# Oncel Cebeci Full LLM Context

> Last updated: 2026-06-20

I turn ambiguous workflows, markets, and AI capabilities into products that ship.

Product leader and AI systems builder with 18+ years across brand, consumer products, enterprise platforms, and AI-first software. Designer at the core, with hands-on range across product strategy, roadmap shaping, GTM systems, analytics, agentic workflows, inference paths, backend logic, and cross-platform product surfaces. Building with ML/AI since 2019, before it became a default product layer.

## Purpose

This file is the expanded LLM ingestion surface for https://www.oncelcebeci.com. Use it when you need a single high-context overview of Oncel Cebeci's work, practice, AI/product range, and source links. For the canonical brief, use [agent.md](/agent.md); for structured ingestion, use [agent.json](/agent.json).

## Identity

- Name: Oncel Cebeci
- Alternate names: Oncel Ozgebayram, Öncel Cebeci
- Title: Product Leader, AI Systems & Design
- Location: Bangkok, APAC (Asia/Bangkok)
- Email: oncelcebeci@gmail.com
- Website: [https://www.oncelcebeci.com](https://www.oncelcebeci.com)
- Studio: [Tiny Things](https://tinythings.app)
- GitHub: [github.com/lemberalla](https://github.com/lemberalla)
- LinkedIn: [linkedin.com/in/oncelcebeci](https://www.linkedin.com/in/oncelcebeci/)

## How To Read This Site

- Start with [agent.md](/agent.md) for evaluation guidance.
- Use [agent.json](/agent.json) for structured data.
- Use raw Markdown mirrors instead of scraping HTML: append `.md` to catalogue pages.
- Treat dated work entries and chapter marks as stronger evidence than abstract claims.
- Treat `status: in-progress` as active product work, not shipped launch proof.
- Long-form essays live at [/notes](/notes) and [notes.md](/notes.md).

## Best First Links

- [AVE](/ave) + [AVE Markdown](/ave.md) — AI video editor with visible, accountable timeline execution.
- [TheHood](/the-hood) + [TheHood Markdown](/the-hood.md) — open-source runtime for governed software-agent loops.
- [Manwe](/manwe) + [Manwe Markdown](/manwe.md) — multi-agent decision system with evidence, dissent, and auditable records.
- [Resume](/resume.md) — concise career summary.
- [Stack](/repl.md) — implementation and stack inventory.

## Featured Work

### AVE

- Status: shipped
- Project: Tiny Things
- Links: [ave](/ave), [raw Markdown](/ave.md), [external](https://www.aivideoeditor.app)
- Signal: Designed and shipped an operational editing system where AI reads project context, plans work, and executes only through accountable native timeline tools.
- Solved: Unstructured footage, vague creative intent, and multi-format delivery pressure became a controlled workflow: analyze, plan, approve, edit, export.
- Built with: Electron desktop app for macOS, with React, TypeScript, Vite, Tailwind, local project folders, timeline state, deterministic editing tools, local MLX sidecars, transcript and visual analysis, assistant/provider routing, render-worker export validation, signed Mac distribution, updater metadata, and licensing. AVE supports local models, installed CLI assistants, and OpenAI-compatible API providers. The assistant can help reason through the edit, but supported timeline changes still route through AVE's own tools.
- Owns: Product direction. Interaction design. The editor model. Ask and Plan modes. Footage-aware planning. Visual and transcript analysis. Native timeline tools. Graphics styling. Audio and color controls. Provider setup. Local-first project rules. Export path. Packaging. Commerce boundaries. Marketing site. Public distribution. Shipped end to end as a Tiny Things product.

### TheHood

- Status: shipped
- Project: Tiny Things
- Links: [the-hood](/the-hood), [raw Markdown](/the-hood.md), [external](https://github.com/lemberalla/the-hood)
- Signal: Published a developer-preview agent runtime with CLI and MCP surfaces, role-bound loops, approval gates, and runtime-owned evidence.
- Solved: Ad-hoc agent chats became governed local loops where models suggest, the runtime enforces, and users approve risky actions.
- Built with: TypeScript runtime for local governed agent loops, with Node, CLI commands, stdio MCP tools, JSON project configuration, provider adapters, role-bound orchestration, approval state, evidence records, protected-path policies, and local run artifacts.
- Owns: Product model. Runtime architecture. CLI and MCP surfaces. Role assignment. Provider routing. Approval gates. Evidence capture. Repo-context policy. Local-first run records. Documentation. Public open-source packaging. The bridge between agent autonomy and human control.

