🤖 Technical Show-and-Tell for AI Builders (AIビルダーのための技術ショー＆テル)

We don’t want just what you built; we want how you built it: the architecture decisions you made, the instrumentation that surfaced the real problem, the eval or metric that finally told you it was working, and the failure that taught you the most. No requirement to use any specific tool-bring whatever stack you shipped on. Pop the hood, show us the wiring, and trade the hard-won lessons that don’t make it into the README.

The best build stories aren’t about the finished product; they’re about the lessons along the way.

私たちが知りたいのは「何を作ったか」だけではありません。「どう作ったか」です。アーキテクチャ上の意思決定、本当の課題を発見した計測や観測手法、うまく動いていると確信できた評価指標、そして最も大きな学びを与えてくれた失敗について共有してください。特定のツールやフレームワークを使っている必要はありません。実際に動かしたスタックを持ち寄り、内部構造を見せながら、READMEには書かれない実践的な知見を交換しましょう。

優れたビルドストーリーは完成品ではなく、その過程で得られた学びにあります。

AI Tinkerers is back in Tokyo for a curated session focused on the architecture and implementation of generative AI systems. This is a high-signal environment for engineers, researchers, and founders who are actively shipping code. We prioritize technical depth over high-level overviews, focusing on the “how” behind the build.

This special event is sponsored by PostHog, providing engineers with the tools to build, test, measure, and ship successful products faster.

AI Tinkerers Tokyoが再び開催されます。今回は生成AIシステムのアーキテクチャと実装に焦点を当てた特別セッションです。実際にコードを書き、プロダクトを構築しているエンジニア、研究者、起業家のためのハイシグナルなコミュニティイベントです。表面的な概要ではなく、実装の裏側にある「どのように作ったのか」を重視します。

本イベントは PostHog の協賛により開催されます。PostHogは、エンジニアがより速く製品を構築・検証・計測・リリースするためのツールを提供しています。

Registration is required. Because our events are popular and space is limited to 50 participants, admission is based on demonstrated technical work to maintain a builder-to-builder atmosphere.

参加登録が必要です。定員は50名のため、技術的な取り組みや実績をもとに参加者を選考し、ビルダー同士が学び合える環境を維持しています。

Submit Your Demo Proposal Here

デモ発表の応募はこちら

🛠 The Format: Demos, Not Pitches (形式：ピッチではなくデモ)

We follow a strict “Show, Don’t Tell” policy. Our demos are 8-minute technical deep dives where builders pop the hood on their stacks. We want to see your code, your architecture diagrams, and your failure cases.

私たちは「Show, Don’t Tell（語るより見せる）」を重視しています。発表は8分間の技術デモ形式です。コード、アーキテクチャ図、計測データ、失敗事例など、実際の実装を見せてください。

What we look for (歓迎する発表内容)

• Agentic Workflows: Autonomous orchestration, tool-use, and multi-agent systems using GPT-5.5 Pro or Claude 4.7.
  GPT-5.5 ProやClaude 4.7を活用したエージェントシステム、ツール利用、マルチエージェント構成。
• Local Inference: Running frontier-level performance on the edge using techniques like TurboQuant or 1-bit quantization (Bonsai 8B).
  TurboQuantや1-bit量子化、Bonsai 8Bなどを活用した高性能なオンデバイス推論。
• Infrastructure: Productionizing LLM systems, MCP-based debugging pipelines, and network-jailing for agent security.
  本番LLMシステム、MCPベースのデバッグ環境、観測性、エージェントセキュリティ。
• Experimental Builds: Novel context engineering, multimodal OCR pipelines, and design-to-code systems.
  コンテキストエンジニアリング、マルチモーダルOCR、デザイン→コード変換などの新しい試み。

Pitch decks are prohibited. If you are working on something unfinished, buggy, or highly experimental, this is the place to share it and trade notes with peers.

