Log-Scale Thinking: When Linear Lies

Some things grow multiplicatively, not additively. Log scales reveal patterns that linear scales hide, especially for anything related to scale or growth.

28 min · Reviewed 2026

Linear Scales Break Down Quickly

Plot GDP of every country on a linear scale and the US, China, and Japan dominate; everyone else looks like zero. Plot on a log scale and you suddenly see that Norway is near the Dominican Republic by income per capita. Log scales make multiplicative differences visible.

When log scale helps

Exponential growth (virus spread, compound interest)
Power-law distributions (wealth, view counts)
Measurements spanning many orders of magnitude (star masses, earthquake energy)
Human perception (sound is heard logarithmically; decibels are a log scale)
Model training loss curves (always plot on log-y)

Model scaling: a perfect log-scale story

Scaling laws for neural networks are almost always shown on log-log plots. The relationship between compute and loss looks like a straight line on log scale, but would be a dramatic curve on linear scale. This is why every paper on training curves uses logarithmic axes.

import matplotlib.pyplot as plt
import numpy as np

training_steps = np.array([100, 1000, 10000, 100000])
loss = np.array([4.8, 3.2, 2.1, 1.4])

fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(10, 4))

ax1.plot(training_steps, loss, marker='o')
ax1.set_title('Linear scale (hides the pattern)')
ax1.set_xlabel('Steps')

ax2.plot(training_steps, loss, marker='o')
ax2.set_xscale('log')
ax2.set_title('Log scale (reveals the pattern)')
ax2.set_xlabel('Steps (log)')

plt.tight_layout()
plt.show()Log scale reveals training dynamics

Two practical tricks

Use log(1 + x) instead of log(x) when data contains zeros
For percentages or ratios, use log-odds: log(p / (1-p))
In pandas: df['log_followers'] = np.log1p(df['followers'])

The big idea: the world runs on multiplication more than addition. Log scales are how you see growth, scale, and skew clearly. Get comfortable reading them, and data starts making much more sense.

