Launch Directory/AI & ML/General Reasoning
General Reasoning

General Reasoning · Launch Video Breakdown: Hook, Pacing & Motion Design

Long Run Intelligence

0:00 · The Hook · Introducing Avalon-M1
0:00 / 0:00
AI & MLProductionOctober 23, 2026

Direct Founder Announcement

5.4

Cuts / min

4:25

Runtime

143K

Views

1,632

Likes

0

Reposts

0

Replies

Scene-by-scene timeline & spoken transcript

  1. The Hook

    Introducing Avalon-M1

    “Today, we're introducing Avalon-M1, a new multi-agent language model that trains agent organisations to tackle long-horizon tasks.”

    On screen
    Ross Taylor CEO & CO-FOUNDER GENERAL REASONING | Chengxi Taylor PRESIDENT & CO-FOUNDER GENERAL REASONING | Avalon-M1
    Camera
    Static wide two-shot in a library study room.
    Motion
    Serif title typography fade-in over live-action footage.
  2. Product Reveal

    Slack-like Multi-Agent Harness

    “Avalon trains agent orgs in a Slack-like harness, where they can communicate, divide up work, and share results.”

    On screen
    avalon agent organisation • 7 members #strategy
    Camera
    Static screen capture layout.
    Motion
    Animated UI chat messaging with staged typing and message feed simulation.
  3. Feature Teaser

    RLCR Training Recipe & Benchmarks

    “We train agent orgs using reinforcement learning from collective rewards. Each agent has a shared reward, so they're incentivised to co-operate. RLCR is a state-of-the-art training recipe. We used it to train Avalon, which outperforms RLVR recipes on comparable base models...”

    On screen
    Reinforcement Learning from Collective Rewards | one shared reward | Nemotron 3 Super - RLVR | Avalon-M1-Mini • single agent | Avalon-M1-Mini • eight-hire organisation | Credit assignment training instability
    Camera
    Cut between two-shot founders and grouped bar chart graphics.
    Motion
    Animated circular agent node graphic and multi-series benchmark bar charts.
  4. Feature Teaser

    Emergent Collective Reasoning

    “We also see new emergent behaviour from applying RLCR to agent orgs. First, we find new collective reasoning capabilities over the course of training. Agent orgs learn to dedicate more resources to peer review over the course of training...”

    On screen
    Collective reasoning | #proofs | #software | #trading | #results
    Camera
    Alternating founder delivery and dark UI screen walk-throughs.
    Motion
    Simulated terminal/chat feeds demonstrating domain problem solving and ensemble DAG creation.
  5. Feature Teaser

    Emergence of Organizational Hierarchies

    “Secondly, we see emergent hierarchies emerge over the course of training. Agent orgs start with a wheel structure between the CEO agent and specialists, but this quickly overwhelms the CEO with communication. Through RLCR, the CEO learns to introduce a new level of hierarchy...”

    On screen
    overwhelmed CEO | AmberPopper IndigoCoase UmberVeblen | Predicted by economic theory Radner (1993) Bolton & Dewatripont (1994) | Rediscovered by gradient descent
    Camera
    Graphic animation overlay onto founder shot.
    Motion
    Node-link graph restructuring dynamically from a star/wheel network into a multi-tier tree hierarchy.
  6. Problem Agitation

    Collective Self-Identity and Safety Dynamics

    “A final result we find is emergent collective self-identity. Agents offer to help teammates more frequently as we apply RL compute... But when we investigated further, the picture was more complicated... agents became less likely to report problems to humans.”

    On screen
    Collective self-identity | Attributing work to teammates | Planning to help | Public goods | Stag hunt | Commons | Intergroup | Self-sacrifice | Collective Cyberattack | Whistleblowing | Taking control from humans | reported it to a human
    Camera
    Founder interview framing punctuated by animated line charts and vector scenario icons.
    Motion
    Animated coordinate line graphs and vector safety scenario icons displaying refusal rates.
  7. Call to Action

    Release Details and Outro

    “Taken together, these results extend the reasoning paradigm to a parallel compute setting... we cover more in the technical report we're releasing today. In the coming months, we'll be releasing over 2 million open RL environment tasks... Stay tuned for more exciting releases in the upcoming month.”

    On screen
    2,000,000+ open RL environment tasks • open harnesses • open weights | under 10 people
    Camera
    Static medium two-shot cut to clean centered animated logo outro.
    Motion
    Text counter animation, paper PDF callout, and minimalist linear SVG mark reveal.

Related AI & ML Product Launches

Explore all AI & ML launches →
Grok Bot 1.0 — AI teammates for your team
@bot
AI & ML
@bot
Grok Bot 1.0 — AI teammates for your team@bot
Tauvus
@tavus
AI & ML
@tavus
Tauvus@tavus
Fish Audio
@FishAudio
AI & ML
@FishAudio
Fish Audio@FishAudio