TRACK 1

🔁 RSI fundamentals

Before discussing benchmarks, dates, and scenarios, understand the idea behind the whole warning: recursive self-improvement (RSI)—AI helping design the next AI, faster. Learn how the loop works, why it appears first in code, and how to separate solid evidence from speculation.

Claude N today’s generation writes tests · debugs designs Claude N+1 better · faster ↻ and helps build Claude N+2—each cycle tends to move faster

How to read it: the current generation (emerald) drives work (cyan: write, test, design) that produces the next generation. The lower arrow shows both the exciting and unsettling part: the result feeds back into the input. That is RSI.

Conceptual illustration: one AI generation designing the next in an ascending spiral
The warning’s visual metaphor: this is not a conscious robot waking up, but a self-reinforcing engineering process that tends to accelerate. Track 1 explains what that really means.
2
Modules
12
Topics
~1h30
Duration
Beginner
Level
Track 1 progress0 of 12 · 0%

Track map

Detailed content

1.1~45 min · 6 topics

🔁 The loop: AI building AI

The warning’s central idea in plain English: what RSI is, where the bottleneck shifts, and why this is a warning, not a reason to panic.

0 of 6
What it is:

RSI is when AI helps design a better AI, which designs the next one, faster. Jack Clark’s (Anthropic) metaphor: “Claude 10 building Claude 11.”

Why learn this:

This is the idea behind the whole course. Without it, the figures in later tracks (METR task horizon, MirrorCode) lack context.

Key concepts:

RSI; frontier model; the self-reinforcing loop.

What it is:

Today, progress is limited by how quickly people can think, test, and code. With RSI, the limits shift to compute, infrastructure, and autonomy.

Why learn this:

Understanding the bottleneck explains why “more chips and energy” become the variables that determine speed—a Track 3 topic.

Key concepts:

Bottleneck; compute; autonomy.

What it is:

Demis Hassabis (DeepMind): today it is “soft” (AI makes engineers faster), not “hard” (a model disappears into a data center and returns superintelligent).

Why learn this:

This separates what is actually happening from a fictional scenario—the line between observation and speculation.

Key concepts:

Soft × hard; superintelligence (ASI).

What it is:

What it means to assign a probability to an idea. The ~60% by 2028 is one person’s calibrated estimate, not a field-wide consensus. (according to the video—unverified).

Why learn this:

This trains you to distinguish an honest probabilistic opinion from a guaranteed prediction—the heart of critical reading.

Key concepts:

Subjective probability; estimate × consensus; solid evidence vs. hype.

What it is:

RSI is not a conscious robot waking up, nor is it “the singularity.” It is a bottleneck shift in an engineering process.

Why learn this:

This avoids two opposite mistakes: science-fiction panic and dismissing everything as hype.

Key concepts:

Singularity; AGI vs. ASI; consciousness.

What it is:

The course’s tone: informed attention, not certainty or despair. And a map of each track: Evidence (T2) and Safety & 2028 (T3).

Why learn this:

This sets the stance that makes everything else useful: read numbers without panicking or ignoring them.

Key concepts:

Attention × panic; the solid-vs-hype method; course map.

View full module
1.2~45 min · 6 topics

💻 Why it starts with code

Self-improvement appears first in software because the loop closes in seconds. Learn how agents run on their own and why humans move up a level instead of disappearing.

0 of 6
What it is:

In software, the cycle is: write → run → check whether it passed → try again—and it happens in seconds.

Why learn this:

This is the mechanical reason self-improvement starts with code, not biology or physics.

Key concepts:

Feedback loop; short cycle; iteration.

What it is:

In the physical world, an experiment takes weeks or months (a cell culture, a chemical reaction). In software, the loop closes in seconds.

Why learn this:

This speed difference explains WHY self-improvement appears first in code.

Key concepts:

Loop speed; experiment cost.

What it is:

An “agent” is AI that carries out steps on its own: run code, edit a file, test, and try again—with no human at every step.

Why learn this:

It is the “worker” that actually keeps the fast feedback loop running for hours.

Key concepts:

Agent; execution autonomy; closed loop.

What it is:

AI proposes a change → tests it → keeps the winner → repeats, searching for solutions faster than humans. The category is well established; the exact name (“Gemini-guided evolutionary agent”) is a claim from the video (unverified).

Why learn this:

This shows how the loop scales: many attempts in parallel, with automatic selection of the best.

Key concepts:

Evolutionary search; winner selection; parallelism.

What it is:

Anthropic publicly says Claude writes most code that gets merged. Exact figures (>80%, ~8× per engineer, ~5 months without typing) come from the video (unverified).

Why learn this:

This is the strongest concrete example of “soft” RSI running today inside a frontier lab.

Key concepts:

Productivity; solid evidence vs. exact figures; internal loop.

What it is:

The human role shifts from typing to directing, reviewing, and deciding what matters. Humans are not “out”; they are “above.”

Why learn this:

It answers the inevitable question (“so, do programmers disappear?”) honestly: the work moves up a level.

Key concepts:

Human above, not out; review; direction.

View full module
← Back to the beginning Track 2 · The evidence →