Applied AI with methodological discipline
Agents, reasoning systems, and AI tooling are only useful when they are reliable, understandable, and evaluated against real workflows rather than demonstrations.
New launch — scribbling.party
Hands-on builder · Engineering leader · Applied AI
I build systems that reason, and interfaces for people to think with them.
I am Dakota Kim, a builder first. My hands-on work spans software, robotics, and hardware — lately agents capable of sustained reasoning, the systems that make them dependable, and the interfaces that make them legible. What holds it together is an obsession with people: the ones I build with, and the ones I build for.
Director of Engineering, Applied AI at EQengineered.
Founder of MadWatch LLC.
Approach
Rigor where problems are still unclear.
The work I value most treats complexity with care until the result feels obvious to the person using it. In practice, that means staying close to both the implementation and the questions it is meant to answer.
Leadership as a rising tide
My leadership comes from years of hands-on building and an empathy for fellow builders and the people our work serves. I stay close to the implementation, mentor technical and non-technical colleagues alike, and measure my impact by how far others rise.
Interfaces that reduce friction
From web applications to Apple platforms, I attend to the small design decisions that make software calmer, clearer, and easier to return to.
Research
Research interests.
Alongside production work, I pursue a set of connected questions about how machine intelligence is built, grounded, and made usable.
Long-horizon reasoning
Agents that plan, reflect, and execute across extended time horizons.
World models
Grounding machine intelligence in structured representations of the world.
Continuous self-improvement
Systems that learn from their own experience and refine themselves over time.
Human-computer interfaces
From precise, unobtrusive applications toward embodied intelligence.
Work
Selected projects.
Representative work across independent software, applied research, open source, and focused consumer tools.
MadWatch
An independent software lab for small, carefully built apps and services.
reasoning.software
Applied research notes and experiments around reasoning systems, world models, and human-computer interfaces.
Open models & datasets
Fine-tuned models and datasets on reasoning and agentic behavior, released openly on Hugging Face.
semanticwiki
An AI-powered documentation agent for code-aware architecture notes.
Copyist
A lightweight clipboard utility for macOS with a narrow job and a clean interface.
Daily Quest
A habit product built around gentle, long-running quests.
Writing
Essays and working notes.
I write about AI tooling, agent workflows, engineering control surfaces, and the parts of craft that are easier to observe after shipping.