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-data-log-scale-thinking

What is the core idea behind "Log-Scale Thinking: When Linear Lies"?
1. Some things grow multiplicatively, not additively. Log scales reveal patterns that linear scales hide, especially for anything related to scale or growth.
2. bootstrap
3. Ask: is the measurement the thing I care about, or a proxy?
4. Fairness has multiple, sometimes conflicting definitions
Which term best describes a foundational idea in "Log-Scale Thinking: When Linear Lies"?
1. log scale
2. logarithm
3. log-log plot
4. exponential
A learner studying Log-Scale Thinking: When Linear Lies would need to understand which concept?
1. logarithm
2. log-log plot
3. log scale
4. exponential
Which of these is directly relevant to Log-Scale Thinking: When Linear Lies?
1. logarithm
2. log scale
3. exponential
4. log-log plot
Which of the following is a key point about Log-Scale Thinking: When Linear Lies?
1. Exponential growth (virus spread, compound interest)
2. Power-law distributions (wealth, view counts)
3. Measurements spanning many orders of magnitude (star masses, earthquake energy)
4. Human perception (sound is heard logarithmically; decibels are a log scale)
Which of these does NOT belong in a discussion of Log-Scale Thinking: When Linear Lies?
1. Measurements spanning many orders of magnitude (star masses, earthquake energy)
2. Exponential growth (virus spread, compound interest)
3. Power-law distributions (wealth, view counts)
4. bootstrap
Which statement is accurate regarding Log-Scale Thinking: When Linear Lies?
1. For percentages or ratios, use log-odds: log(p / (1-p))
2. In pandas: df['log_followers'] = np.log1p(df['followers'])
3. Use log(1 + x) instead of log(x) when data contains zeros
4. bootstrap
What is the key insight about "What a log scale is" in the context of Log-Scale Thinking: When Linear Lies?
1. bootstrap
2. Ask: is the measurement the thing I care about, or a proxy?
3. Fairness has multiple, sometimes conflicting definitions
4. A log scale shows ratios, not absolute differences. Moving one unit up means multiplying by 10 (for base-10 log).
What is the key insight about "Log-scale honesty" in the context of Log-Scale Thinking: When Linear Lies?
1. Log scales can also mislead the untrained eye. A small move on a log plot can represent a massive difference.
2. bootstrap
3. Ask: is the measurement the thing I care about, or a proxy?
4. Fairness has multiple, sometimes conflicting definitions
Which statement accurately describes an aspect of Log-Scale Thinking: When Linear Lies?
1. bootstrap
2. Plot GDP of every country on a linear scale and the US, China, and Japan dominate; everyone else looks like zero.
3. Ask: is the measurement the thing I care about, or a proxy?
4. Fairness has multiple, sometimes conflicting definitions
What does working with Log-Scale Thinking: When Linear Lies typically involve?
1. bootstrap
2. Ask: is the measurement the thing I care about, or a proxy?
3. Scaling laws for neural networks are almost always shown on log-log plots.
4. Fairness has multiple, sometimes conflicting definitions
Which of the following is true about Log-Scale Thinking: When Linear Lies?
1. bootstrap
2. Ask: is the measurement the thing I care about, or a proxy?
3. Fairness has multiple, sometimes conflicting definitions
4. The big idea: the world runs on multiplication more than addition. Log scales are how you see growth, scale, and skew clearly.
Which best describes the scope of "Log-Scale Thinking: When Linear Lies"?
1. It focuses on Some things grow multiplicatively, not additively. Log scales reveal patterns that linear scales hid
2. It is unrelated to foundations workflows
3. It applies only to the opposite beginner tier
4. It was deprecated in 2024 and no longer relevant
Which section heading best belongs in a lesson about Log-Scale Thinking: When Linear Lies?
1. bootstrap
2. When log scale helps
3. Ask: is the measurement the thing I care about, or a proxy?
4. Fairness has multiple, sometimes conflicting definitions
Which section heading best belongs in a lesson about Log-Scale Thinking: When Linear Lies?
1. bootstrap
2. Ask: is the measurement the thing I care about, or a proxy?
3. Model scaling: a perfect log-scale story
4. Fairness has multiple, sometimes conflicting definitions

← Back to interactive lesson

Tendril · Creators · AI Foundations

Log-Scale Thinking: When Linear Lies

Some things grow multiplicatively, not additively. Log scales reveal patterns that linear scales hide, especially for anything related to scale or growth.

28 min · Reviewed 2026

Linear Scales Break Down Quickly

When log scale helps

Exponential growth (virus spread, compound interest)
Power-law distributions (wealth, view counts)
Measurements spanning many orders of magnitude (star masses, earthquake energy)
Human perception (sound is heard logarithmically; decibels are a log scale)
Model training loss curves (always plot on log-y)

Model scaling: a perfect log-scale story

import matplotlib.pyplot as plt
import numpy as np

training_steps = np.array([100, 1000, 10000, 100000])
loss = np.array([4.8, 3.2, 2.1, 1.4])

fig, (ax1, ax2) = plt.subplots(1, 2, figsize=(10, 4))

ax1.plot(training_steps, loss, marker='o')
ax1.set_title('Linear scale (hides the pattern)')
ax1.set_xlabel('Steps')

ax2.plot(training_steps, loss, marker='o')
ax2.set_xscale('log')
ax2.set_title('Log scale (reveals the pattern)')
ax2.set_xlabel('Steps (log)')

plt.tight_layout()
plt.show()Log scale reveals training dynamics

Two practical tricks

Use log(1 + x) instead of log(x) when data contains zeros
For percentages or ratios, use log-odds: log(p / (1-p))
In pandas: df['log_followers'] = np.log1p(df['followers'])

The big idea: the world runs on multiplication more than addition. Log scales are how you see growth, scale, and skew clearly. Get comfortable reading them, and data starts making much more sense.

