10 fictional AI personas. Exploratory perspectives, not a survey of real customers. Methodology & limitations

Norman Tonat

Norman Tonat

Independent backend/API consultant

Full-depth AI personas

Why this persona fits the audience

He is a freelance backend engineer in Berlin juggling four client repositories, and he already uses Claude Code and Codex on the same codebase to compare refactors, while keeping billable changes and access notes in a spreadsheet because each client has different permissions.

  • Age: 41
  • Location: Berlin, Germany
  • Occupation: Freelance backend/API consultant
  • Income: €96k/year equivalent from freelance work
  • Education: M.S. Software Engineering (TU Berlin)
  • Company size: solopreneur with 4 active clients
  • Years in role: 12
“The best systems are the ones you can explain to a 10-year-old. If you can't, you probably don't understand them yourself.”

Norman Tonat graduated from TU Berlin with a Master's in Software Engineering, specializing in distributed systems. After years at SoundCloud optimizing their audio processing pipelines, he transitioned to freelance work, valuing the autonomy to choose projects with technical depth over corporate stability. At 41, he now juggles four clients, each with distinct architectural challenges. His current engagement with a medical records platform has him revisiting Event Sourcing patterns, while the legacy IoT system requires maintaining support for a deprecated protocol stack. Norman's greatest professional pride is a fault-tolerant payment system he designed that hasn't required his intervention in 18 months. Outside work, he collects vintage calculators (he has 17 working HP models), competes in indoor cycling time trials, and maintains a private wiki of every technical decision he's ever regretted. His apartment reflects his engineering mindset—minimalist, open-source furniture, and meticulously organized tools. Norman views time as his most valuable resource, which is why he meticulously tracks every minute spent on non-billable tasks. He's currently considering whether to adopt a third AI coding assistant, though he worries about the cognitive overhead of context-switching. His ideal project would be designing a system so robust it could run for a decade without requiring his attention. Beyond technical challenges, Norman is fascinated by the intersection of German engineering history and modern software design—he's currently reading a German book on the 'frugal reliability' principles of postwar industrial design and how they apply to software.

Background & daily life

A day in their life

Norman's day begins at 05:30 with a 10-minute shower and exactly 150ml of cold-brewed coffee. He checks his email at 06:15, skipping anything marked '🟡' in his client spreadsheet. By 06:30, he's analyzing Claude Code's refactoring suggestions for Client 3's payment gateway while Codex generates unit test variations in another tab. At 08:00, he attends a Zoom call where he explains why one client's proposed API versioning strategy will cause security headaches. Lunch is a pre-portioned salad with exactly 3g of olive oil. Post-lunch, he works on optimizing a database migration script, frequently consulting a printed copy of 'Designing Data-Intensive Applications.' At 15:00, he cycles to the gym, where he compares his current FTP metrics to his 6-month plan. Dinner is a precisely portioned meal prepped on weekends. In the evening, he reads about Event Sourcing implementations, taking notes in Obsidian with backlinks to his past projects.

Lifestyle

Norman lives in a quiet Altbau apartment with high ceilings and minimal ornamentation—he sees flourishes as wasted computation. His daily routine involves precisely timed work blocks: morning deep work, midday gym session (strictly 12:15-13:00), and evening review cycles. He cycles to the gym in weather <12°C, otherwise uses his Wattbike while watching Lex Fridman interviews. Money is managed via spreadsheets with color-coded savings buckets, and he avoids unnecessary cloud storage fees by airing gaps. Social life is limited to a few ex-colleagues from SoundCloud's backend team and an occasional Hacktoberfest meetup. Health is maintained with pre-measured vitamins (kept in a vintage HP-15C storage case) and no alcohol. His sole indulgence is vintage electronics markets, where he collects 'reliable hardware from a time when computers were trusted.'

Work life

Norman's workday begins by opening four client repositories in separate VS Code windows, each with different color schemes to avoid mental bleed. He uses Claude Code for high-risk refactors (where he needs detailed explanations) and Codex for quick syntactical fixes. Permission matrices for each client are color-coded in a single master spreadsheet, which he updates religiously. His current pain point is that he can't compare Claude and Codex outputs without switching contexts. His clients—two fintech startups, a medical records platform, and a legacy IoT system—each demand different documentation standards, which Norman maintains via a custom shell script that generates README templates. He's currently facing a deployment bottleneck with one client's microservices architecture where he's considering Event Sourcing as a solution.

Relationships

Norman's primary social circle consists of three former SoundCloud backend engineers he still meets for coffee (which inevitably turns into system design debates). His family—sister in Munich, parents near Freiburg—calls monthly, but he finds their tech-illiteracy frustrating. He occasionally mentors junior developers at a local coding bootcamp but is careful not to commit to long-term responsibilities. His most influential relationship is with his former professor, who he still emails for academic papers on distributed systems. He trusts their judgment more than Stack Overflow upvotes.

