# Sonnet Projects at AI Tinkerers

> Canonical HTML: https://aitinkerers.org/technologies/sonnet
> Markdown URL: https://aitinkerers.org/technologies/sonnet.md
> Technology record last updated: 2026-02-23T07:13:35Z
> Generated: 2026-09-21T06:47:10Z

Sonnet is Anthropic's powerful, mid-tier AI model, balancing frontier intelligence with high-speed, cost-efficient performance for production-scale deployments.

Sonnet (currently Claude Sonnet 4.5) is Anthropic’s versatile model, optimized for complex agentic workflows and coding tasks. It delivers state-of-the-art performance, achieving 77.2% on the SWE-bench Verified coding benchmark (cite: 2.2, 2.4). The model is engineered for high-volume, real-time applications like customer support automation and financial analysis, supporting a 200K token context window (cite: 2.8). Pricing is set for efficiency: $3 per million input tokens (cite: 2.8). This makes Sonnet the recommended choice for developers needing top-tier reasoning and coding capability at a practical, scalable cost.

- Official technology site: https://docs.anthropic.com/en/docs/about-claude/models
- Public AI Tinkerers demos and talks: 25
- Result page: 1 of 2

## Recent Public Talks and Demos

### [Data Entry automation using Browser Agents](https://dubai.aitinkerers.org/talks/rsvp_Capdf8Q-Kyc)

An AI browser agent that reads a weekly timesheet from Excel and automatically logs every hour into our system no clicks, no copy-paste, no manual entry. In the demo, you'll watch it open a live Salesforce environment, navigate to the correct date, intelligently match projects and tasks, and enter the hours in real time all driven by Claude reading and interacting with the screen just like a human would so that you can focus ur time on more important stuff.

- Event context: AI Tinkerers Dubai - June Demo Day — 2026-06-13 — Dubai
- Public talk page: https://dubai.aitinkerers.org/talks/rsvp_Capdf8Q-Kyc

### [From 1 Agent to 11: Patterns for Multi-Agent Systems That Actually Work](https://columbus.aitinkerers.org/talks/rsvp_nZaEJ2xzX0c)

I built an AI operating system (AIOS) that runs my technology community organization, Cloud Nirvana, using 11 specialized AI agents coordinated from a Mac Mini. Each agent owns a domain: email triage, speaker pipeline QA, partnership tracking, CRM operations, content, community engagement, financial operations, and strategic planning. They share an encrypted CRM, communicate through file-based memory, and operate under a graduated trust model where each agent earns autonomy based on demonstrated reliability. Along the way, everything broke at least once. I documented those failures as reusable engineering patterns (inspired by the Gang of Four) and open-sourced them. The demo walks through the live system, shows agents handling real work in real time, and demonstrates how any builder can discover and apply these patterns against their own codebase.

- Event context: AI Tinkerers - Columbus April Meetup — 2026-04-06 — Columbus
- Public talk page: https://columbus.aitinkerers.org/talks/rsvp_nZaEJ2xzX0c

### [Calendar Club](https://columbus.aitinkerers.org/talks/rsvp_PguV58U8t70)

Ambient deep-research agent that discovers in person events and automatically syncs them to the user’s calendar. Built with LangGraph, the agent runs a multi-step research workflow, asking clarifying questions and gathers feedback during the middle of the research. Finally the result is rendered as agenda of the next week’s related events to sync with the users calendar.

- Event context: AI Tinkerers - Columbus April Meetup — 2026-04-06 — Columbus
- Public talk page: https://columbus.aitinkerers.org/talks/rsvp_PguV58U8t70

