# codecook.ch: Full Content for LLMs > Last updated: 2026-08-13 > Domain: https://codecook.ch > Language: de-CH (Swiss German / German), with a full English version at /en/ ## Canonical service pages All services live on a single offering page per language; the deeper per-service pages were consolidated into it on 2026-08-13. - Offering (all services): https://codecook.ch/angebot/ (DE) · https://codecook.ch/en/offering/ (EN) - AI workshops for companies: https://codecook.ch/angebot/#workshop (DE) · https://codecook.ch/en/offering/#workshop (EN) - MVP Coding Day: https://codecook.ch/angebot/#mvp (DE) · https://codecook.ch/en/offering/#mvp (EN) - Technical second opinion: https://codecook.ch/angebot/#zweitmeinung (DE) · https://codecook.ch/en/offering/#second-opinion (EN) - Long-term development after the MVP: https://codecook.ch/angebot/#weiterbauen (DE) · https://codecook.ch/en/offering/#build-on (EN) --- ## Overview codecook.ch is the digital home for AI, prototyping, and workshops around building with digital ideas. Based in Zürich, Switzerland, we work on AI and tech projects, formats, and experiments. Our conviction: AI is not hype. In the flood of information, we show how to properly contextualize the topics. We demonstrate how coding tools and good concepts produce concrete results, hands-on, no bla bla. Although we come from IT, we consciously write zero lines of code ourselves here, to showcase the possibilities and limits of vibe coding with AI. --- ## Team ### Dima (Dimitry Uzilevsky): born 1986 Where business meets IT: creative, curious, hands-on. **Education:** Fachinformatiker Anwendungsentwicklung / Wirtschaftsinformatik / MBA **Career:** - Gate Gourmet: Head of Supply Chain - Lufthansa (LSG Sky Chefs): Transformation Controller, Deputy Head of Finance - Group Controller at the purchasing JV Deutsche Telekom-Orange - Internal Audit / Investment Controlling - Reporting Dashboards in IT Key Account Sales **Today:** Back to IT roots: Startup Trafa.App, AI projects. **Past affiliations:** Telekom, Orange, Lufthansa, Gate Gourmet, Commerzbank. ### Fönsi (Alfonso Papa): born 1970 Entrepreneur, Business Coach, CRM & Asset Management. **Education:** Bankkaufmann / Betriebsökonom FH / eidg. dipl. Finanzanalytiker, CEFA, CAIA **Career:** - alpa solutions: Strategy, CRM and Organizational Development - Baloise Asset Management: Tribe Lead 3rd Party Clients, Sales and Business Development - Managing Director Switzerland at NN Investment Partners - 26 years in leadership roles as Information Manager, Business Developer and Country Head - 37 years of client-oriented experience in the financial industry **Today:** Entrepreneur and Business Coach, supporting leaders and teams with heart, experience, and digital affinity. **Past affiliations:** Baloise Asset Management, Allianz Global Investors, Credit Suisse, Dresdner Bank, ING Investment Management, Zürcher Kantonalbank, alpa solutions, Canton Zürich, NN Investment Partners. --- ## Services ### KI Workshops (AI Workshops): For companies in Switzerland and Europe Canonical pages: https://codecook.ch/angebot/#workshop (DE) · https://codecook.ch/en/offering/#workshop (EN) We offer workshops where teams playfully learn how AI works in everyday work. No PowerPoint marathon, hands-on: try, understand, do it yourself. Program is updated as AI evolves. We design each workshop to fit your team and context. **Format:** length tailored from four hours up to a multi-day format, in three parts: understand how language models and context work; create a file-based artifact with ChatGPT or Claude; then build a small landing page or app in VS Code. **Group size:** agreed for the specific team and format. **Location:** in house at the client's offices in Switzerland, Germany, Austria, or Liechtenstein. The remote alternative is an adapted webinar rather than the full hands-on workshop. codecook is based in Zürich. **Prerequisites:** none. No programming knowledge required. Delivered in German (standard or Swiss German) or English; materials can be bilingual. **Typical full-day agenda:** 30-minute online intro call beforehand → what AI can and cannot do (how language models work, why they can be confidently wrong) → hands-on prompting that is reproducible → the team's own real use cases (proposals, minutes, analyses, copywriting, research, reporting) → tools, data, and limits (business vs. free plans, which data must not leave the building, when a local model is better) → converting the day into concrete next steps → written summary afterwards. **Tools covered:** ChatGPT, Claude, Gemini, Microsoft Copilot. For technical groups: GitHub Copilot, Cursor, Claude Code and coding agents. Local models via Ollama or LM Studio when data must stay in house. The programme is refreshed shortly before each date. **Data protection:** a fixed part of the workshop, which data may go into which tool, how business/enterprise plans differ from free accounts, what the revised