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-data-log-scale-thinking

What is the core idea behind "Log-Scale Thinking: When Linear Lies"?
1. Some things grow multiplicatively, not additively. Log scales reveal patterns that linear scales hide, especially for anything related to scale or growth.
2. bootstrap
3. Ask: is the measurement the thing I care about, or a proxy?
4. Fairness has multiple, sometimes conflicting definitions
Which term best describes a foundational idea in "Log-Scale Thinking: When Linear Lies"?
1. log scale
2. logarithm
3. log-log plot
4. exponential
A learner studying Log-Scale Thinking: When Linear Lies would need to understand which concept?
1. logarithm
2. log-log plot
3. log scale
4. exponential
Which of these is directly relevant to Log-Scale Thinking: When Linear Lies?
1. logarithm
2. log scale
3. exponential
4. log-log plot
Which of the following is a key point about Log-Scale Thinking: When Linear Lies?
1. Exponential growth (virus spread, compound interest)
2. Power-law distributions (wealth, view counts)
3. Measurements spanning many orders of magnitude (star masses, earthquake energy)
4. Human perception (sound is heard logarithmically; decibels are a log scale)
Which of these does NOT belong in a discussion of Log-Scale Thinking: When Linear Lies?
1. Measurements spanning many orders of magnitude (star masses, earthquake energy)
2. Exponential growth (virus spread, compound interest)
3. Power-law distributions (wealth, view counts)
4. bootstrap
Which statement is accurate regarding Log-Scale Thinking: When Linear Lies?
1. For percentages or ratios, use log-odds: log(p / (1-p))
2. In pandas: df['log_followers'] = np.log1p(df['followers'])
3. Use log(1 + x) instead of log(x) when data contains zeros
4. bootstrap
What is the key insight about "What a log scale is" in the context of Log-Scale Thinking: When Linear Lies?
1. bootstrap
2. Ask: is the measurement the thing I care about, or a proxy?
3. Fairness has multiple, sometimes conflicting definitions
4. A log scale shows ratios, not absolute differences. Moving one unit up means multiplying by 10 (for base-10 log).
What is the key insight about "Log-scale honesty" in the context of Log-Scale Thinking: When Linear Lies?
1. Log scales can also mislead the untrained eye. A small move on a log plot can represent a massive difference.
2. bootstrap
3. Ask: is the measurement the thing I care about, or a proxy?
4. Fairness has multiple, sometimes conflicting definitions
Which statement accurately describes an aspect of Log-Scale Thinking: When Linear Lies?
1. bootstrap
2. Plot GDP of every country on a linear scale and the US, China, and Japan dominate; everyone else looks like zero.
3. Ask: is the measurement the thing I care about, or a proxy?
4. Fairness has multiple, sometimes conflicting definitions
What does working with Log-Scale Thinking: When Linear Lies typically involve?
1. bootstrap
2. Ask: is the measurement the thing I care about, or a proxy?
3. Scaling laws for neural networks are almost always shown on log-log plots.
4. Fairness has multiple, sometimes conflicting definitions
Which of the following is true about Log-Scale Thinking: When Linear Lies?
1. bootstrap
2. Ask: is the measurement the thing I care about, or a proxy?
3. Fairness has multiple, sometimes conflicting definitions
4. The big idea: the world runs on multiplication more than addition. Log scales are how you see growth, scale, and skew clearly.
Which best describes the scope of "Log-Scale Thinking: When Linear Lies"?
1. It focuses on Some things grow multiplicatively, not additively. Log scales reveal patterns that linear scales hid
2. It is unrelated to foundations workflows
3. It applies only to the opposite beginner tier
4. It was deprecated in 2024 and no longer relevant
Which section heading best belongs in a lesson about Log-Scale Thinking: When Linear Lies?
1. bootstrap
2. When log scale helps
3. Ask: is the measurement the thing I care about, or a proxy?
4. Fairness has multiple, sometimes conflicting definitions
Which section heading best belongs in a lesson about Log-Scale Thinking: When Linear Lies?
1. bootstrap
2. Ask: is the measurement the thing I care about, or a proxy?
3. Model scaling: a perfect log-scale story
4. Fairness has multiple, sometimes conflicting definitions

← Back to interactive lesson