### [Repurposing Cursor as an Expert System for Astrology](https://hong-kong.aitinkerers.org/talks/rsvp_rmtmM_-zJ-4)

I will be taking astrology chart data from the Astro Gold app, inputting that data into the context window of Cursor using a model such as Composer or Sonnet, and then prompting the agent live to demonstrate a reading of an astrology chart. I will use it to read the chart of a famous person or entity (such as the formation of AI Tinkerers HK). The agent will be configured using a Markdown file that educates the agent how to read astrology data. Planetary and interpretative data will be exported from Astro Gold for use by the agent. One of the biggest problems with astrology apps is that they are one directional and lack the all important context necessary an interpretation to properly translate. For example, most existing apps such as Costar, AstroGold and The Pattern are one-way broadcast communication, but lack the ability for a user to interact in a conversational way. By using the data from one of these apps we can provide the background data that is needed for an agent to simulate the functionality of an astrology app. However, using the right prompts, we can then have two way communication with the provided interpretations. For example if the app had provided a reading that said "This will be a high energy day" the onus would fall on the user to try and map this interpretation to its meaning. However utilizing a coding agent, we can tell the coding agent the events of that day, to get an interpretation for that specific contextual data. As a practicing astrologer, I know from experience that the most difficult part of reading a chart is seeing through mountains of seemingly contradictory data. When you add up all of the aspects, signs and houses there is something like 13 octillion possible charts. I originally got the idea for this when I realized that artificial intelligence is adept and analyzing large sets of disparate data. I have used Cursor in live readings, with the client's observation, to assist me in providing recommendations. About me: - Two master's degrees: Information Security Policy and Management from Carnegie Mellon. Master's in Business Administration from The Hong Kong University of Science and Technology. - Currently a founding engineer at an AI cybersecurity startup that provides agent observation and governance - 10+ years software engineering experience in blockchain, finTech, healthTech, e-commerce and banking - 8 years professional practicing astrologer with hundreds of clieints I have a YouTube channel where I talk about the intersection of spirituality and pragmatism. This video gives an idea about how I approach astrology from a more scientific point of view: https://youtu.be/A1TZv4a7X8Q?si=KB6elOv0uzV_zSk5

- Event context: AI Tinkerers Hong Kong GBA at the Hive: Creative AI Demos & Technical Show-and-Tell — 2026-03-26 — Hong Kong
- Public talk page: https://hong-kong.aitinkerers.org/talks/rsvp_rmtmM_-zJ-4

### [How to Raise Seed, VC or EU Grant Money Fast — And the Solutions I'm Building to Speed Up Money in Europe](https://valencia.aitinkerers.org/talks/rsvp_-XYoYU91Rc8)

In Europe we have a problem on many planes: bureaucracy, lack of know-how about how money actually moves, what VCs look like, how to become investable, how to find hubs and get exposure for your traction. I'll break down the landscape and show the tools I'm building at EUACC.AI to help accelerate Europe. I'll explain how I fine-tuned my model to be hyper-competent at writing grant applications using agentic tool calling and a new skill training environment called Dojo.md, how it works, and how you could build something similar for your own domain.

- Event context: AI Tinkerers Valencia March Meetup — 2026-03-17 — Valencia
- Public talk page: https://valencia.aitinkerers.org/talks/rsvp_-XYoYU91Rc8

### [Custom Self Improving Subagents with Claude Code](https://seattle.aitinkerers.org/talks/rsvp_Hc7X3ifdjnw)

Most Claude Code setups are static—your agents do the same thing every time. But what if they could learn from their own execution? In this talk, I'll demo a self-improving subagent architecture running in Claude Code that uses a plan→build→improve cycle where agents update their own knowledge after every workflow, maintains expertise.yaml files (500-1000 lines per domain) that evolve based on real execution patterns, and shares learnings across domains via a collective knowledge registry so when one agent discovers something, all agents can benefit. Live demo: I'll show agents improving themselves in real-time, including the .shared/ folder pattern for cross-domain knowledge propagation.

- Event context: AI Tinkerers Seattle Meetup: Dev Tools Track — February 9th, 2026 — 2026-02-10 — Seattle
- Public talk page: https://seattle.aitinkerers.org/talks/rsvp_Hc7X3ifdjnw

### [AI All the Way Down: Building a Production Fuel Pricing System with 5 AI Agents](https://st-louis.aitinkerers.org/talks/rsvp_uWD0V_b2O1k)

I am a former long time CIO at High Tech manufacturer Laird Technologies (15 years), and then DuPont (https://www.linkedin.com/in/mattmizell/). I have recently joined a small local company named Better Day Energy to build the entire technology stack for last mile petroleum delivery. I am using AI to develop, and also run a lot of the backed. As an example: We use Claude Code to build AI agents in n8n and python that parse emails and monitor AI parsing failures. It's AI turtles all the way down—and it actually works in production.

