# Claude-3 Projects at AI Tinkerers

> Canonical HTML: https://aitinkerers.org/technologies/claude-3?page=3
> Markdown URL: https://aitinkerers.org/technologies/claude-3.md?page=3
> Technology record last updated: 2026-02-22T17:01:16Z
> Generated: 2026-08-25T10:16:49Z

Claude-3 is Anthropic's state-of-the-art multimodal model family (Opus, Sonnet, Haiku), setting new industry benchmarks for intelligence, speed, and vision capabilities.

Claude-3, developed by Anthropic, is a powerful family of three generative AI models: Opus, Sonnet, and Haiku. Opus, the flagship, excels in complex reasoning, outperforming peers on key benchmarks (MMLU, GPQA) and supporting a 200,000-token context window. Sonnet offers an optimal balance for enterprise workloads, delivering performance that is 2x faster than its predecessor, Claude 2.1. Haiku is the fastest and most cost-effective option, capable of processing a 10,000-token research paper (including charts) in under three seconds. All three models are multimodal, featuring strong vision capabilities for analyzing charts, diagrams, and PDFs alongside text, enabling advanced data extraction and analysis.

- Official technology site: https://claude.ai
- Public AI Tinkerers demos and talks: 110
- Result page: 3 of 5

## Recent Public Talks and Demos

### [La Evolución Real de un NUEVO Producto de IA desde Customer Success hasta la Toma de Decisiones de Alto Nivel](https://santiago.aitinkerers.org/talks/rsvp_dOtJPRuPVQ8)

El motor de con IA de una empresa tecnológica que gestiona más de mil millones de dólares ha evolucionado en la integración de la Inteligencia Artificial: desde un primer chatbot inteligente enfocado en atención al cliente, pasando por un co-pilot colaborativo para desarrolladores, hasta el diseño de un nuevo producto de IA que hoy empodera a analistas senior de portafolio de inversión. Con herramientas como LLaMA, Claude, Code y capas sobre sistemas legacy, veremos cómo se construye una solución que permite a los hedge funds tomar decisiones críticas en tiempo récord, con precisión, adaptabilidad y escalabilidad real.

- Event context: 🔥 AI TINKERERS – SANTIAGO / CAPÍTULO 6 — 2025-07-31 — Santiago
- Public talk page: https://santiago.aitinkerers.org/talks/rsvp_dOtJPRuPVQ8

### [BPOs in the future post Agentic AI Era](https://delhi.aitinkerers.org/talks/rsvp_cp_pegn5qrU)

This talk explores how Agentic AI is replacing traditional BPOs with cloud-native, autonomous voice infrastructure; delivering scalable, multilingual CX without human agents. Learn why India can lead the next wave of AI-driven outsourcing.

- Event context: AI Tinkerers Delhi - July Meetup — 2025-07-26 — Delhi
- Public talk page: https://delhi.aitinkerers.org/talks/rsvp_cp_pegn5qrU

### [Making-llm-reliable](https://delhi.aitinkerers.org/talks/rsvp_PF6Pb_W1i70)

How to make llm reliable

- Event context: AI Tinkerers Delhi - July Meetup — 2025-07-26 — Delhi
- Public talk page: https://delhi.aitinkerers.org/talks/rsvp_PF6Pb_W1i70

### [Multiple Claude Code collaborating as multi-agent](https://seattle.aitinkerers.org/talks/rsvp_fL0yMDL791U)

This talk showcases how we run multiple Claude Code agents using an open-source MCP (Model Context Protocol) setup — not as a demo, but as production infrastructure. Each agent runs inside a container, with file-mounted context, state, and config. They operate on real tasks (browser automation, LinkedIn browsing, competitive research) and coordinate via pure natural language prompts — just like humans chatting. All logs and exchanges are streamed into a shared chatroom, enabling review, human intervention, or agent-to-agent negotiation. No plugin wrappers. No forced APIs. Just prompt-level orchestration that works.