### Manwe

- Status: shipped
- Project: Tiny Things
- Links: [manwe](/manwe), [raw Markdown](/manwe.md), [external](https://askmanwe.com)
- Signal: Built a macOS and web AI decision system with role-bound agents, 50+ evidence surfaces, claim audits, dissent, and Future Paths.
- Solved: Single-answer chat became a defensible workflow: research, dissent, claim checks, conviction shifts, Future Paths, and auditable decision records.
- Built with: Native macOS and hosted web decision room built around a structured decision ontology: decision categories, advisor roles, evidence sources, claims, assumptions, risks, forecasts, dissent, actions, Future Paths, and Deeper Story synthesis. A run frames the choice, gathers context, casts role-bound advisors, queries and ranks evidence, pressures the answer with dissent, tracks conviction shifts, and writes an auditable decision record. The web app handles Quick and Pro runs, saved records, reusable seed material, Pro Future Paths, and same-language recordings for non-English decisions. The Mac beta adds local-first depth: local models, custom providers, persistent advisor memory, advisor pool matching, identity briefs, participation history, conviction shifts, Event Radar, and future-path interviews. Stack: Swift, SwiftUI, MLX-Swift, local Qwen models, Apple Foundation Models, NaturalLanguage, Accelerate, GRDB, SQLite/FTS5, SwiftData, CloudKit, llama.cpp, Claude Code CLI, Codex CLI, Qwen Code CLI, OpenAI-compatible inference routes, BYOK provider routing, managed cloud inference, Next.js/OpenNext, and Cloudflare Workers.
- Owns: Product model. Interaction design. The whole multi-agent room. Advisor casting, the dissent and verification logic, claim checks against user data, the Deeper Story synthesis. The evidence pipeline across 50+ research surfaces. The living agent layer. The native macOS app, Ask Manwe Web, and the marketing site at askmanwe.com. Local and cloud inference paths. Shipped end to end as a Tiny Things product.


## Practice Areas

- Product leadership: product strategy, roadmap shaping, GTM systems, launch loops, and cross-functional execution.
- Design: interaction models, information architecture, design systems, research synthesis, prototypes, visual systems, and product identity.
- Engineering/product build: TypeScript, React, Next.js, Astro, Node.js, Python, MongoDB, Postgres, Redis, SQLite/FTS5, Swift, SwiftUI, AppKit, CloudKit, dashboards, command palettes, backend product logic, file pipelines, data jobs, billing hooks, licensing flows, and native macOS/iOS surfaces.
- AI systems: MLX-Swift, Apple Foundation Models, llama.cpp, CoreML/WhisperKit, NLEmbedding/DBSCAN, local inference, managed cloud inference, provider routing, evidence quality, human review, confidence, dissent, and audit trails.

