You use Claude every day.
The output is almost good.
I teach you how to fix the almost.

I'm Alberto Granzotto. Software engineer.

I had the same problem you have. Claude gives you something that looks right, but it isn't. And you can't figure out why.

I got obsessed with solving this. I applied the same discipline I use for software engineering, and what I found is that it's a process problem, not a Claude problem.

I built a method (and tools) I use every day.

Now I want to teach it to you. Hands-on, 1-on-1, no theory.

Get the most out of Claude Cowork in 90 minutes.

Module 1
Better input, better output.

  1. We set up Claude Cowork together.

    Your project folder, your CLAUDE.md, your context. You leave with a workspace that's ready.

  2. You learn the protocol.

    Brainstorm, plan, verify, execute. A step-by-step method instead of one-shotting and hoping.

  3. You learn what the AI actually reads.

    PDFs, spreadsheets, folders. How to feed data, verify ingestion, catch hallucinations.

  4. You use the right model for the right task.

    Sonnet for reading, Opus for thinking. You stop burning tokens on the wrong things.

  5. You leave with a cheat sheet.

    The protocol on one side, the common mistakes on the other.

And here's what comes next.

Module 2
Connect your tools, reuse your methods.

You package how you work into skills Claude follows every time. Your own methods, but also stuff other people figured out: negotiation tactics from a book, a copywriting framework, an analysis process. You install them once and Claude uses them on everything. Claude also connects to your email, calendar, and docs.

Learn more
  1. Connect your email, calendar, and Drive. Stop copy-pasting context into every session. Claude reads your actual work.
  2. Build skills from how you work. That analysis you do every week? Teach Claude once. Done in minutes after that.
  3. Turn a method into a reusable skill. Something from a book, a process from a doc. Claude uses it on every project from now on.
  4. Set up project files. Open a project and Claude already knows the background. No warmup.

Module 3
Everything in its right place.

Claude builds your personal knowledge vault in Obsidian. Notes, projects, decisions: structured, connected, searchable. Claude recalls what you said, what you decided, what matters to you. You're not a stranger to Claude anymore.

Learn more
  1. Claude builds your vault from a conversation. You talk, Claude structures. An hour of conversation, months of organizing.
  2. Notes linked in a graph. Ask about a client and Claude pulls up the decision you made six months ago and why.
  3. Query your own knowledge. "What did I decide about pricing last quarter?" You get the answer, not ten documents to sift through.
  4. Claude flags what's off. Last month's decision contradicts this week's plan? Claude catches it.

Module 4
It works while you sleep.

You close your laptop. The system keeps going. It processes your day, monitors your sources, maintains your vault. By morning you know what happened, what matters, and what to do next.

Learn more
  1. Your conversations, processed. That thing you mentioned at 3pm? Captured, filed, connected to the right project.
  2. Links you saved, read. Five articles stashed during the day. By morning they're summarized and linked to what you're working on.
  3. Your sources, filtered. Claude checks Reddit, RSS, newsletters: pulls out what's relevant to your projects.
  4. Your vault, maintained. Broken links, stale tasks, forgotten threads: cleaned up overnight.
  5. Your morning brief. You open your laptop and know the three things that matter today.

Module 1. 90 minutes, 1-on-1.

You get the method, the cheat sheet, and a different way of working with AI.

€250

Interested? Want more information?

No commitment. Just send me an email and I'll get back to you.

Email me

Who I am.

Alberto Granzotto

I'm Alberto Granzotto. Software engineer for 20+ years.

I studied in Padova, Milano, and Southampton, where I worked on semantic web and linked data. My first company was a startup called Urlist, a bookmarking service, in 2012. In 2015 I was a core developer for ascribe, a platform for artists to own digital art the way you own Bitcoin. Nobody was talking about blockchain yet. That pulled me into smart contracts and NFT tools for artists, some of whose work ended up at Centre Pompidou and other museums. I also built a legal framework for DAOs in Estonia that actually runs real companies.

I care about privacy, decentralization, and open source, and I do my best to balance that with the reality of private LLMs.

These days I'm building two AI products: tellfranko.com and mealo.fit. And I teach what I picked up along the way.

Based in Berlin. Registered freelance software engineer in Germany.

granzotto.net · LinkedIn

FAQ

Is it remote or in person?

Remote via FaceTime, Zoom, or Google Meet. If you're in Berlin, you can come to my studio or I come to you.

What do I need to prepare?

Nothing. Bring your laptop.

What software do we use?

Claude Cowork. You need at least a Pro subscription. If you don't have one, I can give you a one-week invite.

Do I need technical experience?

No. This is not about coding. It's about how you work with AI, regardless of your job.

Can you teach me how to build an app with AI?

Yes. I teach vibe coding and how to add guardrails so it doesn't break everything.

Can you automate my workflow?

Yes. If you're doing the same thing every day, I can probably make the AI do it for you.

What about OpenClaw?

It's powerful. It's also easy to lose control: leak secrets, run things you didn't intend, give the AI too much freedom. I teach a different approach: still highly automated, but you stay in control. You know what runs, what gets reviewed, and what stays private.

What else can you do?

I build autonomous sub-agents, custom MCP servers, multi-agent coordination systems. If you need something deeper, let's talk.

I'm not here to tell you you're absolutely right.
The AI already does that.