upcoming ai innovation

mainline intelligence

is coming — the new wave of pyx.

mi is our next ai platform — the umbrella for breakthrough systems that push past today’s assistants into realtime, industrial-grade intelligence. same mainline promise: premium experiences, free to reach.

coming soon

beta · api preview

mainline artificial realtime instant intelligence (marii)

instant when it matters. realtime when the world moves. this is mainline’s leap into ai that keeps up — responses that feel native to the moment, built for creators and players who will not wait on a spinner.

instant responses. real data. all free.

first public beta: pyx assistant (powered by marii) — local-first in the browser, no cloud llm. extremely early and still being improved. path api preview: /api/marii/ask/<request>. read the honest overview: about pyx assistant. browse all betas at /betas.

related: announcer — live mlb play-by-play with game-scoped marii ask.

beta · on-device

MCU — mainline conversational unit

next-gen industrial pyx. a local model built for a 1–3.5B token diet. talk to it, train it, or call it from your own code — same JSON API as an LLM, no key, no cloud run. open MCU → · MCU ask · how to use it

MCU ask

industrial pyx as an API. GET a question. POST { q, docs } to send facts with the request. same on-device model as MCU. no cloud run, no cloud llm.

curl -sS "https://mcu-ask.mainline-mi.workers.dev/what%20is%20mcu"

via the site: curl -sSL "https://pyx-ai.web.app/api/mcu/ask/what%20is%20mcu" · every language: mcu-ask.php

marii ask

same local engine as pyx assistant — sports, math, notebook, weather, wiki. no cloud run, no llm. works from a browser or a terminal anywhere.

curl -sS "https://marii-ask.mainline-mi.workers.dev/how%20is%20judge"

same thing via the site (follow redirects): curl -sSL "https://pyx-ai.web.app/api/marii/ask/how%20is%20judge" · how to use it in your code (many languages): marii-ask.php

same path as the assistant. try “how is judge”, “15% of 80”, or “what is marii”.

// response shows up here
MariiAsk.ask("how is judge").then((data) => {
  // { ok, request, reply, source: "sports", backend: "pyx-assistant" }
  console.log(data.reply);
});

// terminal, anywhere:
// curl -sS "https://marii-ask.mainline-mi.workers.dev/how%20is%20judge"

mi mailing list

drop your email for mi updates — moderator tips, marii/MCU drops, and the occasional “we shipped something weird.” latest: issue 003 (marii ask now runs on pyx assistant).

no, we won’t make it ads. we’ve already done enough crimes against your attention span.

welcome mail comes from no-reply@pixelplaceofficial.com · replies hit support

mi moderator

live now. hit the path with your text. you get something like {"appropriate":false,"score":"700"}. score is 0–1000. default threshold is 700.

get https://pyx-ai.web.app/moderator/check/<text>

lower = stricter. higher = chill.

// response shows up here

add it to your game

one get. look at appropriate, or compare score to your own threshold.

  1. url-encode the text and call /moderator/check/<text>
  2. optional: ?threshold=700 so the api sets appropriate for you
  3. if appropriate is false, block or ask them to rephrase
  4. tune the threshold per mode
fetch("https://pyx-ai.web.app/moderator/check/" + encodeURIComponent(text) + "?threshold=700")
  .then(async (r) => {
    const raw = await r.text();
    try { return JSON.parse(raw); }
    catch { throw new Error(r.status + " " + raw.slice(0, 80)); }
  })
  .then((data) => {
    // { "appropriate": false, "score": "700" }
    if (!data.appropriate) {
      // block or rewrite
    }
  });

dev’s corner

hooks, trainers, and studio tools for pyx + mi. poke around, wire stuff up, or try MCU.

MCU / mi list

MCU is live as an on-device beta. want peeks at the rest of the wave? join the mi mailing list — same inbox, zero ad billboards.