- Event context: AI Tinkerers – St. Louis Meetup: February 4, 2026 — 2026-02-04 — St. Louis
- Public talk page: https://st-louis.aitinkerers.org/talks/rsvp_uWD0V_b2O1k

### [Why I dropped using REST for MCPs and started using GraphQL.](https://poland.aitinkerers.org/talks/rsvp_ZT1zz0ewZj8)

GraphQL offers huge benefits over REST for querying relational data. Yet almost no examples online use it. i'll present how huge impact on quality of tool calling, reducing the network load & tokens usage GraphQL has when implemented as MCPs. I'll cover: - when should MCP use GraphQL - how to implement MCP with GraphQL-like interface - how to use MCPs that use GraphQL-like interface - real-life impact on development speed, infrastructure, llm costs.

- Event context: AI Tinkerers Poland #6 - Meetup in Warsaw (26th November) — 2025-11-26 — Poland
- Public talk page: https://poland.aitinkerers.org/talks/rsvp_ZT1zz0ewZj8

### [Natural Language to Robot Motions](https://seattle.aitinkerers.org/talks/rsvp_xJ2TlLuu5u0)

Zettaware is an agentic robotics platform for robot motion planning.

- Event context: Frontier Builds Demo Night: Experiments at the Edge of AI — 2025-11-13 — Seattle
- Public talk page: https://seattle.aitinkerers.org/talks/rsvp_xJ2TlLuu5u0

### [AG-UI + Ragas for Evals](https://sf.aitinkerers.org/talks/rsvp_ka1abEJdLQY)

There are numerous examples in the Ragas documentation and Github showing how to integrate directly with various agentic frameworks, but they all involve directly invoking the agents through their respective APIs. This demo will show how AG-UI can be used as a more generic integration mechanism, communicating with any AG-UI-compatible agent over HTTP instead, opening up additional frameworks (e.g., ADK) as well as test scenarios (e.g., easily running the same datasets across a wider variety of agents).

- Event context: Evals are Over - Evals are Back - AI in Prod @ Okta — 2025-10-31 — San Francisco
- Public talk page: https://sf.aitinkerers.org/talks/rsvp_ka1abEJdLQY

### [Claude Draws (/ how to automate Claude for Chrome)](https://sf.aitinkerers.org/talks/rsvp_zvyMJwwAAEI)

Claude Draws is an interactive community art project which challenges Claude to bring your idea to life using the hottest graphic design software from 1989, Kid Pix. I'll give a live demo of Claude Draws, built on top of the Claude for Chrome research preview. I'll also talk about how I built Claude Draws, some best practices for building browser agents, and a few lessons I learned along the way.

- Event context: Agents in Production: Human in the Loop Orchestration w/ Temporal — 2025-10-17 — San Francisco
- Public talk page: https://sf.aitinkerers.org/talks/rsvp_zvyMJwwAAEI

### [Using ML for Government Transparency](https://nyc.aitinkerers.org/talks/rsvp_c-_XTku9cAI)

Journalists, academics, and advocates can request government documents under the Freedom of Information Act (FOIA) — but these requests often take months or years to fulfill. While talking with 100+ government employees who process those requests, we learned (1) the biggest blocker to a government agency fulfilling a journalist's FOIA request is request quality, and (2) the vast majority of requests these employees process are terrible. So we built a tool to help journalists draft and submit high-quality, easy-to-process requests; it's currently in beta being used by two dozen journalists (incl. from WIRED, POLITICO, NPR) who have filed ~300 requests. I'll be presenting the tool! We trained a small, BERT-based classifier to rate draft FOIA requests along three dimensions: clarity, scope, and (legal) compliance. When the user rates a draft request, we call a large language model (Gemini 2.5 Flash) that suggests improvements to increase the requests's score on a dimension, which the user can accept with a click. The user then submits the perfected request through our platform alongside an LLM-tailored fee waiver (which has been granted 100% of the time so far). (We also have a "brainstorm" flow chaining calls to Llama 4 via Groq & Gemini 2.5 Flash/Pro where a user can input an article, press release, or other FOIA fodder and receive three high-quality request ideas.)