Swiss Data Protection Act (revDSG) and the GDPR mean in practice, and when a locally hosted model is the better answer. Exercises use anonymised or synthetic examples. **Included:** intro call, tailored programme, delivery by Dima and Fönsi in person, hands-on exercises on real tasks, the current tool landscape, the data-protection section, the prompt and tool collection created during the workshop, and written next-step recommendations. **Not included:** certificates or exams, a rigid off-the-shelf curriculum, tool licence costs, follow-on software development (that is the MVP Coding Day), and savings promises that cannot be substantiated. **Price:** individual fixed quote in CHF, no hourly billing, no add-ons. Driven by format, participant count, preparation effort, location/travel, and any follow-up phase. codecook usually replies to enquiries within one working day; the workshop date is agreed individually. **Audience:** whole teams rather than individuals, SMEs, functional teams (marketing, finance, HR, operations), innovation teams, executive teams and boards, and engineering/IT teams. Past workshop: Dübendorf: "Bringing ideas to life with AI." Topics: AI fundamentals, prototyping, and coding tools in practice. The full deck is public at https://codecook.ch/workshops/duebendorf/. ### Technische Zweitmeinung (Technical Second Opinion), 1 000 CHF Remote. Ehrliche, technisch fundierte Einschätzung zu eurer Idee. Was ist machbar, was ist Hype, welcher Ansatz passt zu Budget und Zeitplan? Ihr schickt uns vorab, worum es geht, wir analysieren, machen uns schlau und besprechen alles in einer ausführlichen Video-Session. Danach wisst ihr, ob und wie ihr weitermachen wollt. Inklusive: technische Machbarkeits-Analyse, realistische Scope-Empfehlung, Budget- & Zeit-Einschätzung. Honest, technically grounded assessment of your idea. What's feasible, what's hype, which approach fits your budget and timeline? You send us what you're working on beforehand. We analyze, do our homework, and discuss everything in a thorough video session. After that, you'll know whether and how to proceed. Includes: technical feasibility analysis, realistic scope recommendation, budget & timeline estimate. ### MVP Coding Day (AI Prototyping Workshop for Startups), 6 000 CHF + 8.1% VAT Vor Ort in Pfäffikon ZH. Ein Tag (9:00: open end), ein Prototyp. Wir entwickeln gemeinsam eure App oder Website, ihr sitzt daneben, entscheidet und testet. Am Abend steht der testbare Kernablauf; Deployment, Source Code im passend gewählten Stack, Git-Repository und Dokumentation folgen spätestens zwei Tage danach. Maximal zwei Leute von eurer Seite. Verpflegung ist inklusive. On site in Pfäffikon ZH. One day (9:00: open end), one prototype. We build your app or website together. You sit beside us, make decisions, and test. By evening the testable core flow exists; deployment, source code in the appropriate stack, Git repository, and documentation follow no later than two days afterwards. Maximum two people from your side. Food and drinks are included. Canonical overview: https://codecook.ch/angebot/#mvp (DE) · https://codecook.ch/en/offering/#mvp (EN) Detailed MVP pages: https://codecook.ch/mvp-workshop/ (DE) · https://codecook.ch/en/mvp-workshop/ (EN) The MVP Coding Day is designed for startups, founders, and pre-seed teams with a new digital product idea, plus innovation teams inside SMEs. It focuses on one testable core user flow for early feedback and investor conversations; it is not a promise of a production-ready, scaled product. All three services are compared on one overview: https://codecook.ch/angebot/ (German) and https://codecook.ch/en/offering/ (English). **Outcome:** the smallest functioning market test: one end-to-end flow that a real test user can complete independently with real code and logic through a testable web link or iOS build. Example flows include uploading a document and receiving an AI analysis, entering customer data and generating a quote as a PDF, or submitting a request and notifying the right specialist. The actual flow is agreed during scoping. **Sequence:** (1) remote scoping call of one to three hours, cutting the idea down to the one flow testable in a day and clarifying audience, platform, and success criterion; (2) preparation, repository, project skeleton, data-model sketch, and deployment path set up in advance so the shared day starts with building rather than setup; (3) the on-site coding day one to two weeks after scoping, morning for the skeleton (screens, navigation, data model, the one interaction that matters) and afternoon for realistic data and sharpening; (4) deployment no later than two days afterwards to the client's domain or as an iOS build via TestFlight; (5) handover of source code, Git repository, documentation, and an honest estimate of what building it out would cost; (6) bug fixing for two weeks after the coding day, covering fixes only and no new features. **Not included:** a