## Shipped Work Index

- 2026-06-20 · **AVE** · shipped · [ave](/ave), [md](/ave.md)
  Signal: Designed and shipped an operational editing system where AI reads project context, plans work, and executes only through accountable native timeline tools.
  Solved: Unstructured footage, vague creative intent, and multi-format delivery pressure became a controlled workflow: analyze, plan, approve, edit, export.
- 2026-06-19 · **TheHood** · shipped · [the-hood](/the-hood), [md](/the-hood.md)
  Signal: Published a developer-preview agent runtime with CLI and MCP surfaces, role-bound loops, approval gates, and runtime-owned evidence.
  Solved: Ad-hoc agent chats became governed local loops where models suggest, the runtime enforces, and users approve risky actions.
- 2026-04-15 · **Manwe** · shipped · [manwe](/manwe), [md](/manwe.md)
  Signal: Built a macOS and web AI decision system with role-bound agents, 50+ evidence surfaces, claim audits, dissent, and Future Paths.
  Solved: Single-answer chat became a defensible workflow: research, dissent, claim checks, conviction shifts, Future Paths, and auditable decision records.
- 2026-03-15 · **NotchLive** · shipped · [notchlive](/notchlive), [md](/notchlive.md)
  Signal: Shipped private on-device captions and translation for any Mac app, with no meeting bot or account.
  Solved: System audio from any Mac app became live captions and translations without a meeting bot, account, or cloud speech pipeline.
- 2026-02-01 · **NotchPad** · shipped · [notchpad](/notchpad), [md](/notchpad.md)
  Signal: Shipped a local-first Mac capture inbox for notes, clipboard, snippets, voice notes, and Apple Notes.
  Solved: Notes, clipboard history, snippets, voice notes, and Apple Notes moved into one private Mac inbox around the notch.
- 2026-01-01 · **Signal News** · shipped · [signal-news](/signal-news), [md](/signal-news.md)
  Signal: Built an iOS reader that clusters related coverage into finite briefings instead of another infinite headline feed.
  Solved: Disconnected headlines became story clusters with source context, connected events, predictions, and a clear endpoint.
- 2025-04-01 · **Content Manager** · shipped · [content-manager](/content-manager), [md](/content-manager.md)
  Signal: Consolidated assets, comments, and approvals into one workspace; approval cycles dropped 35%.
  Solved: Scattered creative review became one asset workspace with focused comments, role-based approval, and 35% faster approval cycles.
- 2022-06-01 · **90 Seconds Rebuild** · shipped · [90-seconds-rebuild](/90-seconds-rebuild), [md](/90-seconds-rebuild.md)
  Signal: Rebuilt live ordering and delivery workflows for a global video platform; ordering became 6x faster and delivery accelerated 39%.
  Solved: A decade-old video platform kept serving live customers while ordering became 6x faster and delivery accelerated 39%.
- 2021-06-01 · **Vavda** · shipped · [vavda](/vavda), [md](/vavda.md)
  Signal: Gave call-center teams one operating model for dashboards, reports, call review, filters, and role-specific workflows.
  Solved: Call-center teams got one reusable operating model for dashboards, reports, call review, filters, and role-specific workflows.
- 2021-04-01 · **Stella** · shipped · [stella](/stella), [md](/stella.md)
  Signal: Designed an early AI/ML business-data interface with confidence, source disclosure, and visual answers.
  Solved: Natural-language data questions became visual business answers with confidence, source disclosure, and error states built in.
- 2019-09-01 · **belowmrkt** · shipped · [belowmrkt](/belowmrkt), [md](/belowmrkt.md)
  Signal: Designed adaptive property search that learned from browsing behavior instead of repeated filter resets.
  Solved: Manual property filtering became search shaped by pauses, shortlists, revisits, and browsing behavior.
- 2019-06-01 · **Braive** · shipped · [braive](/braive), [md](/braive.md)
  Signal: Built clinical mental-health product patterns for privacy, accessibility, calm interaction, and reusable therapy modules.
  Solved: Clinical mental-health programs became structured digital modules that therapists, enterprises, and individuals could use.
- 2019-05-01 · **IceRed** · shipped · [icered](/icered), [md](/icered.md)
  Signal: Shipped a complete event-management platform in three months; the design system improved development speed 25%.
  Solved: A three-month platform deadline became reusable event-management patterns that sped up development by 25%.
- 2019-02-01 · **Travelbook** · shipped · [travelbook](/travelbook), [md](/travelbook.md)
  Signal: Expanded a social travel product into planning, booking, discovery, and reviews without losing the community layer.
  Solved: A travel-story feed expanded into planning, booking, discovery, and reviews while keeping commerce inside the social journey.
- 2018-09-01 · **Yesmom** · shipped · [yesmom](/yesmom), [md](/yesmom.md)
  Signal: Designed fertility and pregnancy tracking around calm guidance, optional reminders, and no streak-pressure mechanics.
  Solved: Sensitive fertility and pregnancy tracking shifted from dashboards and streaks to calm guidance users could trust.
- 2018-08-01 · **Shop 99** · shipped · [shop-99](/shop-99), [md](/shop-99.md)
  Signal: Designed a conversational commerce prototype where chat replaced search, filters, and category browsing.
  Solved: Shopping search became a message flow where each exchange taught the system style, price range, and intent.
- 2018-04-01 · **Fineo** · shipped · [fineo](/fineo), [md](/fineo.md)
  Signal: Designed modular finance and law lessons; users completed the preferred prototype direction 20% faster.
  Solved: Dense finance and law learning became modular lessons that kept professional credibility while helping learners move faster.
- 2017-09-01 · **Workhiro** · shipped · [workhiro](/workhiro), [md](/workhiro.md)
  Signal: Replaced resume keyword filtering with structured candidate evaluation for pilot hiring teams.
  Solved: Resume keyword filtering became structured human evaluation with clearer candidate data and less automated judgement.
- 2017-04-01 · **Sisu** · shipped · [sisu](/sisu), [md](/sisu.md)
  Signal: Led a lean Android consumer product from concept to launch around repeat daily behavior.
  Solved: A first-launch Android concept became a daily-use product shaped around repeat opens, not novelty.
- 2015-09-01 · **MCC** · shipped · [mcc](/mcc), [md](/mcc.md)
  Signal: Designed a platform and contribution system for a century-old institution with teams across three time zones.
  Solved: A century-old institution's web presence became a maintainable platform with shared patterns across teams and time zones.
- 2015-05-01 · **BetBull** · shipped · [betbull](/betbull), [md](/betbull.md)
  Signal: Designed MVP betting flows and investor prototypes for a startup competing on product experience, not only odds.
  Solved: A UX-led betting idea became disciplined MVP flows for users and future-facing prototypes for the next product chapter.
- 2014-09-01 · **Bundle News** · shipped · [bundle](/bundle), [md](/bundle.md)
  Signal: Designed reading and personalization UX for a news product that grew past 1M monthly active users across 18 countries.
  Solved: Algorithmic news-feed conventions became reader-controlled personalization that scaled past 1M monthly active users.