- Event context: Building AI Agents with Google Cloud AI — 2025-07-25 — Seattle
- Public talk page: https://seattle.aitinkerers.org/talks/rsvp_fL0yMDL791U

### [AI Cloud Architecture Diagram Generator](https://houston.aitinkerers.org/talks/rsvp_sFaClewvj4I)

AI Cloud Architecture Diagram Generator

- Event context: AI Tinkerers - Houston Inaugural Meetup (July) — 2025-07-02 — Houston
- Public talk page: https://houston.aitinkerers.org/talks/rsvp_sFaClewvj4I

### [Swarm of AI coding agents divide-and-conquer your project while sharing knowledge through an intelligent memory bank](https://boston.aitinkerers.org/talks/rsvp_JJe3Ej1Ei_A)

Imagine coordinating dozens of AI agents simultaneously, each working on different aspects of your project while sharing knowledge through an intelligent memory bank. The initial version of Wakala-v2 was built autonomously using claude-sparc, a shell script that leverages Roo Code and Claude Code, but recent development was done using claude-flow, the ultimate multi-terminal orchestration platform that revolutionizes how you work with Claude Code. What is Wakala, you may ask? In many African countries, WhatsApp *is* the internet for the vast majority of the population, and South Africa is no exception. This is the way most South Africans conduct business, book appointments, pay for services and merchandise, and look up information. Wakala is a WhatsApp-native assistant designed to empower Cape Town township residents by providing easy-to-use, data-light tools for running daily businesses, accessing vital services, and learning new skills directly through WhatsApp. Wakala's mission is to lower the barriers to digital and economic participation for township residents in Cape Town. It offers accessible, affordable, and multilingual functionalities for financial transactions, business management, logistics, information access, and skill development.

- Event context: AI Tinkerers Boston June Meetup @ MIT! — 2025-06-30 — Boston
- Public talk page: https://boston.aitinkerers.org/talks/rsvp_JJe3Ej1Ei_A

### [Mining opportunities](https://santiago.aitinkerers.org/talks/rsvp_ktST36XMY-k)

We will present Evalitics, an AI-powered platform that helps suppliers in the mining and construction industries pre-evaluate whether they’re truly capable of fulfilling a public tender before they apply. The platform reads unstructured procurement documents and supplier profiles, cross-checks missing or inconsistent information, and uses LLMs like GPT-4 to generate a “match score” and actionable feedback. All through an intuitive interface designed for non-technical users. We’ll walk through: a) How we extract and structure public data using web scraping. b) How we use semantic matching and risk simulation to detect execution problems before they happen. c) A preview of our conversational AI assistant that answers supplier questions based on current and past tender rules.

- Event context: AI Tinkerers-Santiago Cap.#5 (Humano 95%) + (5% IA) — 2025-06-26 — Santiago
- Public talk page: https://santiago.aitinkerers.org/talks/rsvp_ktST36XMY-k

### [Patch Party: Live-Fixing LLM Agents](https://london.aitinkerers.org/talks/rsvp_-7V7yhBZXMQ)

This is a live demo of a feedback loop that patches autonomous agents in real time. We’ll show how we built a minimal actor-critic framework that catches step-level agent failures—like hallucinated facts, logic errors, or missed tool calls—and injects precise critiques to recover mid-task. We'll walk through: How we used τ‑Bench and DA-Code to build a taxonomy of agent failure types How we tagged real traces and analyzed which errors matter most (e.g., reasoning failures > tool bugs) How we implemented the critic loop using FastAPI + GPT-4o or Claude as the agent, and optional models/humans as critics How a 1-2 sentence critique can improve success rates by 30%, without replanning or retraining How to plug this loop into your own stack using just a trace, a tagger, and a hook We’ll live-debug an agent solving a DA-Code task, show failure in action, and patch it with a model-generated critique. It’s messy, practical, and shows how real-time judgment can boost reliability with minimal overhead.

- Event context: AI Tinkerers London - June Meetup — 2025-06-25 — London
- Public talk page: https://london.aitinkerers.org/talks/rsvp_-7V7yhBZXMQ

### [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

### [Pokémon rl 2 gen 3](https://sf.aitinkerers.org/talks/rsvp_RHvzJtHRnLk)

Use rl to play pokemon emerald using transfer learning from pokemon on red

- Event context: How It’s Made: Architecting Planning-Based AI Systems ft. AI21 Maestro — 2025-05-23 — San Francisco
- Public talk page: https://sf.aitinkerers.org/talks/rsvp_RHvzJtHRnLk

### [Generating & Self-Healing Custom Integrations with Laminar](https://toronto.aitinkerers.org/talks/rsvp_9aTVCW9uXwA)

Building an integration workflow with on Laminar from prompts. Handling workflow errors and letting AI self-heal the integrations when errors occur.

