folio № 001Paris area · 48.8566° NFounded June 2025
Fade in.

Most AI ships as a first draft.
We ship the production cut.

We deploy AI inside companies, and we build our own AI products. Production-grade engineering, evals before launch, monitoring after.

// What we do

Two activities, one studio.

№ 01

Deploy AI inside your business.

You already have a company, customers, and a list of places where AI would save real hours. We come in as senior engineers, scope the surface that moves a number, and ship it to production. Integration, agents, evaluations, monitoring. We stay until it runs without us, then hand back code you own.

No proofs of concept. No internal team to manage.

In productionCertiDealAn AI customer-support agent at CertiDeal. 45% of support tickets resolved without human intervention. First deployed 2025, renewed monthly since.
№ 02

Build our own AI products.

The other half of the studio is our own product line. Products where AI is the reason the product exists, not a feature added on top. We design, build, evaluate, and ship them under our own name.

  • SheediOS app for real-life matchmaking. Private beta on TestFlight.
  • EnsembleInternal applied-AI product. In active development.
  • FormloopInternal applied-AI product. In active development.
In private betaSheed iOS app screenshotSheed, our iOS matchmaking app. Private TestFlight access on request to jerome@manibor.com.
// How we build

Production stack, picked for compounding.

Senior engineers only. AI as the velocity multiplier. Every line of code is written by an engineer with 10+ years on the job, never by a junior learning on your dime. Same stack on client work and on our own products: mature where maturity matters (TypeScript, Postgres, Next.js), opinionated where the AI is load-bearing (Claude, Vercel AI SDK, agent patterns, evals before launch).

  • LanguageTypeScript
  • FrameworkNext.js · App Router · React Server Components
  • ModelsClaude (Anthropic) · GPT (OpenAI) · Mistral
  • AI orchestrationVercel AI SDK · agent patterns · tool use · streaming
  • DataPostgres · pgvector · object storage
  • Eval & observabilityCustom evals before launch · prompt regression suites · monitoring
  • InfrastructureVercel · Cloudflare · CI on GitHub Actions
  • MobileSwift / SwiftUI for native iOS (Sheed)
// Technical depth

Where the AI is load-bearing, the engineering is not just integration. Custom eval suites on 1,000+ labelled examples before every production release. Prompt regression catches drift on the next API or model update. Agent patterns specialised for the workflow we ship into, not generic chat. Latency budgets enforced end-to-end, cost-per-resolution measured every day. We treat AI like production code, with the unglamorous list that production needs.

// How we engage

Scoped, shipped, handed back.

No mystery procurement, no proposal cycles. You bring a paragraph, we send a written brief with a number on it. Then we ship.

  • ScopeOne signed brief, one ship date. 4–12 weeks per engagement. Quote sent within 48 hours of the call.
  • StaffingEngineering team scoped to the mission and named in the signed brief. From founder-led on the smallest scopes, to founder plus one to three senior contractors mobilised from our network when the surface demands it. Substitution clauses on every contract. No juniors on your project. No agency layer.
  • PricingTypical engagement €50k–€250k. Quoted scope-fixed, not time-and-materials. Invoiced 50 / 50, net 7.
  • Working rhythmDaily PRs into your repo. Weekly written digest. Code is yours from day one, documented, picked up by another engineer with zero friction.
  • Post-launch30-day warranty included. Optional monthly retainer for AI agents in production: evals, monitoring, model updates.
  • Compliance & securityGDPR-ready: EU data residency on request, DPA on file. Model providers (Anthropic, OpenAI) under enterprise contracts when scope requires, with no training on your data. PII minimised at ingestion, redacted in logs. Your repo, your CI, your security gates. Read the full compliance pack →
J. B. · founder4×5

Jerome Borenstejn. Founder.

Twelve years of senior engineering at venture-backed scale-ups: hospital scheduling at Doctolib across 350+ facilities, back-end lead at Kolsquare, CTO and co-founder at Roadstr with 100k+ users. Trained in applied mathematics, EMLYON, ESCP, and as a screenwriter. Founded Manibor in 2025.

Legal identity.

A real company, registered in France, with a real address.

Legal nameMANIBOR SAS
SIREN988 603 908
SIRET (HQ)988 603 908 00018
EU VATFR70 988 603 908
RCSCréteil · 2025B05236
NAF / APE62.01Z — Computer programming
Incorporated27 June 2025
Registered office15 avenue des Mérisiers, 94130 Nogent-sur-Marne, France
HostingVercel Inc. · Walnut, CA

Get in touch. .

Send one paragraph: what you want shipped, or what you want to investigate. We answer within two working days, with a written reply, not a calendar link.