- Event context: Demos and Dim Sum with Deel and Apollo — 2025-06-03 — New York City
- Public talk page: https://nyc.aitinkerers.org/talks/rsvp_c-_XTku9cAI

### [Zero Trust-Proofing and Containerizing AI Agents for Enterprise Deployments](https://toronto.aitinkerers.org/talks/rsvp_9fibOugPX6M)

Identity Machines' SDK and CLI tool for AI agent developers, the Iron Book API for agentic multi-vector security, compliance, and privacy testing (vulnerability scanning with Claude Sonnet 3.7 on Bedrock and OSV, sentiment analysis, security scoring, foundational model and behavioral testing per OWASP LLM Top 10, cryptographic attestation, SBOM creation), AI agent packaging into an OCI-container deployed on EKS, and ongoing work with Google on trust-related enhancements for the A2A protocol (the Iron Vow mechanics).

- Event context: AI Tinkerers Toronto - May 2025 Meetup: AGENTS at Ada — 2025-05-22 — Toronto
- Public talk page: https://toronto.aitinkerers.org/talks/rsvp_9fibOugPX6M

### [What Happens When You Let AI Build a Game?](https://toronto.aitinkerers.org/talks/rsvp_D5YVCtG_VDs)

Over a period of 2 weeks, I used AI (mostly Cursor + Claude Sonnet 3.7 + ChatGPT o1 Pro) to build two playable browser-based games with Three.js and Nuxt.js. These weren’t polished, long-term projects, they were experiments in rapid prototyping through “vibe coding,” where I gave AI rough ideas and let it generate both the requirements and most of the implementation. In this 6-minute demo, I’ll show: - The actual games in action - The workflow I used with Cursor (messy prompts, generated code, and debugging side-by-side with AI) - How the AI tackled surprisingly complex problems like 3D scene setup, multiplayer sync, and game state transitions - What worked, what didn’t, and what it felt like to “co-develop” with an LLM I’ll walk through code, hacks, and surprises, all from the perspective of someone trying to build real things, fast, with AI in the driver’s seat.

- Event context: AI Tinkerers Toronto - April 2025 Meetup at Shopify — 2025-04-24 — Toronto
- Public talk page: https://toronto.aitinkerers.org/talks/rsvp_D5YVCtG_VDs

### [How to build a Gamemaster](https://seattle.aitinkerers.org/talks/rsvp_qAOZpkD0eMY)

Building a voice-based agent for *entertainment* is a huge challenge. Game development is a unique craft that most engineers struggle with: iterating to "find the fun" feels unintuitive. I present our D&D-like experience delivered by an AI Gamemaster, and share how combining game systems, human-authored content, and the craft of gamemastery, has resulted in a compelling experience that players spend hours enjoying.

- Event context: AI Tinkerers Seattle - March Meetup — 2025-03-28 — Seattle
- Public talk page: https://seattle.aitinkerers.org/talks/rsvp_qAOZpkD0eMY

### [Comparing Multimodal AI on Analytical Reasoning](https://boston.aitinkerers.org/talks/rsvp_j5G8xNtOHV0)

Many different deep learning models, especially LLMs, can be focused and fine-tuned on specific data modals. One very interesting feature of today's LLMs is multimodal input; this has its uses in encoding image, text, structured, and other sorts of data into the same embedding space. This talk will briefly go over the technical use cases of multimodal models by running an in-code comparison across different AI models and their ability to analyze and reason with data.

- Event context: AI Tinkerers Boston February 2025 — 2025-02-24 — Boston
- Public talk page: https://boston.aitinkerers.org/talks/rsvp_j5G8xNtOHV0

### [AWS Bedrock Claude Chatbot](https://austin.aitinkerers.org/talks/rsvp_U2hBZUvpX7Q)

I have recently gotten a powerful customizable open-source conversational AI agent project from AWS Japan up and running in an AWS private cloud. It is based on the Anthropic Claude 3.5 Sonnet foundation model. It includes the same UI that comes with the public Claude Pro subscription. But it adds web search and easy no-code ways of building a RAG context repository. I will give a brief tour of the principal capabilities.