- Event context: AI Tinkerers Toronto - May 2025 Meetup: AGENTS at Ada — 2025-05-22 — Toronto
- Public talk page: https://toronto.aitinkerers.org/talks/rsvp_9aTVCW9uXwA

### [SlateFront AI: A new way to learn with AI, visually.](https://abu-dhabi.aitinkerers.org/talks/rsvp_GkYfScCkcU0)

SlateFront is an AI-powered tool that generates mathematical animations and full explainer videos for math and physics concepts instantly, from just text or drawings.

- Event context: AI Tinkerers - Abu Dhabi Meetup #4 (May 2025) — 2025-05-22 — Abu Dhabi
- Public talk page: https://abu-dhabi.aitinkerers.org/talks/rsvp_GkYfScCkcU0

### [How to build an MCP server](https://nyc.aitinkerers.org/talks/rsvp_a5SrCxX0ofk)

Explaining MCP and demoing our platform for building and managing MCP servers.

- Event context: How It’s Made: Architecting Planning-Based AI Systems ft. AI21 Maestro — 2025-05-19 — New York City
- Public talk page: https://nyc.aitinkerers.org/talks/rsvp_a5SrCxX0ofk

### [A Realtime Vocal Assistant on an ESP32](https://paris.aitinkerers.org/talks/rsvp_J1qlamXmC3M)

Hi, I have built a Realtime Vocal Assistant that input/output from an ESP32 board, and Langchain agent running on local node server. Your local Alexa but 100 times better.

- Event context: AI Tinkerers - Paris Meetup on May 15th — 2025-05-15 — Paris
- Public talk page: https://paris.aitinkerers.org/talks/rsvp_J1qlamXmC3M

### [Placeholder: The Open Robotics Stack for the AI Era](https://paris.aitinkerers.org/talks/rsvp_AFA7fC7oUSo)

Placeholder is a collaborative platform redefining how robots are built, shared, and scaled. It brings together everything robotics teams need—hardware files, configurations, compatible components like sensors, documentation, and data—into a single, versioned, cloud-based workspace. Placeholder simplifies sourcing by showing where to buy parts, full kits, or download 3D-printable components, making it seamless to go from concept to physical prototype. It integrates directly with GitHub for code and control software, and with Hugging Face for hosting related datasets. Users can keep designs private or open-source, fork robots, track changes to both hardware and software, and share modular builds. By unifying hardware, software, and data, Placeholder enables scalable and accessible robotics development for teams, researchers, and the open-source community.

- Event context: AI Tinkerers - Paris Meetup on May 15th — 2025-05-15 — Paris
- Public talk page: https://paris.aitinkerers.org/talks/rsvp_AFA7fC7oUSo

### [nanoVLM: Build a VLM in pure pytorch in <750 lines](https://paris.aitinkerers.org/talks/rsvp_FTf3k9qqBR0)

I'll be presenting nanoVLM, a minimal, open-source PyTorch library for training Vision-Language Models (VLMs) from scratch in just ~750 lines of code. Inspired by nanoGPT, nanoVLM is simple, readable, and efficient — achieving competitive performance (35.3% on MMStar) with just 6 hours of training on a single H100 GPU. It combines a SigLiP-ViT encoder and LLaMA-style decoder, and is light enough to run in a free Google Colab.

- Event context: AI Tinkerers - Paris Meetup on May 15th — 2025-05-15 — Paris
- Public talk page: https://paris.aitinkerers.org/talks/rsvp_FTf3k9qqBR0

### [Using State Machines to Orchestrate Multi-Agent Systems](https://paris.aitinkerers.org/talks/rsvp_2_IfqwaoeBQ)

I'll demo the framework we built for a Deep Research style agent focused on statistical modelling and analysis. The framework is based on a state machine approach - to illustrate this I'll show how the state machine evolves in our command console as the agent goes about solving a task. This is a follow up from a demo I gave at a previous Tinkeres on an agent for building custom machine learning models - at the time the framework was quite new but it has since evolved significantly.