営業資料や投資家向けピッチデックは禁止です。未完成のもの、バグのあるもの、実験的なものこそ歓迎します。

🕕 Schedule (スケジュール)

🧪 Presentations & Networking (発表とネットワーキング)

We focus on builders sharing projects they have been working on and their learnings. Topics include the latest models, tools, patterns, and techniques for creating amazing AI solutions. It is an informal venue for code walkthroughs and deep technical questioning that might not fit the main stage.

参加者同士が取り組んでいるプロジェクトや、そこから得た学びを共有することを重視しています。最新モデル、開発ツール、実装パターン、評価手法、本番運用の知見など幅広いテーマが扱われます。通常のカンファレンスでは扱いにくいコードレビュー、アーキテクチャの議論、深い技術的な質疑応答ができる場です。

Event photos

AI Tinkerers - Tokyo

AI Tinkerers - Tokyo

AI Tinkerers - Tokyo

AI Tinkerers - Tokyo

AI Tinkerers - Tokyo

AI Tinkerers - Tokyo

Sponsors (スポンサー)

AI Tinkerers is proud to be supported by partners who provide the infrastructure and tools that enable the builder community.

AI Tinkerersは、ビルダーの挑戦を支えるインフラやツールを提供するパートナー企業によって支えられています。

All your developer tools in one place. PostHog gives engineers everything to build, test, measure, and ship successful products faster.

開発に必要なツールをひとつのプラットフォームで提供し、プロダクト開発を加速します。

Sponsorship opportunities for future events are available. See sponsorship opportunities for details.

今後のイベントへのスポンサー参加も募集しています。

AI Tinkerers Tokyo Stats (AI Tinkerers Tokyo 実績)

• Attendees / 参加者数: This active community of 314 technical professionals consists of 43% AI/ML researchers, 35% full-stack AI developers, and 22% technical leaders. Members specialize in LLM fine-tuning, RAG pipelines, and MLOps. Key distinctions include a high concentration of builders from premier institutions like Sakana AI and Woven by Toyota, alongside deep expertise in bilingual AI deployment and robotics.
• Companies Represented / 参加企業数: Featuring tech giants like Microsoft, Google, Amazon, and NVIDIA, alongside AI pioneers Sakana AI, Hugging Face, ElevenLabs, and Shisa.ai, and innovative startups such as Woven by Toyota, Mercari, Playco, and more
• Demos / デモ発表数: 45 demos were submitted and 34 were presented. The most engaging themes have explored agentic workflow automation, secure and governed agent execution, and local-first/privacy-constrained deployments. Technically, attendees have focused on RAG (including multimodal variants), multi-agent orchestration, observability/debugging, and on-device inference optimization, alongside evaluation via measurable experimentation and structured knowledge/claim tracking.
• Testimonials / 参加者の声:

  “This was a really good demo for me, I could learn the process used for a fact checking system. I was very interested in fact checking from a long time ago, and this age definitely needs it.”

  — CEO, search & AI

  “Log analysis via cclogviewer is a necessary step for debugging subagents. The stats on skill activation provide a clear, data-driven path for optimization.”

  — COO, energy tech

  “Having lots of interesting talks”

  — Founder, Japanese AI startup

  “Keep the live code and demos over slides format. Sharing deep implementation details and what broke provides much more value than typical product pitches.”

  — COO, energy tech

  “I really liked this demo. I had an idea to use MCP to do something similar and was glad somebody was doing it before me.”

  — CEO, search & AI

  “Thank you for organizing such a high-quality builder community in Tokyo. The networking was incredibly stimulating.”

  — COO, energy tech

  “Path-based character recognition over standard OCR is the right call for handwriting. It makes the SRS flow feel seamless for Japanese learners.”

  — COO, energy tech

📍 Location (会場)

The meetup will be held in Shinagawa, Tokyo. The exact venue address and access instructions will be sent only to accepted and registered attendees.

会場は東京都品川エリアです。正確な住所および入館方法は、参加確定者にのみご案内します。