- Event context: AI Tinkerers February Meetup: Community AI Demos - North Austin — 2025-02-14 — Austin
- Public talk page: https://austin.aitinkerers.org/talks/rsvp_U2hBZUvpX7Q

### [Scratchpad - working memory for LLM applications](https://nyc.aitinkerers.org/talks/rsvp_WeM1BWpMbuQ)

An experiment in overcoming catastrophic amnesia of LLMs while constraining token consumption. A non-vector-based human readable approach empowering models with a scratchpad for each project / conversation topic / context. How much value can you get out of having the LLM automatically update a persistent text area that is loaded into context when returning to the conversation in the future? Demonstration of a LLM-power (via API) app that retains knowledge about multiple contexts as demonstrated by a journaling and project management solution.

- Event context: Building with Claude | An evening with Anthropic — 2024-10-21 — New York City
- Public talk page: https://nyc.aitinkerers.org/talks/rsvp_WeM1BWpMbuQ

### [Claude.vim - full-featured SWE copilot in under 1000 lines of code](https://prague.aitinkerers.org/talks/rsvp_c-JyIHioR4c)

Claude 3.5 Sonnet knows how to use vim! (Controlling vim is just typing letters, after all.) And... you can do anything from vim: edit long files, manipulate code, navigate codebase, execute shell commands and snippets of python, open websites into buffers, and so on. Plus, vim can serve as a fully editable Claude chat window. Therefore, embedding Claude in vim means you get a fully agentic AI pair programmer in <1000 LoC. We will demo how you can get a full-featured AI chat as a UNIX hacker, use Claude as a pair programmer co-pilot in vim, and show the insane power of Claude tool use (something claude.vim relies on extensively, and you are NOT getting full potential out of Claude in your app if you don't use tools).

- Event context: AI Tinkerers Prague Launch Event - September 2024 — 2024-09-04 — Prague
- Public talk page: https://prague.aitinkerers.org/talks/rsvp_c-JyIHioR4c

### [Harmonising AI and Music: Playa Music's Innovative Playlist Generation and Discovery](https://paris.aitinkerers.org/talks/rsvp_D0RsV07-nFI)

We'll demonstrate two cutting-edge AI models that are revolutionizing music discovery and playlist creation: groove-scout-1.3: An advanced playlist search engine that generates unique album artwork. party-time-1.4: A generative playlisting engine, akin to a "PlaylistGPT". Our demo will showcase how we've integrated multiple language models (including Claude-3.5-Sonnet, GPT-4, phi3:medium, and others) with our proprietary Cynthia search engine to create a groundbreaking musical experience. We'll present live demonstrations of both models, highlighting the technical challenges overcome and the innovative solutions implemented.

- Event context: AI Tinkerers - Paris Inaugural Meetup featuring Mistral AI — 2024-09-03 — Paris
- Public talk page: https://paris.aitinkerers.org/talks/rsvp_D0RsV07-nFI

### [Transforming Code Collaboration with Aider: AI-Powered Pair Programming](https://fort-wayne.aitinkerers.org/talks/rsvp_fd5h26pQoac)

Meet Aider —an open-source tool that's redefining how developers work with large codebases. Aider seamlessly integrates with Git, leveraging AI models like GPT-4o and Claude Sonnet 3.5 to offer real-time code suggestions, multi-file edits, and automatic commits, all within your terminal. In my upcoming presentation, I'll be showcasing how Aider can be used to create real, functional code while demonstrating its powerful capabilities. We’ll explore how this tool can make the impossible possible, whether you're refactoring, fixing bugs, or adding new features. Check out Aider on GitHub and learn more at https://aider.chat 🌐

- Event context: AI Tinkerers Fort Wayne August 20th 2024 Meet up — 2024-08-20 — Fort Wayne
- Public talk page: https://fort-wayne.aitinkerers.org/talks/rsvp_fd5h26pQoac

### [websim](https://sf.aitinkerers.org/talks/rsvp_Xnz3wl4pw2E)

explore, create, and remix your hallucinated internet with websim

- Event context: AI Tinkerers - San Francisco - Summer Edition - July 2024 — 2024-07-12 — San Francisco
- Public talk page: https://sf.aitinkerers.org/talks/rsvp_Xnz3wl4pw2E

### [Open source Anthropic's Artifacts on steroids](https://sf.aitinkerers.org/talks/rsvp_7JFyTnsNFAs)

I'll show a demo of an open source template for building your own custom Artifacts UI - similar to what Anthropic has in the latest Claude version. On top of that, I'll show how to make these AI artifacts apps deployable. Features: - Sonnet 3.5 codegen & code execution - Deployable AI-generated apps - Vercel's AI SDK for streaming - Use custom LLM

