Crazy 8s

Status: Folded · Evidence: P · Family: Divergent ideation and idea expansion · Verdict: fold (2026-06-03)

Use instead: Brainwriting 6-3-5 / NGT

What it is

Crazy 8s (also written Crazy Eights) is a fast divergent-generation drill: you fold one sheet of paper into eight panels, set a timer, and sketch eight distinct ideas in eight minutes - roughly one minute per panel. The point of the ferocious pace is to blow past your first, most obvious idea (usually the least original) and force the obvious-but-unspoken second, third, and eighth options out of hiding before the inner critic can intervene. The sketches are deliberately rough; they only have to communicate a concept, not look good, which is what makes the exercise accessible to people who “can’t draw.” In its home setting, the Google Ventures design sprint, it sits on the “Sketch” step: every team member runs their own Crazy 8s silently and in parallel, and the panels feed the next step (a more detailed solution sketch, then dot-voting).

Stripped of the sketching format and the catchy name, the durable move is narrow and clean: a timeboxed, quota-driven burst of silent parallel generation. Three forcing functions do the work, and only their combination is what people mean by “Crazy 8s”:

1. A fixed quota (eight) - a target that pushes past the two or three ideas a relaxed session stops at.
2. A hard timebox (eight minutes) - a deadline tight enough to suppress evaluation and self-editing mid-generation.
3. Silent, parallel generation - everyone generates independently before anything is shared, so no one anchors on the first voice in the room.

The “eight in eight minutes, sketched” packaging is a memorable wrapper. The mechanism underneath is the generic quantity-over-quality, defer-judgment, generate-before-you-converge stance - run under a stopwatch and a count, in a visual medium. As the verdict section argues, that mechanism is already a shipped move, and the wrapper (a number, a clock, a sketch format) is a mode of it, not a separable cognitive operation.

When it helps / when it misleads

Crazy 8s helps when you need breadth fast and the group’s real enemy is premature settling: the team has latched onto one solution too early, the obvious ideas are dominating, and you want a quick, low-stakes way to widen the option set before any selection. The timebox is genuinely good at defeating the “we must produce something polished” pressure, the quota defeats stopping at three, and the silent-parallel format defeats anchoring on whoever spoke first. In a sketch-friendly, human-facilitated workshop it is a deservedly popular warm-up for the divergent phase.

It misleads or wastes effort when:

• The quota manufactures filler. Demanding exactly eight ideas guarantees eight panels whether or not the problem has eight worthwhile directions. The last few panels are often padding, repetition, or jokes - and they look like content because they occupy a panel. The structure rewards filling the grid over producing distinct ideas.
• It is treated as the answer rather than the raw material. Crazy 8s is pure divergence. It produces candidates, not decisions; eight rough sketches with no convergence, viability check, or selection is an idea dump, not a result. The famous version always pairs it with a voting and narrowing step for exactly this reason.
• The “go fast, no judgment” framing is imported where there is nothing to relieve. This is the decisive caveat for this library. The pace and the silence exist to suppress human self-censorship and human social anchoring - an agent generating ideas has no inner critic to outrun and no first-speaker to conform to. The timebox and the quota are solving a problem the AI context does not have; what survives is the bare quota-and-quota mechanism, which is already shipped.
• The sketch medium is the point but the medium does not transfer. Crazy 8s’ visual format is a real asset for human teams (a sketch externalizes a UI concept faster than a sentence). In a text transcript that asset collapses into a list of eight short idea statements, at which point you are running a timeboxed brainwriting pass and have lost the only thing that distinguished the drill.

What the evidence says

The honest grade is P (practitioner). Crazy 8s is a well-known, widely-taught design-sprint technique with no controlled study of its own; its standing rests on a popular book and a borrowed body of brainstorming research that supports the underlying mechanism but not the eight-panel packaging.

