What does it cost to work with you?

It depends. I know "it depends" isn't satisfying, but it's honest: audit, project rescue and fractional CTO have different sizes and timelines. I give you the price in our first call, after I understand what you actually need. What's certain: it's written down before we start, and it's fixed. No quotes that keep growing, no surprises halfway through.

How quickly will I see real results?

Audit: written report in your hands in 2-3 days. Project rescue: first changes live within 1-2 weeks, completion in 2-4 weeks. Fractional CTO: from the first week you have a written technical roadmap and the first decisions made — you don't wait a month to see what's changing.

Do you only work remote or also in person?

Remote, mostly. I come on-site when it actually matters — kick-off, strategic workshops, on-site audits. Veneto, Lombardia and Emilia covered easily; other regions with travel expenses. No theater of mandatory presence.

Do you also work full-time somewhere else?

Yes — I'm a Tech Lead at an enterprise company, and that's exactly where the technical level I bring to your project comes from. That's why I work with very few clients at a time, with clear scope: audit, time-boxed rescue, light monthly retainer. My availability is limited by design, not despite everything. If you need someone full-time every day, you need an in-house dev — and I'll be the first to tell you.

What happens if after the audit we decide not to continue?

Nothing bad. The report is yours, take it home — it contains everything another dev or team needs to act. The initial 30-minute call is free, the audit has a fixed price agreed up front. Zero ambiguity, no hidden clauses.

Do you only work with startups or also with SMBs?

Both. Founders validating an idea, CTOs inheriting a codebase, SMBs with stuck internal software or a developer who stopped replying to emails. I bring value in all these cases — it's not the company size that matters, it's whether you actually need a senior technical reference.

Can I have a fractional CTO while I already have internal developers?

Yes, and it's one of the most common cases. You have devs (internal or external) but no one makes technical decisions or coordinates. That's the role I play: I don't replace the team, I lead it. Direct Slack, weekly call, no disappearances. If there's a promising senior dev, my goal is also to help them grow until they can replace me.

What stacks are you specialized in?

Next.js, TypeScript, Node.js, Python, FastAPI, PostgreSQL, AWS, Vercel — that's where I'm fastest. But fractional CTO isn't about the stack: it's about picking the right one for your stage and executing. I also work with inherited codebases on other stacks — actually, that happens often. If I think the stack choice is wrong, I'll tell you.

How does invoicing work?

Italian invoice with VAT. Audit and project rescue: 50% upfront, 50% on delivery. Fractional CTO: monthly invoice, renewable quarterly contract (no annual lock-ins). Payment by bank transfer. Clean, invoiced, traceable.

AI & modernization

Claude Code, for real: my senior experience with Anthropic's AI agent

I've used Claude Code on real work — refactors, migrations, even running this very site. What a terminal AI agent actually does well, where a senior's judgment is still essential, and how I use it with clients.

May 27, 20267 min read

Everyone talks about AI coding agents, and very little of it is clear. So let me say it as someone who actually works with them: Claude Code— Anthropic's terminal-based AI agent — is one of the tools that has changed how I work the most recently. Not because of the buzz, but for a concrete reason: it does the grunt work across entire codebases, fast, leaving me the decisions that matter. This very site, for one, is largely run that way.

What Claude Code actually is

It's not a chatbot you copy-paste code into. It's an agent that lives in your terminal (or your IDE) and works directly on your project: it reads files, understands the structure, runs commands, and applies changes across dozens or hundreds of files in one go. The difference from "AI that suggests" is enormous: here the AI acts on the codebase, under your control.

Where it truly makes a difference

Refactors and migrations. Swapping a library or upgrading a framework across an entire codebase: hours instead of weeks, with consistent changes file by file.
Audits of inherited codebases. Mapping a project you don't know, understanding architecture and dependencies, finding the fragile spots — in a fraction of the time.
The boring work. Missing tests, documentation, dead code, small fixes repeated across many files. It's exactly where an AI agent excels and where a senior would rather not burn hours.

Where a senior's judgment is still essential

Here's the part AI marketing glosses over: Claude Code executes brilliantly, but it doesn't decide. The choices that make the difference remain human.

Architecture. Which structure can sustain growth, where to draw the boundaries, which trade-off to accept. The agent offers you plausible paths; knowing which one is right for your stage is a different craft.
Validation. Code that compiles is not code that's correct. For anything that touches production, human review isn't optional — it's the real job.
Direction. What to do and what to leave alone: the AI doesn't know your business constraints. Without that context, it very efficiently optimizes the wrong thing.

The summary I keep repeating: AI accelerates execution, not the taking of responsibility.The value isn't "AI that does everything", it's a senior who, with AI, does the same work much faster, without lowering the bar.

How I use it with clients

On projects — technical audits, unblocking stalled products, modernizing legacy systems — Claude Code is the multiplier behind the scenes. I decide architecture and strategy, the agent does the heavy lifting, and I review every step that goes to production. For whoever hires me, the result is simple: shorter timelines, with the same rigor. An audit that used to take a week wraps up in a couple of days. This is the practical account, from the inside; if you'd rather see how all of this applies to modernizing legacy software at the business and project level — what really speeds up and where a human decision is still needed — I laid that out there, from the point of view of whoever has to decide.

And it's not just "I work faster": I also get internal teams working well with these tools — which tasks to delegate, with which guardrails, where to put human review. That's my AI enablement service: hands-on work or training for your team.

The bottom line

Claude Code isn't hype and it isn't magic: it's a serious tool that, in senior hands, compresses the time real code work takes. Anyone who thinks it eliminates the need for an experienced engineer has it wrong; anyone who uses it to amplify a senior has it exactly right.

Frequently asked questions

Is Claude Code useful for a senior developer, or only for beginners?

It's genuinely useful, just in a different way. For me as a senior it's not a tool for learning to code — it's an execution multiplier. I let the agent do the grunt work across entire codebases — refactors, migrations, the boring stuff — while I keep the decisions that matter. That's the whole point: a senior who, with AI, does the same work much faster without lowering the bar.

What does Claude Code actually speed up, and what doesn't it?

It speeds up execution: refactors and migrations across the whole codebase, audits of inherited projects, missing tests, documentation, dead code, repeated fixes across many files. It doesn't speed up the taking of responsibility: architecture, validating what ships to production, and direction stay human calls. The agent executes brilliantly, but it doesn't decide.

Can I trust the code an AI agent produces?

You can trust the process, not any single output sight unseen. Code that compiles isn't code that's correct: for anything touching production, human review isn't optional for me — it's the real job. That's exactly how I use it — the agent does the heavy lifting, I review every step that goes live.

How is it different from an AI assistant that suggests code in the editor?

The difference is enormous. An assistant suggests, you copy and paste. Claude Code is an agent that lives in your terminal or IDE and acts directly on the project: it reads files, understands the structure, runs commands, and applies changes across dozens or hundreds of files in one go — under your control.

Do you use Claude Code on client projects too?

Yes, it's the multiplier behind the scenes. I decide architecture and strategy, the agent does the heavy lifting, and I review everything that ships to production. For whoever hires me the result is simple: shorter timelines with the same rigor. I also get internal teams working well with these tools — which tasks to delegate, with which guardrails, where to put human review.

Have a codebase to modernize, or a team to bring onto AI the right way? Let's talk for 30 minutes, free. I reply within 24 hours.