Pachinko pachinko
Pachinko pachinko
AI CORE FEATURES

Functions:
AI That Runs Every Time You Drop

Set up context and a prompt once. Drop new notes on it forever.

Pachinko Functions

① Select notes as context. ② Write a prompt for what the AI should do.
③ Create a function from context + prompt. ④ Drop new notes onto it - function fires automatically.

HOW FUNCTIONS WORK

Context + Prompt + Drop = Automatic

A function is a section with a brain. You select notes as context - your stock portfolio, your brand guidelines, your research, whatever the AI needs to understand. Then you write a prompt: "for each new article, tell me which positions are affected." This creates a function section. Now, every time you drag and drop a new note onto it - an article, a tweet, a YouTube transcript - the function fires automatically. Your context and prompt run against the new note and you get results without lifting a finger.

Function context is intentional AI context. When a Function runs, Pachinko sends the saved context, prompt, and dropped notes to the AI provider you configured. Hide from AI keeps projects out of automatic discovery, but notes you deliberately select from hidden projects can still be part of a Function's context.

function object PCK-RN / build path
01 context

Select the notes, references, examples, or rules the function should carry into every job.

02 prompt

Describe the repeatable work once, in the voice and format you want to keep using.

03 settings

Choose web search, thinking level, map, reduce, and whether raw inputs should stay around.

04 drop

Send new material to the function whenever you want the same job done again.

  • step 1 Select notes as context (your portfolio, guidelines, research)
  • step 2 Write a prompt ("for each new article, tell me...")
  • step 3 This creates a function section
  • step 4 Drop new notes onto it-articles, tweets, transcripts
  • step 5 The function fires automatically every time you drop
  • the point Set it up once. Use it forever.
  • privacy Saved context is sent when the Function runs, including hidden-project notes you selected on purpose
  • FUNCTION SETTINGS

    Fine-Tune How Your Function Processes Notes

    Functions have controls that change what they know, how hard they think, and how they process what you drop on them. Start with context, then choose the processing behavior that matches the job.

    function control legend PCK-RN / settings
    01

    Context

    The material the function carries into every job: notes, references, examples, portfolio holdings, brand rules, or anything else the prompt needs.

    Use when the answer should depend on your project, not generic model memory.
    02

    Extended Thinking

    Gives the AI more reasoning time before it writes. It is slower, but better for prompts that require comparison, judgment, or multi-step analysis.

    Use when the output needs a real decision, not a quick rewrite.
    03

    Web Search

    Lets the function look outside your workspace during each run, so time-sensitive questions can use current sources instead of old context.

    Use when prices, news, products, companies, or facts may have changed.
    04

    Map

    Runs the same prompt separately across each dropped item. Each article, note, job posting, or review gets its own pass.

    Example summarize each article in one sentence.
    05

    Reduce

    Processes the dropped items together as one set. The function can find patterns, conflicts, consensus, or a single answer across the batch.

    Example what do these five articles agree on?
    06

    Auto-Delete

    Trashes the dropped source after processing. Useful when the original item was just raw material and the output is what you want to keep.

    Use when a function is acting like a pipeline, not an archive.

    Reduce is the default. Map transforms items one by one. Combine web search, extended thinking, and auto-delete for a more automated pipeline.

    GETTING GOOD RESULTS

    Refine Your Prompt, Catch Bad Outputs

    Your first prompt won't be your best prompt. The point is to watch what your function produces and iterate. Hover over the function icon in your function section and an edit button appears - click it to refine the prompt. Drop a new note, see what comes back, adjust, repeat.

    AI gets things wrong. It hallucinates facts, misses obvious points, and sometimes ignores your instructions entirely. Since every function output lands as a note in your workspace, you can read them like any other note - spot check a few, delete the ones that are wrong, and tighten your prompt if you see a pattern of bad output.

  • iterate the prompt Hover the function icon → edit button → refine
  • spot check outputs Function results are notes - review them like any other note
  • bad output pattern? Tighten the prompt, add examples, or add a reference that clarifies
  • try extended thinking For complex prompts, extended thinking often produces much better results
  • WHAT PEOPLE BUILD WITH FUNCTIONS

    Set Up Once, Use Forever

    function library use cases / reusable jobs
    RN-01 Stock tracking portfolio context / market articles
    RN-02 Competitive intel competitor list / industry news
    RN-03 Newsletter draft voice samples / saved links
    RN-04 Job hunting resume context / job postings
  • stock tracking Context: your portfolio. Prompt: "which positions are affected?" Drop: market articles.
  • design evolution Context: your portfolio. Prompt: "how should I evolve my style?" Drop: design trend articles.
  • competitive intel Context: your competitors. Prompt: "what are they missing that we could own?" Drop: industry news.
  • health tracking Context: your blood work. Prompt: "transform this into something healthier for me." Drop: recipes.
  • newsletter Context: your voice and style. Prompt: "write me a newsletter draft." Drop: batches of saved tweets.
  • job hunting Context: your resume. Prompt: "write a tailored cover letter." Drop: job postings.
  • SHARE YOUR FUNCTIONS

    Export, Import, Swap

    Built a function that works? Export it and send it to someone. Got a friend's function? Import it and run it on your own notes. Functions are portable - the context + prompt combo travels as a single file.

    Review the context before you export. If the Function includes notes or sections from a project hidden with Hide from AI, those items travel in the export too.

  • export Right-click the function → Export Function
  • import File → Import Function...
  • what travels The prompt and the context notes or sections, together
  • privacy Check for private or hidden-project context before sharing an export
  • the point Good functions are worth sharing - and stealing
  • WALKTHROUGHS

    Build One, Then Automate It