- Event context: AI Tinkerers - San Francisco - Summer Edition - July 2024 — 2024-07-12 — San Francisco
- Public talk page: https://sf.aitinkerers.org/talks/rsvp_7JFyTnsNFAs

### [BAML: Using algorithms to guarantee JSON](https://seattle.aitinkerers.org/talks/rsvp_vCVU8mkZnms)

When using an LLM, its important to be able to parse the response reliably, many times JSON.parse doesn't work due to missing quotes, new lines, or when it does work, the exact schema may not be followed. You could have wanted an array of numbers, but gotten just 1. Using knowledge of your schema, BAML fixes those errors for you and provide a layer of guarantee, it will either be your schema or an exception will be raised. (0 additional LLM calls, runs locally in the package, built in rust) During the talk we'll see: 1. a few examples of BAML fixing the errors 2. walkthrough parts of the algorithm 3. Talk about edge cases like markdown and missing commas and how we address them

- Event context: AI Tinkerers Seattle - July 2024 Meetup — 2024-07-12 — Seattle
- Public talk page: https://seattle.aitinkerers.org/talks/rsvp_vCVU8mkZnms

## Related Technologies

- [Claude](https://aitinkerers.org/technologies/claude) ([Markdown](https://aitinkerers.org/technologies/claude.md)) — 173 public demos
- [Claude-3](https://aitinkerers.org/technologies/claude-3) ([Markdown](https://aitinkerers.org/technologies/claude-3.md)) — 110 public demos
- [Anthropic](https://aitinkerers.org/technologies/anthropic) ([Markdown](https://aitinkerers.org/technologies/anthropic.md)) — 36 public demos
- [Claude Code](https://aitinkerers.org/technologies/claude-code) ([Markdown](https://aitinkerers.org/technologies/claude-code.md)) — 215 public demos
- [React](https://aitinkerers.org/technologies/react) ([Markdown](https://aitinkerers.org/technologies/react.md)) — 219 public demos
- [AWS Bedrock](https://aitinkerers.org/technologies/aws-bedrock) ([Markdown](https://aitinkerers.org/technologies/aws-bedrock.md)) — 9 public demos
- [Gemini](https://aitinkerers.org/technologies/gemini) ([Markdown](https://aitinkerers.org/technologies/gemini.md)) — 188 public demos
- [GPT-4](https://aitinkerers.org/technologies/gpt-4) ([Markdown](https://aitinkerers.org/technologies/gpt-4.md)) — 529 public demos
- [GPT-4o](https://aitinkerers.org/technologies/gpt-4o) ([Markdown](https://aitinkerers.org/technologies/gpt-4o.md)) — 57 public demos
- [Grok 3](https://aitinkerers.org/technologies/grok-3) ([Markdown](https://aitinkerers.org/technologies/grok-3.md)) — 2 public demos
- [Haiku](https://aitinkerers.org/technologies/haiku) ([Markdown](https://aitinkerers.org/technologies/haiku.md)) — 7 public demos
- [JavaScript](https://aitinkerers.org/technologies/javascript) ([Markdown](https://aitinkerers.org/technologies/javascript.md)) — 40 public demos
- [LLM](https://aitinkerers.org/technologies/llm) ([Markdown](https://aitinkerers.org/technologies/llm.md)) — 123 public demos
- [Python](https://aitinkerers.org/technologies/python) ([Markdown](https://aitinkerers.org/technologies/python.md)) — 662 public demos
- [Three](https://aitinkerers.org/technologies/three) ([Markdown](https://aitinkerers.org/technologies/three.md)) — 13 public demos
- [ADK](https://aitinkerers.org/technologies/adk) ([Markdown](https://aitinkerers.org/technologies/adk.md)) — 5 public demos
- [AGENTS](https://aitinkerers.org/technologies/agents) ([Markdown](https://aitinkerers.org/technologies/agents.md)) — 20 public demos
- [AG-UI](https://aitinkerers.org/technologies/ag-ui) ([Markdown](https://aitinkerers.org/technologies/ag-ui.md)) — 4 public demos

## More Results

- Next: https://aitinkerers.org/technologies/sonnet.md?page=2