What the record supports. Crazy 8s is a real, named, attributable method with a clear lineage (Jake Knapp and the Google Ventures design-sprint team; codified in the 2016 book Sprint and the Google Design Sprint Kit). Its three forcing functions are each grounded in the broader ideation literature. The defer-judgment / quantity-breeds-quality stance is Osborn’s foundational claim (Applied Imagination, 1953), and at least one quantitative study supports the quantity-quality link directly: Muñoz Adánez (2005), across 69 groups, reported a strong correlation between the number of ideas produced and their rated quality (r = .893). The silent-parallel half of the move is the better-evidenced half: decades of replicated work show that silent, independent, written generation outproduces interacting verbal brainstorming in both quantity and quality, because it removes production blocking and anchoring (Diehl & Stroebe 1987; Mullen, Johnson & Salas 1991 meta-analytic integration). So the mechanism Crazy 8s runs on is real and, in its silent-parallel component, strongly supported.

What the record does NOT support, and the laundering trap. There is no controlled or comparative study I can locate that tests Crazy 8s itself - the eight-sketches-in-eight-minutes drill - against a plain idea list, a longer session, a different quota, or any other generation method, on idea quantity or quality. The design-sprint sources (the Sprint book, the Google Design Sprint Kit, and the dense layer of consultancy and blog write-ups) are practitioner accounts and procedure descriptions, not experiments. The supportive numbers belong to the general mechanism, not to this packaging: Muñoz Adánez’s r = .893 is a finding about quantity and quality across brainstorming groups, not a measurement of the eight-panel format, the eight-minute timebox, or sketching; the brainwriting meta-analytic results are about silent parallel generation in general. Citing any of them as evidence for Crazy 8s specifically would be exactly the transferred-evidence laundering this library exists to prevent. And the evidence is genuinely mixed even at the level of the general claim: Rietzschel, Nijstad and Stroebe (2006) showed that “productivity is not enough” - groups that generate more ideas do not reliably select the better ones, so a high-quantity drill like Crazy 8s, run without a disciplined selection step, can pile up options without improving the decision. The popular “if you push to eight ideas you get more innovative ones” claim has no traceable primary source measuring eight as a threshold; it is a plausible restatement of the quantity heuristic, and any specific yield figure attached to it traces to no study and is excluded from the grade.

Transfer caveat (required). All of the located evidence is from human subjects in group, classroom, and workshop settings; none of it studies Crazy 8s, or the quota-plus-timebox mechanism, performed by or with an AI agent. The evidence is transferred from human contexts and has not been validated for AI-augmented use. The conservative governing grade is therefore P: an established practitioner technique whose own effectiveness is untested, whose supportive evidence belongs to the general ideation mechanism it shares with shipped skills, and whose human-facilitation rationale (outrun the inner critic, beat the clock) does not transfer to an agent.

Why it is / is not a skill here

Verdict: Fold into brainwriting. Crazy 8s is, mechanically, a timeboxed mode of brainwriting - the same silent-parallel-generation engine, run under a stopwatch and an idea quota, rendered as sketches. It does not clear the Build burden, which is to name one distinct, durable cognitive move that no shipped skill already produces above the roughly 20% overlap ceiling.

The shared machinery is direct. brainwriting (S-tier, this library’s strongest ideation anchor) already owns “generate several genuinely independent idea streams in parallel, without anchoring on the first voice, then build on and consolidate them.” Crazy 8s is that same move with two parameters pinned: a fixed quota (eight) and a hard timebox (eight minutes), plus a sketch rendering. A preset for parameters an existing skill already exposes - how many ideas, how fast - is a mode of that skill, not a new skill. Tellingly, the catalog already recorded this when it graded brainwriting: its reasoning states that brainwriting “absorbs silent-writing-before-discussion and the timeboxed Crazy-8s mode” by design. The fold is not a demotion discovered after the fact; the timeboxed-quota variant was folded in as part of what makes brainwriting the family’s anchor.

