How LLMs and coding tools work9 min read

Language Models are Few-Shot Learners

Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, et al. · 2020 · NeurIPS 2020 · Advances in Neural Information Processing Systems 33

GPT-3 showed that a big enough language model can be steered by examples in the prompt alone — no fine-tuning, no gradient updates, no training run of your own.

Read the paper ↗

The short version

→A 175-billion-parameter model, ten times larger than any dense model before it, evaluated on dozens of tasks without ever being trained on them.
→The steering happens entirely in the prompt: show a few examples of the task, then the real input. The paper calls this in-context learning.
→No weights change during in-context learning. Nothing is learned in the ordinary sense — the examples condition the next-token prediction and are gone the moment the request ends.
→This is the paper that made 'prompt engineering' a job. Every few-shot example you paste into a system prompt is this result being cashed in.

Subscription

The Library

Everything running a software business asks of you and none of it is code: onboarding that activates, churn you can actually see coming, marketing with no budget, SEO that finds buyers, and the AI-plus-Ahrefs workflow for writing the blog that feeds it. One new playbook a week, plus the whole archive from day one.

$9/ month · cancel any time

✓A new playbook every week — onboarding, churn, growth, SEO, content
✓The full back catalogue the moment you join, not drip-fed
✓Vendor explainers: Stripe, Supabase, Resend, Vercel, the AI APIs
✓The research shelf — real papers, decoded, on learning and on LLMs
✓One email when a new one lands, off by default in one click
✓Cancel any time; you keep reading to the end of the month

first — it takes ten seconds.

No card is asked for yet — the subscription is recorded and really opens your account. See all plans →