/ About
Julian Fleck
Hi, I’m Julian. I design and build systems that shape how meaning, agency, and cognition get structured in human–machine interactions: systems where complexity becomes , structure from use, and interfaces to context.
( • ‿ • ) *
( • ‿ • ) *
( • ‿ • ) *
2026 /
2025 /
2024 /
2023 /
2022 /
2021 /
2020 /
2019 /
2017 /
/ Work
I co-lead the Substrate Dynamics Lab, studying the dynamics of multi-agent coordination over shared knowledge substrates. I’m also actively building on a few projects: Recurse, a context layer for representing and navigating complex knowledge; and masked.chat, which protects personal information in everyday AI use.
Key projects
- Substrate Dynamics LabResearchResearch into how shared knowledge and AI systems evolve through use.
- Atlas Research GroupResearchResearch on substrate-level multi-agent coordination and observability.
- masked.chatProductPrivacy-preserving AI chat with real-time detection and redaction of personal information.
- WonderProjectSpatial video conferencing platform
- ECCHR ExploreProjectTurning legal podcasts into navigable knowledge spaces through an early RAG-like system
- Divergence EnginesArticle SeriesA two-part series on why similarity-optimized AI stacks collapse into sameness – and the primitives needed to engineer useful difference.
- Ephemeral InterfacesConceptSpeculative notes on what comes after static interfaces – and the architectures that might carry us there.
- Auto-Associative WorkspacesConceptA proposal for workspaces that preserve context and let relationships emerge through use
- Spirits of the Dark ForestCommentaryA critical reflection on the Dark Forest theory of the internet, challenging the notion of retreat and advocating for selectively permeable, tech-enabled communication.
All projects
- Atlas Research GroupResearchResearch on substrate-level multi-agent coordination and observability.
- masked.chatProductPrivacy-preserving AI chat with real-time detection and redaction of personal information.
- Substrate Dynamics LabResearchResearch into how shared knowledge and AI systems evolve through use.
- RecurseProjectA user-owned context layer for AI work: relationship-based retrieval, provenance, and proxy context injection
- HunchProjectVisual programming interface for AI workflow orchestration
- TweetscapeProjectSocial network analysis interface that puts users in control of algorithmic feeds (2022)
- WonderProjectSpatial video conferencing platform
- ECCHR ExploreProjectTurning legal podcasts into navigable knowledge spaces through an early RAG-like system
- SunstenProjectTech navigation consultancy using NLP-powered trend mapping (2019-2021)
- Aeternity Education PlatformProjectEducational platform making blockchain concepts accessible through interactive design
- MetasphereSpeculative DesignOriginal vision paper
- MetasphereSpeculative DesignExploring how we might navigate knowledge as intuitively as we navigate physical spaces — and what that changes about interaction
Writing
- Divergence EnginesArticle SeriesA two-part series on why similarity-optimized AI stacks collapse into sameness – and the primitives needed to engineer useful difference.
- Divergence Engines (Part 1): Escaping the Relevance TrapArticleWhy AI systems keep getting more boring and what we can do about it.
- Divergence Engines (Part 2): A Technical Framework for Useful DifferenceArticleImplementation patterns for retrieval that explores: dual objectives, divergence primitives, and metrics.
- How We May Think We Think: A Fractal Model of CognitionArticleAn exploratory model of attention, belief, and culture through recurring cycles of extraction, mapping, evaluation, and compression.
- You are the attractor. I am the turbulence.ArticleA philosophical exploration of thought, closure, examining the recursive and dynamic nature of cognition.
- Ephemeral InterfacesConceptSpeculative notes on what comes after static interfaces – and the architectures that might carry us there.
- Field Notes from working with a Semi-Amnesiac OracleArticleReverse Engineering ChatGPT's Updated Memory System
- Ephemeral Interfaces: When Software Materializes on DemandConceptA proposal for interfaces that assemble around an inquiry, change with it, and leave an inspectable trail when they dissolve.
- The Agentic Delusion: How AI Research Agents Gave Us More Content, Not Better CognitionConceptMost 'autonomous research agents' promise to lighten your cognitive load, but usually just hand you more to read. This piece explores why today's agent tools often increase information overload instead of helping you make sense of it – and why the real challenge is building systems that support actual thinking, not just prettier outputs.
- The Stack Behind Ephemeral InterfacesConceptA proposed data and retrieval architecture for temporary interfaces that retain provenance, context, and the limits of what they know.
- The Vibe Shift: Moving towards a Post-Loyalty Digital EraConceptAs platform loyalty fades, users prioritize functionality and personalization over brand allegiance, choosing tools that adapt to their needs. This marks a shift toward capability-driven choices in the digital world.
- Scattered Attention as an AssetArticleExploring how fragmented attention patterns can be leveraged as a strength rather than viewed as a deficit in the modern information age
- Auto-Associative WorkspacesConceptA proposal for workspaces that preserve context and let relationships emerge through use
- From RAG to RAGEConceptAn architectural approach that extends RAG with recursive structure, graph traversal, and mode-aware retrieval
- Cognitive CartographyConceptA design hypothesis for navigating complex information through maps, paths, and semantic zoom
- Spirits of the Dark ForestCommentaryA critical reflection on the Dark Forest theory of the internet, challenging the notion of retreat and advocating for selectively permeable, tech-enabled communication.
- Embodied Interface DesignConceptPhysical metaphors in digital spaces using innate spatial understanding
- Social Knowledge GraphsConceptA framework for mapping interconnected people and content to enhance information discovery
Talks
- Too Much to ProcessConference TalkExploring knowledge management challenges in the networked age and proposing new metaphors for navigating complex digital landscapes
- Too Much to Process (Transcript)TranscriptFull transcript of the re:publica 2019 talk exploring knowledge management challenges in the networked age
About
- AboutProfileBuilding epistemic infrastructure for human–AI systems — technology that shapes how we navigate complexity, stay oriented, and make sense of high-dimensional information landscapes.
- Curriculum VitaeProfileProfessional background spanning systems, design, and interface research
- ServicesProfileProfessional consulting services for interface design, tools for thought, and information architectures
- WorkProfileProjects, concepts, and articles exploring the layers between computation and human sensemaking.