rojkind arquitectos · ai workshop · topic 2
from chatgpt to an agent in rhino
The full workshop material: every part can be read as a guide or presented full screen. The glossary with every term is at the end.
contents
- 00Opening
What the workshop is about and how it is organized.
- 01Demystifying AI
Companies, models and apps; what an LLM is; reasoning models; how to classify a model; why AI costs money.
- 02Agents, context, and cost
The agent loop, harness parts, the context window, tokens and pricing, ways to pay.
- 03Connecting agents to Rhino
What MCP is, the Rhino MCP server, and Lattice with its own bridge.
- 04Code literacy
How a script talks to Rhino, the six pieces of any program, and GitHub.
glossary
- Agentpart 2 →
- A model inside a program that gives it tools and a loop: it decides a step, runs it, reads the result and continues until the task is done.
- APIpart 1 →
- The door that lets a program use a model directly, paying per token, without going through the chat.
- Branchpart 4 →
- A parallel line of work, on a copy, to try things without affecting the main one.
- Clone, push, pullpart 4 →
- Copying the repository down, sending your changes up, and bringing others' changes in.
- Commitpart 4 →
- A saved version of the code, with a note on what changed.
- Company (lab)part 1 →
- Builds models and sells access to them. Examples: OpenAI, Anthropic, Google.
- Condition (if)part 4 →
- Doing something only when a condition is met.
- Connectionpart 1 →
- How an agent acts on another program: MCP or a direct bridge.
- Context windowpart 1 →
- The maximum amount of text, in tokens, a model can read in one request.
- Dictionarypart 4 →
- A table of key → value pairs, like named size → radius.
- Edge casepart 4 →
- An unusual situation the code must account for, like overlapping circles or a nested layer.
- Frontier modelpart 1 →
- The best model a company has at a given time: the most capable and the most expensive.
- Functionpart 4 →
- A named, reusable recipe: it takes data and returns a result.
- Hallucinationpart 1 →
- A plausible, confident, but false answer. It happens because the model reconstructs; it doesn't look things up.
- Harnesspart 1 →
- Everything wrapped around a model to make it useful: memory, instructions, tools and a loop. ChatGPT and Lattice are harnesses.
- Inferencepart 1 →
- Using a trained model: from your text, it predicts the next piece, one at a time.
- Input and output tokenspart 2 →
- Input: everything sent to the model, including the context. Output: what the model writes. Output costs several times more per token.
- Latticepart 3 →
- A desktop agent and code editor for Rhino. It connects to Rhino through its own bridge and uses models through OpenRouter.
- Listpart 4 →
- An ordered collection of values, like the list of circle centers.
- Loop (for)part 4 →
- Repeating the same instructions for each item, like drawing a circle at each center.
- MCP (Model Context Protocol)part 3 →
- A standard for connecting agents to programs. The program publishes a menu of tools; the agent reads it, orders, and receives results.
- MCP serverpart 3 →
- The small program that exposes a piece of software's actions (for example Rhino) as MCP tools.
- Memorypart 2 →
- Notes the harness saves and reads back between sessions. The model by itself remembers nothing.
- Modalitiespart 1 →
- The kinds of input a model understands: text, images, audio, video.
- Model (LLM)part 1 →
- A file of numbers (weights) that predicts the next piece of text. Examples: GPT, Claude, Gemini.
- Open weightspart 1 →
- A model whose weights file can be downloaded and run on your own machine. Often called “open source”, even when the training data and code are not public.
- Pull requestpart 4 →
- A proposal to merge a branch into the main line, with review.
- Reasoning modelpart 1 →
- A model that writes intermediate steps (a scratchpad) before the final answer. Better at multi-step problems; costs more and takes longer.
- Repositorypart 4 →
- A project's folder on GitHub, with the full history of every change.
- RhinoCommonpart 4 →
- The (.NET) programming interface Rhino is built on. Used from Python, C# and Grasshopper.
- rhinoscriptsyntaxpart 4 →
- A simpler Python library built on top of RhinoCommon: one line per action, less complete.
- Router (OpenRouter)part 1 →
- A service that gives access to models from many companies with one account and one key.
- Skillpart 2 →
- A package of instructions for one specific job. The agent only sees its title and opens it when the task needs it.
- System instructionspart 2 →
- Fixed rules the agent receives before every task: its role, its style, and what it must never do.
- Tokenpart 1 →
- The piece of text a model works with: a short word or part of a longer one. Prices are charged per token.
- Toolpart 2 →
- An action the agent can request: read a file, run a script, query Rhino. The model asks for it in writing; the harness runs it.
- Trainingpart 1 →
- Adjusting the weights so the model reproduces a huge set of texts as well as possible. Done once, at great cost.
- Variablepart 4 →
- A name that holds a value, for example radio = 1.0.
- Weights (parameters)part 1 →
- The numbers a model is made of, like the control points of a surface. A large model has tens or hundreds of billions.
- Workspacepart 2 →
- The files and folders an agent can see and change.