- Event context: AI Tinkerers - Paris Meetup on May 15th — 2025-05-15 — Paris
- Public talk page: https://paris.aitinkerers.org/talks/rsvp_2_IfqwaoeBQ

### [SPEAKER: Vaibhav, CEO of Boundary](https://nyc.aitinkerers.org/talks/rsvp_MRaxIRyS5a0)

Vaibhav CEO of Boundary and creator of BAML

- Event context: Advanced AI Engineering Camp with 🦄 ai that works — 2025-05-10 — New York City
- Public talk page: https://nyc.aitinkerers.org/talks/rsvp_MRaxIRyS5a0

### [GenAI Poker Coach App](https://las-vegas.aitinkerers.org/talks/rsvp_onikQ6wLQeM)

I am building an app that uses reinforcement learning to model poker, and an LLM to chat and coach you on your play in a personalized manner. I am coding in python This project is in early phase of development. I will show my process of scoping and project, making a roadmap for it, exciting an MVP, and iterating to create a good working product with product market fit. I will talk about the mathematics and why I'm using Reinforcement Learning and LLMs for the specific chosen capabilities of the bot and why these are the appropriate mathematical frameworks for the problems I'm looking to solve. I will also show you my sloppy python code.

- Event context: AI Tinkerers Las Vegas - May Inaugural Meetup — 2025-05-09 — Las Vegas
- Public talk page: https://las-vegas.aitinkerers.org/talks/rsvp_onikQ6wLQeM

### [Funcformer: Generalizing Transformers with Applicative Functors](https://milan.aitinkerers.org/talks/rsvp_GEWuU9IVwJU)

This talk presents a systematic approach to extending Transformer models beyond sequence data. Using applicative functors, we derive a generalized Transformer framework that can operate on arbitrary data structures. We demonstrate a step-by-step Haskell derivation that reconstructs Transformer architecture abstractly, showing how attention mechanisms and feedforward networks can be generalized. This derivation provides the mathematical foundation for practical implementations in Python and other ML frameworks. If time allows, we will present a reference model, "Funcformer," which operates on continuous functions represented as Chebyshev polynomials, showing connections to neural operators from the literature on learning PDE solutions. This abstraction enables Transformers to process new data types and creates opportunities for implementing cross-modal models that can simultaneously handle different data structures using consistent underlying principles. This talk is based on: https://cybercat.institute/2025/02/12/transformers-applicative-functors/

- Event context: AI Tinkerers Milan - May 8, 2025 — 2025-05-08 — Milan
- Public talk page: https://milan.aitinkerers.org/talks/rsvp_GEWuU9IVwJU

### [Creating Your AI Benchmarks in Minutes](https://dc.aitinkerers.org/talks/rsvp_4o03sWmDJDE)

I am going to create a fun and live AI benchmark using RELAI's data agents and evaluate a Gemini model on it!

- Event context: Google Cloud AI Tinkerers - DC Metro Meetup (May 7th, 2025) — 2025-05-07 — DC
- Public talk page: https://dc.aitinkerers.org/talks/rsvp_4o03sWmDJDE

### [Real‑Time AI Coach for Mock Product Manager Interviews](https://dc.aitinkerers.org/talks/rsvp_rgdgm5jyLhE)

My wife was recently on the job market, and wanted to get more practice doing some of the specific interview types often required for Product Manager roles. Using online resources about PM interviewing along with the OpenAI Realtime API, I put together an application that would come up with practice interview topics and perform the interview via voice. Then I used several different LLMs to grade the interview and provide diverse perspectives on how she did, and what she could improve on in the future. I will do a short live demo of the interview functionality and the post-interview grading, and walk through the code to show how I implemented the different pieces.

- Event context: Google Cloud AI Tinkerers - DC Metro Meetup (May 7th, 2025) — 2025-05-07 — DC
- Public talk page: https://dc.aitinkerers.org/talks/rsvp_rgdgm5jyLhE

### [Real-Time Sentiment Analysis of YouTube Live Chats with LLMs](https://montreal.aitinkerers.org/talks/rsvp_zfsW81aYAMY)

Harnessing the power of Large Language Models (LLMs) to gain instant insights into the emotional tone of YouTube Live Chat conversations as they unfold. This demo provides real-time sentiment analysis, enabling a deeper understanding of audience reactions and engagement during live streams.  

