Deep Dive · XiaoHu Explains

The Claude Fable 5 Field Guide: 8 "Finding Your Unknowns" Prompt Patterns, From Blind Spot Scans to Post-Read Quizzes

A talk transcript from Claude Code team member Thariq Shihipar: the bottleneck with new models is no longer the model itself — it's whether you can clearly articulate your own unknowns
TL;DR
  • Anthropic Claude Code team member Thariq Shihipar gave a talk titled "A Field Guide to Fable" at the AI Engineer World's Fair, with a companion official blog post published the same day — reportedly racking up about two million reads within three days.
  • Core methodology: break down "what you know" into four quadrants — known knowns, known unknowns, unknown knowns, unknown unknowns. Reducing and pre-planning for these "unknowns" is the core skill for getting the most out of the new generation of models.
  • The talk lays out 8 ready-to-copy prompt patterns, organized by before, during, and after you start work, each with a ready-made example prompt.
  • A counterintuitive first-hand disclosure: Claude Code's system prompt was recently cut by 80% — for the new model, examples actually constrain it more than they help, and the direction has shifted from "giving constraints" to "giving context."
  • Thariq used this exact method to cut Fable's launch video from scratch — video editing was a completely unfamiliar domain for him.
⚑Stance flag: This is Anthropic's (Claude Code team's) own talk at the AI Engineer conference, plus a companion official blog post of the same title, about how to get the most out of their own Claude Fable 5 — vendor content. The talk contains almost no benchmarks, architecture, or hard numbers for Fable; it's mostly methodology and first-hand observations on internal prompt engineering.
1What This Talk Is About

What exactly is this talk about

Anthropic Claude Code team member Thariq Shihipar gave a keynote at the AI Engineer World's Fair, simultaneously publishing an official blog post of the same title, "A field guide to Claude Fable 5: Finding your unknowns."

This talk is about one very specific thing: how capable the Fable generation of models has become, the bottleneck has shifted away from the model itself and onto you — onto whether you can clearly articulate "what you don't know." The talk contains almost no hard numbers or architecture details for Fable; it's entirely about "how to collaborate with a more capable model."
📌
Why it's worth watching: Claude Code's system prompt was recently cut by 80%, shifting direction from "giving constraints" to "giving context" — a first-hand look at internal prompt-engineering changes at Anthropic. It also distills "how to get the most out of a more capable model" into 8 copy-paste-ready prompt patterns, each with a ready-made example.

One word keeps coming up in the talk: unhobbling. The idea is that what actually constrains a model is often the framework you've put on it yourself, and your old prompt-writing habits — frameworks that are essentially a reflection of how well you understand the model. Change them, and you can unlock capability the model already had.

An analogy

Take a horse that can already run, and remove the unnecessary reins and blinders.

~2 million
Reported read count for the companion official blog post within three days of publication
80%
The portion of Claude Code's system prompt recently trimmed, shifting from "giving constraints" to "giving context"
8
The number of copy-paste-ready prompt patterns the talk distills
Full talk · bilingual Chinese/English subtitles (about 19 minutes, click the cover to load)　·　Original video on YouTube (English) ↗
2Background · Prompts Doing Less

Models get smarter in "spikes" — prompts, meanwhile, need to do less

To understand why this generation of models needs "fewer examples, more context," look at two pieces of background first. This section is setup — the main point comes later.

The first piece of background is capability overhang: models don't get stronger evenly across the board — they suddenly leap ahead on certain specific tasks. There's a viral post online asking "why can't large models say which Pokémon names end in AW" — a regular chat model can't answer, even though it clearly knows all the Pokémon names. But Claude Code can answer, because it will fetch every Pokémon and write a script to filter for names ending in AW. Same "knowledge," but whether or not you give it a code-execution tool makes a world of difference in the result.

An analogy

Like ore veins buried at different depths underground — tools are the pickaxe that decides which vein you can reach. Half the work of using a new model well is figuring out what new possibilities actually exist now.

The second piece of background is more counterintuitive: system prompts are getting smaller. Thariq splits prompt-engineering best practice into three stages, and you can see the trend.