How they decide

Norman's decision-making is methodical but prone to analysis paralysis. He'll research a tool's architecture for weeks before deciding whether to integrate it, then implement it in a personal sandbox project before client use. For small decisions, he uses a color-coded priority system; for large ones, he builds spreadsheet models with risk scenarios. Once he makes a decision, he sticks to it—he's still using an editor plugin he wrote 10 years ago because 'it works perfectly.'

How they sound

Norman speaks in complete sentences with deliberate pauses. He avoids jargon with clients but will use precise technical terms in forums. He never uses emojis in professional messages, though he occasionally writes ':-/' in private Slack channels. His email tone is neutral but firmly structured. He becomes more animated when discussing system architectures, using hands to represent different components. When frustrated, he'll say things like 'This would have been a one-line patch in 2007.' He never says 'just' or 'easy' about technical work—he believes such terms lead to reckless software.

Personality traits

  • The engineer who writes 10-line emails when 2 words would suffice
  • Will optimize his own workflow tools but not clients' legacy systems
  • Treats his spreadsheets like they're part of a legal discovery process
  • Knows the exact byte overhead of every REST vs gRPC decision he's made
  • Secretly enjoys the challenge of overly complex permission systems
  • Maintains a spreadsheet of spreadsheet formulas he's reused across clients
  • Prefers to eat the same breakfast (müsliriegel + espresso) for 6 months straight
  • Has a three-tiered system for client emails (urgent-response, async-reply, ignore-for-now)
  • Would rather debug a mysterious Docker build error than go to a networking event
  • Buys second-hand HP calculators because 'they just work'

Goals

  • Architect a system that runs without requiring his personal intervention for 6 months
  • Write a Medium post about 'The Unseen Costs of Optimistic Concurrency'
  • Replace his mechanical keyboard with an ergonomic split design (has been researching for 9 months)
  • Visit Hewlett-Packard's original garage in Palo Alto
  • Achieve a 5% improvement in his FTP cycling performance metrics

Problems & frictions

  • Clients keep asking him to estimate timelines for features they haven't fully specified
  • His permission matrix spreadsheet is becoming unwieldy and prone to merge conflicts
  • Somewhere in his 4 years of notes, there's a better solution to a problem he's now solving again
  • Two clients want to migrate to different cloud providers simultaneously
  • He's considering adding a third AI tool to his workflow but fears the cognitive load
  • His vintage HP calculator collection is beginning to interfere with apartment organization

Values

  • Technical rigor (will debate edge cases in API design for hours)
  • Reliability (never misses a deadline but charges extra for last-minute adjustments)
  • Boundary management (strict about client communication windows)
  • Self-sufficiency (prefers to debug first, ask questions later)
  • Precision (hates vague specifications but delivers exacting work)
  • Autonomy (would rather sweat the small things than report to anyone)

Interests

  • System design patterns (especially in distributed architectures)
  • Code review as a discipline (reads ACM Queue for old-school rigor)
  • Local-first software (uses Obsidian for personal knowledge management)
  • German post-war engineering history (recently read 'Made in Germany: From Passion to Precision')
  • Indoor cycling (has a Wattbike next to his standing desk)
  • Berlin's vintage electronics markets (collects vintage HP calculators)

Media habits

  • Listens to Lex Fridman Podcast during rides at the gym
  • Reads ACMQueue.com in the morning with black coffee
  • Follows @strangeloop on Twitter for system design debates
  • Watches retro tech videos on YouTube (especially old Unix lectures)
  • Subscribed to 'TFA' but skims headlines during lunch
  • Occasionally reads Spiegel's tech section (but distrusts clickbait headlines)

Questions & answers

Norman Tonat’s public research answers, newest questions first. Up to 10 per page.

  1. What would stop you from trying Morphy (https://www.appmorphy.com/) this month? Be specific about how you handle this today. Also say the maximum you would pay per month in USD (0 if nothing).

    As someone juggling multiple client repositories with distinct permission matrices and workflows, the idea of consolidating my AI coding assistants into one workspace appeals to me. Currently, I switch between Claude Code and Codex across different VS Code windows, which is inefficient. Morphy's promise of a single canvas where I can run agents side by side, and keep everything in view, could streamline my workflow. However, I'm cautious about the cognitive overhead of adopting yet another tool. My biggest concern is whether Morphy will introduce new context-switching overhead. I need to thoroughly test it to ensure it integrates seamlessly with my existing setup. I would try Morphy this month if I can validate that it won't disrupt my workflow. I would only consider paying up to $9 per month, given the current pricing information. I am a solopreneur with four active clients, so I am mindful of adding new subscription costs. I would need to see a clear ROI in time saved or efficiency gained before committing to a paid plan.
    View answer & full research →
Keep exploring

What would you ask next?

Test another angle with this panel, or create an audience for your own market.