- Event context: AI Tinkerers Montreal - May 2025 Meetup at LightSpeed — 2025-05-07 — Montreal
- Public talk page: https://montreal.aitinkerers.org/talks/rsvp_zfsW81aYAMY

### [I am coding the same agents in all LLM frameworks to compare them side-by-side](https://amsterdam.aitinkerers.org/talks/rsvp_BPdGQe_Nw1c)

"I am coding the same agents in all LLM frameworks to compare them side-by-side: - DSPy - Langgraph - Google ADK - PydanticAI - InspectAI - No framework and more..." Want to build your own local AI agent app? In this session, I’ll show how I wired together a simple full-stack agent playground to build AI agents in different frameworks. We’ll dive into the architecture, look at the streaming layer, and talk about what went right (and weird) while building them. No slides. No marketing. Just code, glitches, and lessons from messing around with agent stacks for fun. It's based on this X post I posted this week: https://x.com/_rchaves_/status/1914083928647885122

- Event context: AI Tinkerers Amsterdam - May Edition — 2025-05-06 — Amsterdam
- Public talk page: https://amsterdam.aitinkerers.org/talks/rsvp_BPdGQe_Nw1c

## Related Technologies

- [GPT-4](https://aitinkerers.org/technologies/gpt-4) ([Markdown](https://aitinkerers.org/technologies/gpt-4.md)) — 529 public demos
- [Llama-2](https://aitinkerers.org/technologies/llama-2) ([Markdown](https://aitinkerers.org/technologies/llama-2.md)) — 227 public demos
- [LangChain](https://aitinkerers.org/technologies/langchain) ([Markdown](https://aitinkerers.org/technologies/langchain.md)) — 444 public demos
- [OpenAI API](https://aitinkerers.org/technologies/openai-api) ([Markdown](https://aitinkerers.org/technologies/openai-api.md)) — 518 public demos
- [PyTorch](https://aitinkerers.org/technologies/pytorch) ([Markdown](https://aitinkerers.org/technologies/pytorch.md)) — 273 public demos
- [Transformers](https://aitinkerers.org/technologies/transformers) ([Markdown](https://aitinkerers.org/technologies/transformers.md)) — 148 public demos
- [Sonnet](https://aitinkerers.org/technologies/sonnet) ([Markdown](https://aitinkerers.org/technologies/sonnet.md)) — 25 public demos
- [Python](https://aitinkerers.org/technologies/python) ([Markdown](https://aitinkerers.org/technologies/python.md)) — 654 public demos
- [Gemini](https://aitinkerers.org/technologies/gemini) ([Markdown](https://aitinkerers.org/technologies/gemini.md)) — 187 public demos
- [React](https://aitinkerers.org/technologies/react) ([Markdown](https://aitinkerers.org/technologies/react.md)) — 216 public demos
- [Anthropic](https://aitinkerers.org/technologies/anthropic) ([Markdown](https://aitinkerers.org/technologies/anthropic.md)) — 36 public demos
- [GPT-4o](https://aitinkerers.org/technologies/gpt-4o) ([Markdown](https://aitinkerers.org/technologies/gpt-4o.md)) — 57 public demos
- [AWS Bedrock](https://aitinkerers.org/technologies/aws-bedrock) ([Markdown](https://aitinkerers.org/technologies/aws-bedrock.md)) — 9 public demos
- [FastAPI](https://aitinkerers.org/technologies/fastapi) ([Markdown](https://aitinkerers.org/technologies/fastapi.md)) — 178 public demos
- [Llama 3](https://aitinkerers.org/technologies/llama-3) ([Markdown](https://aitinkerers.org/technologies/llama-3.md)) — 38 public demos
- [AI21 Labs API](https://aitinkerers.org/technologies/ai21-labs-api) ([Markdown](https://aitinkerers.org/technologies/ai21-labs-api.md)) — 1 public demo
- [AI agent](https://aitinkerers.org/technologies/ai-agent) ([Markdown](https://aitinkerers.org/technologies/ai-agent.md)) — 8 public demos
- [AI agents](https://aitinkerers.org/technologies/ai-agents) ([Markdown](https://aitinkerers.org/technologies/ai-agents.md)) — 35 public demos

## More Results

- Previous: https://aitinkerers.org/technologies/claude-3.md?page=2
- Next: https://aitinkerers.org/technologies/claude-3.md?page=4