Urgent.News

What's breaking now, across thousands of outlets.

AI

The Missing Interface in AI-Generated Animation Is seek(t)

A deterministic seek(t) interface makes AI-authored animation reproducible, testable, and easier to edit, whether written by Opus 5.5 or another coding model.

The Missing Interface in AI-Generated Animation Is seek(t)

A deterministic scene contract is crucial for AI-generated animation, regardless of whether a language model "made a video" or merely wrote code for a scene that a renderer then turned into frames. The key contract is a function, seek(t), which returns the frame that belongs at time t. This allows for reliable export and jumping to any timestamp.

To implement this, the model should generate a self-contained HTML scene with a declared duration and an asynchronous seek function that clears the canvas, calculates the state for a given time, and draws it. The renderer should own the frame rate, awaiting completion before taking a screenshot. The scene should be tested by calling seek(t) twice and comparing the resulting frames.

The example provided uses functions of time rather than accumulated mutations to control motion, ensuring repeatability and avoiding issues related to randomness or autonomous animation loops.

Written by urgent.news from HackerNoon's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at hackernoon.com →

More in AI

sifter: a 23 MB model that sorts your expenses locally, for free

There is a kind of chore that is too small to hire someone for and too repetitive to keep doing by hand. Filing expense lines. Sorting a folder of customer messages. Tagging inventory notes.

  • sifter is a local text sorting tool for expense categorization
  • Users provide labeled examples to train the open-weight model
  • sifter operates offline, free on user's machine, 23 MB download

Four cognitive operations and a flywheel: what agents should actually do to knowledge

Most agent systems answer one question: how do I get the model to write something for me? That is the wrong first question. The right one is: what kind of cognitive work should the agent do at all?

  • Four cognitive operations: systematize, theorize, academicize, systematize again as architecture
  • Flywheel process: documents to structured nodes, extraction, pattern discovery, challenge
  • Agent as editor, not author; gate questions before each batch of work

Cubicle v0.22: a wider office, approvals you can read, and four new agent feeds

Cubicle shows your AI agents as people in a pixel-art office: they walk to a desk when they start working, raise a hand when they need you, and go for coffee when they’re done.

  • Cubicle office expanded with additional desk wings matching window width
  • Approvals feature improved to show title, summary, and risks of raised hands
  • Reduced false alarms with "Last run failed" label removal and "needs you" marking

More from Monday 12 October →