finished scalable product, App Store release and review handling, a full design system or branding, legal texts (terms, privacy policy), ongoing operations/support/maintenance, and third-party costs such as domain, hosting, or an Apple Developer account. **Compared to alternatives:** an agency delivers more functionality after weeks and a typically five-figure budget; a freelancer is usually hourly with an open-ended scope. The MVP Coding Day instead answers the prior question, does the idea hold at all, for a fixed CHF 6,000 plus 8.1% Swiss VAT within a few weeks. If the scoping call shows the idea is too big for one day, codecook says so before any money changes hands and proposes either a smaller slice or the CHF 1,000 technical second opinion. A client who decides against the build after scoping is billed CHF 500. **Code ownership:** the client receives the complete Git repository with all source code and documentation. No lock-in, no proprietary platform, no recurring fee. Any development team can build on it. For founders outside Switzerland, scope and preparation happen remotely, while the collaborative Coding Day remains on site in Pfäffikon ZH. codecook serves teams in Switzerland, Germany, Austria, and across Europe. Travel and accommodation are not part of the fixed price. **Project examples:** - **Airhockey Duo**: Coded in a few hours over beers and burgers. Available on the Apple App Store. (Project 01) - **Trafa.App**: A full-scale startup product, not a one-day MVP. A "Reddit for Maps" with shared maps, pins, and interest-based communities. Built with developers from Ukraine, representing approximately 1,300 hours of development work. (Project 02) - **copilot24.de**: AI copilot for occupational physicians and workplace safety, with clear domain guardrails and domain-specific knowledge. Essentially ChatGPT with workplace safety rules as guardrails. (Project 03) --- ### Weiterentwicklung nach dem MVP (Long-term development and support) Canonical pages: https://codecook.ch/angebot/#weiterbauen (DE) · https://codecook.ch/en/offering/#build-on (EN) An MVP Coding Day answers one question: does the idea hold up? When the answer is yes, what follows is not another single day but weeks and months of development, operation, and support. For that work codecook continues with an established circle of developers in Eastern Europe, mostly from Ukraine, that it has worked with for years. There are three valid next steps after handover: keep the prototype and test it first; hand the complete code, Git repository, and documentation to an internal team or another agency; or continue with a team assembled through codecook's network. **Track record:** Trafa.App: a full-scale startup product with roughly 1,300 hours of development work, was built with this team, not in a one-day format. **Team selection:** codecook assembles the right team for the project from a long-standing partner network and scales it up or down with the need. **Two ways to run it.** (A) *Direct*: the client continues the project directly with the developers. All of them speak English, so no intermediary is required. (B) *Link and quality gate*. Codecook stays involved, handles the technical coordination in the developers' own language, and reviews the quality of what is delivered, so the client keeps speaking only to codecook. **How it is set up:** this is not anonymous outsourcing or a handover to unknown freelancers. If the client wants them to, Dima and Fönsi remain the contacts, hold the technical lead, and review what is delivered. The time zone matches Switzerland. Under option B billing can run in Swiss francs through codecook, so the client keeps a single contract and a single point of contact. **Scope:** team selection, continued feature development after the MVP, ongoing operation, maintenance, and support. **Price:** quoted per engagement rather than published. Effort, team size, and hourly rate are agreed transparently before work starts. This is the one part of the codecook offering that is not a fixed-price one-off, because sustained development cannot honestly be priced as one. **No lock-in:** the client owns the code and the Git repository throughout and can continue with any other development team at any time. --- ## Tutorials & Tool Recommendations ### LLM Providers (current overview) Our current view on which models we track for chat and coding. This is an assessment, not a complete market overview. | Provider | Chatbot | Coding | |----------|---------|--------| | OpenAI | ChatGPT | Codex | | Anthropic | Claude | Claude Code | | Google | Gemini | Gemini Code | | Alibaba / Qwen | Qwen Chat | Qwen Coder | | Mistral | Le Chat | Devstral | | Meta | Meta AI | Llama | | Kimi | Kimi | Kimi Code | | DeepSeek | DeepSeek | DeepSeek Coder | | Grok | Grok | Grok Code | | GLM / z.ai | GLM | GLM Coding | ### Coding Tool Setup Our recommended setup: VS Code is where Codex, GitHub Copilot, Claude, and Qwen run in one