Taste Lab turns “feed my family this week” into a real grocery basket — priced against the actual store shelf. It targets the €100B+ weekly shop, barely 3% online, where the biggest failure is buying too much. And new EU law just made that over-buying a liability.
Built & shipped by one founder — production stack and all
The problem isn't recipes — it's the €100B+ weekly grocery shop, where the dominant failure is over-buying. The EU's food-waste cut is now legally binding, and that waste is concentrated in homes. Every corporate response so far is supply-side: markdowns, surplus apps. Taste Lab is the only one working the demand side — planning plus pantry-state. That turns a binding target into three things: product positioning, non-dilutive grant capital, and a B2B door into retailers.
Horizon Europe CL6 (~€4M/project), EIT Food FAN, EIC Accelerator — a grant track that extends runway without dilution.
The WRI "$1 → $14" waste case doubles as the signed-retailer credibility unlock — the milestone that kills scraper-risk.
A binding regulatory target turns demand-side planning from a nice-to-have into a policy tailwind — opening non-dilutive capital and a B2B door into retailers.
Not a prototype: a chat-first planner, a swipe-trained taste graph, and a grocery rail — 212 of 229 modules live, 56 database tables, 5 apps. Beyond the consumer app sit four internal platforms most pre-seed teams never build — Mise (recipe-generation engine), Ledger (behavioral analytics), Taster (automated chat QA) and Council (an autonomous operating team) — the machine that lets one person ship like a Series-A team.
These aren't side-projects — they're why one person ships like a Series-A team and the production scales at ≈$0 marginal cost. The internal platforms are the capital-efficiency story, not a distraction from it.
Chat-first weekly planner with a swipe-trained taste graph that learns each household's palate — and a chat that renders real cards, forms and approvals instead of walls of text.
Moat: a proprietary preference dataset that compounds every session — can't be bought, only earned.
Point it at any grocery and it returns a verified, photographed catalog — mapping the store's real products onto one clean ingredient model, then generating recipes and studio photos at volume.
Moat: marginal cost ≈ $0 — a grocery becomes a million plates in compute hours, not headcount. The defensible IP.
Behavioral analytics, built in-house — every plan, tap and basket captured, with full replay of any household's journey through the product.
Moat: the company watches its own demand loop in real time — the instrument that prices the take rate and proves retention. (This deck is tracked through it.)
Automated chat QA: synthetic test users with a motivation engine hold real conversations, scored by deterministic gates + judges across 7 rubrics.
Moat: ship fast without breaking safety, allergen and quality guarantees — the work that usually needs a QA team.
A standing autonomous team: four role-agents — Strategy, Delivery, Mise Quality and Engineering — each a persistent Claude session running scheduled duties, supervised through a live Observatory dashboard.
Moat: the org itself becomes software — standing company functions run autonomously, pushing the team-scale-output thesis from building the product into operating the company.
Designed & scoped — ready to build
The household taste-intelligence brain — taste & ingredient ontology, user·recipe·SKU embeddings, acceptance & repeat-cook prediction, substitution and constraint-aware ranking — behind one model registry and inference API.
Moat: answers the question that compounds — given this household, context and the food on hand, what will they safely cook, enjoy and repeat? The transferable, proprietary IP every other app consumes.
The translation from food intent to purchasable reality — canonical ingredient→retailer-SKU graph, product matching & substitution, live price & availability, pantry-aware quantities, and basket optimization across budget, nutrition and preference.
Moat: owns the rail from recipe to checkout — retailer adapters, basket export and the attribution/affiliate economics. Where the monetizable unit, the basket, actually gets built.
Mise generates the data → Palate turns it into intelligence → Taste Lab serves it → Basket takes it to checkout. Ledger, Taster and Council record the outcomes, benchmark the quality, and watch the whole loop — so each app makes the others sharper.
Recipes and chefs are batch outputs, not handcrafted entries — long-running autonomous runs that produce them at volume, each fully verified and styled. Marginal cost ≈ $0; the only ceiling is run time.
The SKU importer + canonicalization engine snap raw store inventory onto one ingredient model — vendor-neutral, allergen-safe, deterministically sourceable.
Today · 4 stores · 20,725 SKUs → any regionEach dish is generated, validated against safety/nutrition/realism gates, then shot by a food-photo swarm. No human in the loop.
Tested capacity · 10,000 verified recipes & photos / dayReaching a million verified, styled recipes and thousands of chef personas is compute hours, not headcount or capital.
Today · 136 chefs · 309 batches · 1.06M stepsReal, published output — swipe the deck, then open the full detail
Point the engine at a store it has never seen and let it run. In a single autonomous day — no human in the loop — it absorbs the inventory, casts a kitchen, writes the menu, and shoots every plate.
The importer + canonicalization engine snap a store's raw inventory onto one canonical ingredient model — vendor-neutral, allergen-safe, deterministically sourceable.
Persona generation gives each chef a distinct voice, palate and ingredient bias; a portrait swarm shoots their likeness. A whole collective, from scratch.
Each dish runs skeleton → enrichment → validation behind safety, nutrition and realism gates — grounded in the very SKUs absorbed at hour zero.
A food-photo swarm styles and shoots each recipe — studio-grade, one image per dish. No set, no stylist, no shoot day.
A grocery the engine had never seen is now a verified, photographed catalogue — in one autonomous run. Marginal cost ≈ $0; the only ceiling is run time.
Taste Lab is built AI-first: agents do the work, running jobs end-to-end through MCP. But nothing reaches a real person on trust alone — every dish passes the same fixed checks, and anything that sticks needs a human's sign-off. The output is verified and sourceable, not AI slop.
