What's New

Recent product updates

See what’s new in emergentic.ai, including new workflows, clearer tools, and fixes across the parts of the product you use most.

Latest Release Version 1.19.11

City places and movement tools

Choose a New York starting location, explore addresses with the city camera, and assign native walking and wayfinding tools to agents.

LatestNovember 7, 2026

New

Start a city world at an address, coordinates or a point on the map. The picker checks available 3D coverage; agents need outdoor walking space within 75 metres.

Search addresses and coordinates from the city camera’s jump menu. Distant views load while agents continue moving.

Improved

Native scene walking, city walking and wayfinding appear in world and agent action selection, with their execution requirements shown in action details.

City cameras frame tall buildings and preserve manual camera adjustments while a selected address loads.

Fixed

Coordinate movement respects movement settings and assigned tools, including permission changes made during route planning.

Returning to a background tab no longer reports a city timeout solely because the browser paused rendering.

City rendering preserves exposure and highlight detail with post-processing disabled. Procedural walkers keep a foot on the ground and their lighting follows their poses.

Recent Releases

Archive

A running feed of user-facing product updates.

Run budgets and clearer usage

Set a dollar budget for a simulation, compare service usage, and explore alternative topic analyses of the same run.

Version 1.19.10

November 4, 2026

New

  • Run budgets stop simulations when recorded cost reaches the limit. Calls already in progress can add to the final cost. Administrators can set each user’s maximum per run.
  • Account credits cover LLM calls, media generation, speech and transcription across browser and API requests.
  • Choose canonical, hierarchical or adaptive topic analysis, with separate saved results for each mode.

Improved

  • Appearance can follow your device’s light or dark setting. Choose System, Light or Dark in the account menu.
  • Sign in directly from the site header on desktop and mobile.
  • Updated provider disclosures, privacy and credit terms, including Modal hosting and the absence of automatic expiry for purchased credits.
  • Core API Usage shows service totals in a sortable table. Recent Logs stays open with sortable columns, precise timestamps, request durations and error details.
  • Run budget, agent memory and background execution have their own collapsible sections. Help text appears in tooltips, and selected agents appear as a compact avatar row.
  • Topic timelines include sortable views and recorded plot events, with links back to supporting messages.

Fixed

  • Removed credit purchases, subscription plans and team billing controls from Settings and account menus.
  • New World Designers use Gemini 3 Flash again. Existing designers retain their selected models.
  • Usage limits include metered speech, transcription and grounded web-search costs, and account credit queries avoid scanning full usage histories.

Gemini model updates

New requests using Gemini 2.5 Flash or Flash Lite run on Gemini 3.5 Flash Lite. Gemini 2.5 Pro requests run on Gemini 3.8 Flash.

Version 1.19.9

October 27, 2026

Improved

  • Existing worlds and clients can keep their saved model selections; new requests use the replacement model automatically.
  • Model menus and client examples use current Gemini choices.
  • Recorded studies and past usage retain their original model names and prices.

Editable documents and simpler agent studies

Agents can revise shared documents and scene objects. New client examples show how to build a world, run conversations and read selected regions of a shared canvas.

Version 1.19.8

October 27, 2026

New

  • Agents can search shared documents, read selected passages and make precise revisions, with earlier versions retained and conflicting edits detected.
  • World creation can include beliefs, motivations, desires and intentions as properties that agents update during a simulation.
  • A shared-canvas client tutorial sets up groups in different locations, runs a study and reads rectangular regions of pixels.

Improved

  • Agents can find scene objects by name, type, tag or location, inspect their details and change individual properties.
  • Client applications can start agent conversations without an open browser and reconnect using a simulation’s saved access settings.

Fixed

  • Interrupted client studies stop their simulations and release their connections. API timestamps include their UTC timezone.
  • Platform prompts remain private while your authored guidance stays editable.

Study galleries and replays

Between Machine and Machine and its recorded world and canvas timelines load again.

Version 1.19.7

October 15, 2026

Fixed

  • Restored missing recordings used by the gallery, population matrix and interactive study timelines.