place with files, Git, and terminal. Key tools: - **Google Antigravity**: Agent-first IDE from Google. Runs as a native development environment on your machine, not in the browser. - **GitHub Copilot**: Currently the best price-performance ratio for many. Students often get it free. Compare multiple models directly in the editor. - **Claude (Claude Code)**: Strong in chat for structure, explanations, and longer coding tasks. - **Qwen Code**: Best via the VS Code plugin to try Qwen directly in the editor. - **Codex**: If OpenAI, use Codex instead of ChatGPT for coding. Strong for agentic work directly in the tool and editor. Prompt tips: Describe goal, user, platform, and desired result as concretely as possible. Best to prompt and code in English currently. German works too but usually slightly worse, most models are primarily trained on English and the programming world's language is English. For beginners: Start with a small project rather than a huge idea. Let the model build a simple file and folder structure first. Ask the model to briefly explain each step and only make small changes at a time. Good first projects: Small landing page, To-do app, Notes app, Simple data/CSV viewer. Good working method: First let it plan, then implement step by step. Briefly test after each step. Copy errors directly back into the tool and ask for a concrete fix. Cross-platform mobile (iOS + Android): React Native Framework, Expo Runtime. Cross-platform desktop (macOS + Windows): Tauri Framework. macOS native: Swift Framework, Xcode Project. Always needed: Git/GitHub for version control, a terminal, a good editor like VS Code, a clear README with goal and setup. ### Local LLMs Everything runs locally on your machine. Slower than cloud AI providers, but very secure and costs only electricity, no subscriptions. Almost all local LLM programs are based on `llama.cpp`, a key open-source library. Tools: - **Ollama**: Very simple `llama.cpp` wrapper. ChatGPT-style for local models. - **LM Studio**: Slightly more complex but still simple wrapper with more configuration options. - **llama.cpp**: For pros who want to integrate local LLMs into their own tools or work directly with the command line. - **Hugging Face**: Database for current and distilled models. Distilled models: Smaller, trimmed versions of large models. Faster, cheaper, and often good enough for many practical tasks. Typical sizes: Simple (3B: 7B, ~2–5 GB), Medium (8B, 14B, ~5–10 GB), Advanced (32B+, 20 GB+). Rule of thumb: 16 GB RAM or VRAM is the absolute minimum. More memory usually means larger models, more context, and smoother work. Requirements: Computer with at least 16 GB RAM or a strong Nvidia GPU. The more memory, the better. Key terms: 4B, 6B, 8B etc. roughly indicate the number of parameters. More parameters = more knowledge, nuance, and complexity, but also more memory and compute needed. GGUF is the standard format for many local `llama.cpp` setups, usually the best choice on Windows. MLX is especially interesting for Apple Silicon and local models on Mac. Best local coding model: Qwen3 Coder Next (45–85 GB). Other models we use locally for coding: Qwen, Ministral, GLM. For image/video processing: Qwen and SmolVLM. ### Hardware Setup Currently, Macs are the best and simplest setup for local coding with AI, mainly because of Unified Memory: CPU, GPU, and Neural Engine share the same memory pool. This often makes local models on Mac more straightforward and efficient. For professional local LLM tasks: Mac Studio with at least 32 GB memory. For Windows-specific development: If you're working with Unreal Engine, for PlayStation, or generally Windows-heavy development, a Windows machine is the better choice. Our devices: - MacBook Pro (Apple M Chip) with 16 GB RAM - Mac Studio (Apple M Chip) with 32 GB RAM - Windows PC with RTX 4070 Super and 64 GB RAM ### Claws (AI Agents) Claws are personal AI assistants or agent runtimes that don't just chat. They take on real tasks: web, apps, files, terminal, and entire small workflows. Key players: - **OpenClaw**: The most important open-source Claw at the moment. Originally started by Peter Steinberger. Local, flexible, and clear in the agent/assistant space. - **Kimi Claw**: Cloud version around OpenClaw. One-click setup, 24/7 online, and significantly easier for beginners. - **ChatGPT agent**: OpenAI's direction for real computer-use and web tasks. Can browse, use files, fill forms, and execute actions. - **Claude Computer Use**: Anthropic's computer-use approach. Can see screens and directly control mouse and keyboard. What's happening: Established providers are releasing their own personal assistants or programs that can directly control the computer. The direction is clear: less chat, more real execution. Names to know: OpenAI (ChatGPT agent, formerly Operator), Anthropic (Claude Computer Use), Google (Project Mariner), Kimi (Kimi Claw, Kimi Agent Swarm). Security warning: These tools often