⇄ Swipe to see every gate
Five fixed checks, then a person signs off — every dish, before it's ever saved. Hover a step for what it catches. It's why the catalogue is real food you can actually buy and cook, not AI slop.
Agents operate the platform through MCP servers — they run the long autonomous jobs (1.06M step-executions across 309 batches). Built tool-first for machines, with humans supervising, not clicking.
Mise · Kanban · Registry — all agent-drivableA two-phase approval engine gates every durable change — profile, grocery list, budget — and every inbound message passes a prompt-injection sanitizer first. Agents move fast; the gates decide what sticks.
Explicit approval before any durable writeBeyond the gates, Taster runs synthetic users against 7 scoring rubrics, and Zod validates every trust boundary in a TypeScript-strict codebase (no any). Trust is enforced in code, not hoped for.
7 QA rubrics · Zod at every boundaryThe stack underneath
Taste Lab turns "feed my family this week" into a grocery basket priced against real store inventory. The basket is the monetizable unit: commission on referred grocery spend (typically 3–5%) plus a thin premium tier. Spain is the live beachhead — 19M households, a €100B+ grocery market barely 3% online, with Bonpreu's 20,725 products already mapped. Adjust the inputs below.
Revenue/HH = basket × baskets/mo × 12 × grocery commission. Implied valuation applies the scenario's forward-ARR multiple (7× bear · 8× base · 10× bull), cross-checked against a 0.35–0.6× influenced-GMV multiple. A US household is worth ~1.6× the EU one ($110 basket, 2.5×/mo → ~$165/yr) but faces multiples-higher CAC; US lives in the bull case.
The product is live and in the hands of beta households. The pre-seed's first job is to turn this early signal into a proven conversion loop — baskets per household, week-over-week retention, and a take rate evidenced against a real retailer.
Every visit to this page is itself instrumented through Ledger — the same analytics stack that runs the product. Live engagement numbers drop into the tiles above as the beta scales; this slide is where the conversation should start, and where the next milestone is earned.
One playbook, ported country by country — Portugal, Italy, France — each reusing the same engine plus a local catalog. The curve traces the illustrative base-case trajectory, quarter by quarter, as each market comes online; the two financing rounds mark where the ramp is funded.
⇄ Implied base-case valuation · illustrative
The engine run once for the live beachhead — Mercadona / Carrefour / Dia — proving the conversion loop at national scale.
The first port: same machine, a local catalog. The step that proves the playbook travels.
Two of Europe's largest grocery markets, each reusing the engine — the step the curve above is built on.
As of Dec 2025, the full meal-intent-to-checkout loop ships inside ChatGPT via Instacart, and rich chat UI is a documented Apps-SDK pattern — so neither is a differentiator. But incumbents do Product-level inference, not a dish-level taste graph; genuine pantry-state lives only inside a $4,000 fridge; and no verified product combines dish-level taste + software pantry-state + WhatsApp + a canonical-catalog production stack into one assistant. That's the bet.
| Player | Loop → checkout | Canonical catalog | Software pantry-state | Dish-level taste graph | WhatsApp channel | Gen-UI in chat |
|---|---|---|---|---|---|---|
| Taste LabUS + Spain/EU | Planned | Yes | Partial | Shipped | Phase 1 | 10 intents |
| Instacart × ChatGPTUS | Yes | Yes | no | Product-level | no | Yes |
| Kroger + GeminiUS | Rollout | Yes | no | Product-level | no | some |
| Samsung Food + Fridgeglobal, split | redirect | no | Fridge HW | Product-level | no | app UI |
| MercadonIASpain, unofficial | no | scraped | no | no | no | text |
⇄ Swipe to compare all six players
Recreating just the published catalogue — 3,102 developed recipes, each styled and photographed — would cost ~$0.78M of human recipe-dev + food-styling + photography ($100 dev + $150 photo per dish), on top of the engineering floor. And it recurs at ~$0 marginal cost.
104 active chef personas (136 built) and a swipe-trained, dish-level affinity model — a proprietary preference dataset that compounds with every session and can't be bought, only earned. Plus Mise (catalog), Taster (eval loop), Ledger (analytics): a stack that lets a one-person team out-ship incumbents with orders of magnitude more headcount.
Les Echem — founder. Twelve years building digital products as a software engineer, design leader and AI developer, across a Fortune 100, luxury hospitality, Michelin-starred restaurants and food e-commerce. Award-winning design met a lifelong passion for culinary experimentation — and the rare overlap of all three (engineering, design, and the kitchen) is exactly what this product demands. The entire build above — 5 apps, a production stack, 3,102 photographed recipes and a live taste graph — was architected and shipped solo, building the intelligent operating system for the home kitchen.
The capital efficiency isn't the valuation — it's the signal. A founder who turns a weekend-away budget into Series-A-scale output is exactly the bet a pre-seed cheque is making: rare execution velocity, before the market has priced it in.
A de-risked, eight-person-year build with a production stack most teams reach only after a Series A — and a working product. The pre-seed funds one thing: turning shipped IP into proven demand.
Use of funds: prove Spain's conversion loop, sign a first retailer, evidence the commission rate, and ship software pantry-state — on the shipped IP, the taste model and the production stack.
Then, on evidence — not roadmap or geography:
Once the loop is proven, across Mercadona / Carrefour / Dia.
After Portugal proves the playbook ports — funding Italy and France.
This is a private look at what one founder has built. May we have your name?