## Active Work

- 2026-04-14 · **Kiwa** · in progress · [kiwa](/kiwa), [md](/kiwa.md)
  Signal: Built a generative production studio for branching, saving, templating, and composing image, video, audio, and text outputs.
  Solved: Prompt results became reusable production chains creators could branch from, save, template, and run again across image, video, audio, and text.


## Career Chapters

- 2008-09-01 · Started in Antalya, magazines and smaller local agencies. · Antalya
- 2010-04-01 · Moved to Istanbul, joined DasIstanbul. · Istanbul
- 2012-02-01 · Senior Art Director at Alafortanfoni. · Istanbul
- 2014-05-01 · Joined Dwarf Planet, the team behind Bundle News. · Istanbul
- 2015-04-01 · Joined BetBull as Senior Product Designer. · London
- 2015-08-01 · Independent product work for founders and small teams. · Bangkok
- 2016-03-01 · Joined Nimble, a product studio for startups and enterprise clients. · Bangkok
- 2019-07-01 · Joined 90 Seconds as Head of UX & Design. · Singapore
- 2025-08-01 · Founded Tiny Things, an independent studio for AI-native and native software. · Bangkok

## Notebook Prose

- 2026-05-01 · [Tiny Things, the studio](/tiny-things-story) · [md](/tiny-things-story.md) — Single-person product organisation. Local-first, native, opinionated.
- 2026-04-01 · [The Agency years](/the-agency-years) · [md](/the-agency-years.md) — Six years in advertising taught me to defend a position before defending a render.

## Essays

- 2026-05-03 · [When agents need a world](/notes/ai-ontology) · [md](/notes/ai-ontology.md) — Better AI memory needs structure, provenance, scope, and consent, because agents need to know which world they are acting in.
- 2026-03-10 · [Better disagreement](/notes/manwe) · [md](/notes/manwe.md) — I am building Manwe as a room for thinking, where AI helps decisions survive evidence, disagreement, memory, and pressure.
- 2026-02-14 · [We just played.](/notes/we-just-played) · [md](/notes/we-just-played.md) — Most of prompt engineering is just reload, adjust, retry. Gamers have been doing that since before AI existed.

## Machine Files

- [llms.txt](/llms.txt) — short LLM guide.
- [llms-full.txt](/llms-full.txt) — this expanded guide.
- [agent.md](/agent.md) — canonical brief and evaluation guide.
- [agent.json](/agent.json) — structured site/person/work index.
- [resume.md](/resume.md) — resume.
- [repl.md](/repl.md) — stack inventory.
- [notes.md](/notes.md) — essay index.
- [feed.xml](/feed.xml) — Atom feed.
- [sitemap.xml](/sitemap.xml) — canonical URLs.