get extensive access: browser, files, terminal, logins, clipboard, or entire apps. Prompt injections are a real problem, a website or file can slip hidden instructions to the agent. Running such a tool with broad permissions is essentially voluntarily installing a trojan in the worst case. Therefore: only test on machines with little sensitive data, grant few permissions, treat skills and plugins like real code, and never blindly confirm everything. --- ## Location & Contact Based in Zürich, Switzerland. We serve startups and companies in Switzerland, Germany, Austria, and across Europe. - **AI workshops:** from four hours to multi-day, on site at the client's offices in Switzerland, Germany, Austria, and Liechtenstein. A fully remote workshop is not offered; the remote alternative is a webinar in an adapted form. - **MVP Coding Day:** on site in Pfäffikon ZH; scoping and handover remote. - **Technical second opinion:** fully remote across Switzerland and Europe. - **Languages:** German (standard German or Swiss German) and English. - **Long-term development after an MVP:** codecook assembles a partner developer team in Eastern Europe, mostly from Ukraine. Either the client works with them directly in English, or codecook stays on as the link and quality gate. Remote, same time zone. - **Commercial model:** the three core formats are one-off engagements at a fixed price or fixed quote, no retainer, no minimum term, no subscription. Long-term development after an MVP is quoted separately, with effort and rate agreed before work starts. ### Clients outside Switzerland Billing is in Swiss francs at a fixed price regardless of the client's country. For teams in Germany, Austria, and Liechtenstein the split is simple: the technical second opinion is remote; the AI workshop, from four hours to multi-day, runs on site at the client's offices, in which case the client additionally covers travel costs, hotel, and meals for the codecook team, while the remote alternative is a webinar in an adapted form; and the MVP Coding Day is exclusively on site in Pfäffikon ZH, with scoping and handover remote. Travel and accommodation are not part of the fixed price. Exchange rate, bank fees, and the VAT treatment of cross-border services are for the client's own accounting to determine. GDPR remains the governing framework for EU clients; none of this substitutes for legal advice. - Website: https://codecook.ch - LinkedIn Dima: https://www.linkedin.com/in/uzilevsky/ - LinkedIn Fönsi: https://www.linkedin.com/in/alfonso-papa/ - Email: hello@codecook.ch --- ## Common questions answered on this site - **What does an AI workshop for a company in Switzerland cost?** codecook quotes the workshop individually; it runs from four hours up to a multi-day format, and the price depends on duration, participant count, preparation effort, and location or travel. A fixed quote in CHF follows a short intro call. The remote alternative is an adapted webinar. - **What does it cost to have an MVP built?** The MVP Coding Day is CHF 6,000 fixed plus 8.1% Swiss VAT, covering the scoping call of one to three hours, preparation, the on-site day one to two weeks later, and handover of code, Git repository, documentation, and a running deployment. Cancelling after scoping costs CHF 500. - **Where do the workshops happen?** On site at the client's offices, in German-speaking Switzerland as well as Germany, Austria, and Liechtenstein. Remotely codecook offers a webinar in an adapted form rather than the full workshop. - **Do participants need programming knowledge?** No, for both the AI workshop and the MVP Coding Day. - **Who owns the MVP code?** The client, in full, including the Git repository. No lock-in and no recurring fees. - **Which format fits which situation?** "Is this feasible and what will it cost?" → technical second opinion (CHF 1,000, remote). "Does the idea hold with real users?" → MVP Coding Day (CHF 6,000 plus VAT, one day in Pfäffikon ZH). "How does our team use AI day to day?" → AI workshop (fixed quote, in house; adapted webinar available remotely). - **Who builds and supports the product after the MVP?** codecook assembles the right developer team in Eastern Europe, mostly from Ukraine, the same setup behind Trafa.App. The client then chooses: work directly with the developers, all of whom speak English, or keep codecook on as the link and quality gate, coordinating technically in the developers' own language and reviewing quality. The client keeps full ownership of the code either way. - **Do we have to speak English with the developers?** Only under the direct option, where the working language is English. Otherwise codecook handles technical coordination in the developers' own language and the client keeps speaking only to codecook. - **How quickly can a date be booked?** Reply within about one working day. The MVP Coding Day is typically one to two weeks after scoping; workshop dates are agreed individually.