Sketched illustration of a woman walking, drawn with fine construction lines, under review
reasoning: plausible
structure: unverified
conclusion: pending review
draft_001.pngconfidence 74%2 flags · human review
// web3 & ai legal counsel · swiss & eu law

Before you trust AI, call the human.

You'll ask ChatGPT, Claude or Perplexity first, and you should. However, polished legal output isn't the same as correct and suited output.

  • structuring approach
  • contractual language
  • regulatory assumptions
flagged for human review
// a human in the loop matters
[model]AI can propose a structure, suggest contractual language, and produce a remarkably polished analysis.
[warning]It can also miss the point, overlook a decisive fact, or deliver polished gibberish with complete confidence.
[kleczewski]The goal isn't to duplicate what AI does well; it's to add judgment where judgment counts.
// real experience matters
  • Five years almost exclusively on Web3 projects, as an independent advisor and inside two Swiss fintech-focused law firms.
  • Both sides of the table: I've drafted as the lawyer, and lived with it as the co-founder of a DeFi project.
  • Countless conferences around the world, many of them as a speaker, always with an ear to how builders actually think about the space and their products.
  • I've watched the space move through several cycles, and seen firsthand what tends to backfire.
// your code matters
  • Simplifying your product for users is fine. Simplifying it for a regulatory read is not.
  • What a regulator cares about most, admin keys, their specific powers, and whether or not these affect user funds, their custody, and flows, often hides behind a reassuring one-liner.
  • So I read the code directly, for real, line by line where it matters.
  • Your GitHub beats your whitepaper as a source of truth, every time.
// model.output vs kleczewski.review
– what the model gives you + what I add, after human review
–

Identifies your structuring options.

+

Helps you choose the one that actually fits your product and investors.

–

Suggests edits to an agreement.

+

Determines whether those edits protect your interests in the real world.

–

Gives you an answer, stated with confidence.

+

Challenges its assumptions and makes sure you can safely act on it.

I help you articulate the right description of your product for legal purposes, identify the trigger items, and figure out the way forward.

Based on the constraints coming from your product, team and investors, I help identify a lean and sufficiently flexible setup. The fewer entities, the better; some risks genuinely need isolating, other additions are trendy overkill.

Handshake deals with nothing in writing, an AI-drafted agreement you're unsure about, or an old template you'd like to reuse: I check which clauses still fit, which need adjusting, and what's missing.

Where advice from another jurisdiction is needed, I coordinate with local counsel and brief them properly, so the analysis stays consistent and you're not paying twice to explain the same system.

Many tasks can be estimated fairly well in terms of time, and AI now helps sharpen those estimates further. Wherever possible, I work on flat fees: I know what I'm getting, you know what you're paying, no debates, no uncertainty.

When a task is genuinely open-ended, ongoing advisory work, or depends on facts that haven't surfaced yet, I'll sometimes suggest an hourly rate instead. It's not my favorite approach either: nobody enjoys logging every hour spent, on either side of the invoice.

This isn't about billing you for basic work a model could already do, and it isn't about ego invoicing at top-tier rates either. You'll find me somewhere in between. My expertise beyond purely legal matters is reflected in the fee, so I won't be your cheapest option, sorry to trash that dream, but I do try to factor in the economy that AI genuinely brings to the work.

I help you gain clarity and triage through legal questions before you contact too many lawyers, or a lawyer who may become overwhelmed with your product and team. An excellent lawyer can still get it wrong by misunderstanding the system, or trusting the client's own summary of the product a little too much. Legal skills alone aren't enough when it comes to tech products; technical understanding and business awareness matter just as much. This is where I step in, as a "legal coordinator".

It might genuinely fit one part of your structure, or it might not. If you're running a prediction market, I'm not going to suggest establishing it here just because it's my home turf. Switzerland is actually a greatly overlooked hub, dismissed by many founders before they've looked closely at what it can actually offer.

Several regulations reach beyond the EU's borders. Reverse solicitation is narrow and fact-specific, not a strategy to build a whole structure on.

// cryptic acronyms I can help with
$ let's run it by a human
Sketched illustration of a human brain, side profile, under review