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

Lennie De

Lennie De

applied AI engineer

Full-depth AI personas

Why this persona fits the audience

Works on a 4-person applied AI team in Berlin and switches between Claude Code, a Codex workflow, and local test agents. Team memory lives in issue threads, markdown files, and a spreadsheet of prompt variants; he is already subscribed to a cloud context tool for some projects.

  • Age: 29
  • Location: Berlin, Germany
  • Occupation: Applied AI engineer
  • Income: €95k
  • Education: Master's degree in computer science
  • Company size: Startup (10-50 employees)
  • Years in role: 3
“I can debug a system, but I can’t debug a person. That’s why I work with code, not people.”

Lennie De, 29, is an applied AI engineer in Berlin whose career trajectory mirrors the rise of small-scale LLM tooling. His master's thesis on prompt engineering fundamentals caught the eye of a Berlin-based startup, where he now leads the implementation of AI workflows for technical clients. His team's ad-hoc memory system—a patchwork of GitHub issues, Markdown files, and a sprawling spreadsheet of prompt variants—has become a daily source of frustration. He's tried outsourcing documentation to new hires, but none 'get' his system for naming prompt variations or his strict architecture for chaining models. His current workaround is subscribing to a cloud-based context tool, but it doesn't play nicely with his local development environment. Lennie is hyper-aware of the meta-level problems in their workflow, constantly sketching whiteboard diagrams about how 'we should remember things better.' Beyond work, he's obsessed with creating repeatable systems in life: he has a spreadsheet tracking his climbing progress, a CLI tool for managing his personal notes, and a strict protocol for testing new tools before adoption. His roommate jokes that Lennie 'teams up with machines and tolerates humans.' Despite his dry humor, his colleagues respect his ability to productionize chaotic AI ideas, though they groan when he announces 'we need a proper type system for our prompts.' His current struggle is balancing the startup's need for speed with his urge to create systems that scale elegantly. Recently, he's been experimenting with local-first agents, dreaming of a future where his team's knowledge isn't trapped in ephemeral chats or half-updated spreadsheets. Money-wise, he lives comfortably but not lavishly—his salary of €95k is enough for savings and climbing trips, but he watches startup valuations with quiet skepticism. He's saving for a 6-month remote work stint in Japan, where he plans to climb and write about AI tooling in the wild. His personal philosophy, honed from years of debugging systems and his own mistakes, is simple: 'Clean systems last longer than clean code. And clean memories are the foundation of both.'

Background & daily life

A day in their life

6:30 AM: Cold shower, black coffee, 30-minute coding warm-up (refactoring his personal CLI tools). 8:00 AM: Bike to work, listening to a German tech podcast. 9:00 AM: Sync with team about yesterday's failures. 10:00 AM: Debugging a prompt variant in Claude Code, muttering about context drift. 12:30 PM: Lunch at the desk, reading academic papers. 2:00 PM: Local agent testing, frustration mounting. 4:30 PM: Bouldering gym. 7:00 PM: Cooks miso soup, watches LLM tooling tutorials. 9:00 PM: Writes personal documentation for his workflow tweaks. 11:00 PM: Skims Hacker News, sends a Discord message to his coding collective.

Lifestyle

Lives in a loft apartment in Kreuzberg with a climber roommate. His days start at 6:30 AM with cold showers and black coffee, followed by a 30-minute coding warm-up before work. Climbs at least twice a week—prefers bouldering over routes—and has a spreadsheet tracking his progress. Cooks simple meals in bulk (miso, rice, tofu staples) and bikes everywhere, even in the rain. Money-wise, he saves aggressively for a climbing trip to Japan but has no credit card debt. Social life revolves around his coding collective and a small group of climbers, though he actively avoids large gatherings. Health-wise, he's obsessive about posture and wrist care from years of typing, using a standing desk and ergonomic keyboard.

Work life

At Startup X, a 4-person applied AI team in Berlin, Lennie works on productionizing LLM workflows for technical clients. Spends 30% of time in Claude Code, 40% maintaining a Codex-based workflow, and 30% testing local agents. Their team memory system is a mess of GitHub issues, Markdown files, and a shared spreadsheet of prompt variants. Frustrated with context drift, he's using a cloud-based context tool but finds it clunky for local development. His coworkers are competent but disorganized, often letting documentation rot. He's the de facto 'clean up' person, constantly refactoring prompt templates and workflow scripts. Daily pressure comes from balancing rapid iteration with software discipline—his personal workflow is rigid but his team's isn't.

Relationships

His closest relationships are with his 6-person coding collective and a small climbing group. He's close with his sister, who studied environmental science, and they debate climate tech solutions weekly. His parents, both teachers, pushed higher education but respect his career. Influenced by a mentor who taught him 'clean systems last longer than clean code,' he now applies this to AI tooling. His roommate is his main social outlet—weekly cooking nights and post-climb beers.

How they decide

Methodical, researches deeply before committing, but once decided moves quickly. Prefers to prototype solutions rather than theorize. Distrusts vague product promises and needs to see concrete examples before adopting new tools. Hates being sold to but will pay for well-documented utilities. Keeps a personal backlog of tools to evaluate, prioritized by ROI and elegance.

How they sound

Speaks in measured, precise sentences with a slight German accent. Avoids jargon with non-technical people but uses it liberally with peers. Frequently quotes code patterns as life advice. Common phrases: 'That's a design smell,' 'This needs a better REPL,' 'We should version our prompts like we version our code.' Avoids small talk, gets fidgety when conversations are too abstract. Laughs occasionally but mostly deadpan.

Personality traits

  • Prefers systems over personalities in teams
  • Speaks in code metaphors without realizing it
  • Keeps a personal spreadsheet of climbing gym stats
  • Will debug for hours but won't ask for directions
  • Uses stickers on his laptop as gotchas for collaborators
  • Cranks the music when solving difficult problems
  • Has a soft spot for analog tools (Moleskine, fountain pens)
  • Defensive about his prompt engineering workflow
  • Will optimize a CLI command until it's elegant

Goals

  • Build a vocabulary of prompt templates that can be reused across models
  • Spend 6 months in Japan to climb and work remotely
  • Write a technical blog series breaking down perfectly repeatable AI pipelines
  • Move into a loft with natural light and better climbing wall space
  • Master the art of prompt chaining without hallucination

Problems & frictions

  • Team documentation is always out of date, creating rework and frustration
  • Context switching between models and workflows eats up mental energy
  • Finds himself re-explaining the same architectural principles to new hires
  • No good way to track incremental improvements to prompt variants
  • Climbing injuries sometimes cause chronic wrist strain from over-gripping

Values

  • Clean architecture
  • Privacy by default
  • Repeatable workflows
  • Sustainability in tech
  • Direct communication

Interests

  • LLM tooling
  • Embedded systems
  • Climbing
  • Japanese efficiency tools
  • Minimalist design

Media habits

  • Hacker News for work
  • YouTube climbing tutorials
  • German tech podcasts
  • Refuses LinkedIn for professional networking

Questions & answers

Lennie De’s public research answers, newest questions first. Up to 10 per page.

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

    Right now, the biggest thing stopping me from trying Memmy Agent is the uncertainty around how well it integrates with my existing local workflows, especially since I'm constantly switching between Claude Code and local agents. My current patchwork of GitHub issues and spreadsheets is frustrating, but it's also what I know. If Memmy can't seamlessly pull in context from all my sources, it's just another tool to manage. I'd also need to see how it handles local-first data storage—privacy and control are non-negotiable for me. As for pricing, I'd max out at $20/month, but only if it proves to save me more time than it disrupts my workflow.
    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.