The three things that make Crazy 8s feel distinct each resolve to a mode or a non-transferring asset, not a mechanism:

• The eight-idea quota is an idea-quota - and quotas are a standard knob on parallel generation, already present in brainwriting’s “aim for a quota per stream.” Pinning it to eight is a setting, not an operation.
• The eight-minute timebox is a facilitation parameter. Its value (suppress mid-generation self-editing, force past the obvious) is real for human groups but is precisely the human-psychology benefit that does not transfer to an agent, which has no self-editing reflex to outrun. For the library’s AI-agent context the timebox is moot, which removes the main thing that could have argued for a standalone skill.
• The sketch medium is a genuine asset for human design teams, but it is a rendering format, not a cognitive move, and it degrades to a list of short idea statements in a text transcript - at which point Crazy 8s simply is a timeboxed brainwriting pass.

Why fold rather than recipe or reject. It is not a recipe: a recipe is a fixed chain of two or more separable shipped skills, whereas Crazy 8s is a single skill (brainwriting) run with a pinned quota and timebox - one mode, not a chain. And reject would be less informative than fold: the technique is real, famous, taught everywhere, and genuinely useful in its home setting, so the honest service is to point the reader to where its durable move already lives rather than pretend it does not exist. This is the same call the catalog made when it folded lotus-blossom and fishbone-ishikawa into issue-tree - a famous, evocatively-named technique whose fame is its packaging (a number, a grid, a fish, a stopwatch) rather than a move the library does not already ship. The learning value of the NO: a beloved workshop drill is not automatically a skill. Crazy 8s is a stopwatch-and-count wrapper on silent parallel generation, and a library that ships mechanisms for AI agents rather than timeboxed sketch formats for human rooms documents it and folds it into brainwriting.

Lineage and who to read

Crazy 8s was popularized by Jake Knapp during his decade at Google and Google Ventures, where he created the design-sprint process; it is codified as part of the sprint’s “Sketch” day in the 2016 book Sprint, written with John Zeratsky and Braden Kowitz (both former Google Ventures design partners), and in Google’s open Design Sprint Kit. The drill’s deeper roots are in the divergent-ideation tradition: Alex Osborn’s “defer judgment / reach for quantity” principles (Applied Imagination, 1953) are the rationale for the quota and the no-criticism framing, and the silent-parallel half traces to the brainwriting and nominal-group lineage (Rohrbach’s 6-3-5, Delbecq and Van de Ven’s Nominal Group Technique) that the shipped brainwriting skill documents. For the honest limits, read Diehl and Stroebe (1987) and the Mullen, Johnson and Salas (1991) meta-analysis on what actually drives ideation productivity, and Rietzschel, Nijstad and Stroebe (2006) on why high idea-quantity does not automatically yield good idea-selection. “Crazy 8s” and “Crazy Eights” are generic descriptive names in common use within the design-sprint community with no located trademark registration; this entry is documented descriptively, with credit to Knapp and the Google Ventures team, and is not flagged as branded.

Named sources

• Jake Knapp, John Zeratsky and Braden Kowitz, Sprint: How to Solve Big Problems and Test New Ideas in Just Five Days (Simon & Schuster, 2016). The book that codified the design sprint and the Crazy 8s sketch drill within it; a practitioner method, not a study. Practitioner / foundational. (P)
• Google Ventures, “Crazy 8’s,” Design Sprint Kit (designsprintkit.withgoogle.com). The canonical procedure: fold one sheet into eight, sketch eight ideas in eight minutes, one minute per panel, then share and vote. Practitioner reference; cites no controlled study. (P)
• Alex F. Osborn, Applied Imagination (Scribner, 1953). Origin of the “defer judgment” and “reach for quantity” principles that justify the quota-and-no-criticism stance. Practitioner / foundational. (P)
• Alfredo Muñoz Adánez, “Does Quantity Generate Quality? Testing the Fundamental Principle of Brainstorming,” The Spanish Journal of Psychology 8(2) (2005). Across 69 groups, reported a strong quantity-quality correlation (r = .893). Supports the general mechanism Crazy 8s runs on, NOT the eight-panel packaging; cited to locate the strongest supportive evidence and to mark that it does not test the drill. (P / quantitative)
• Michael Diehl and Wolfgang Stroebe, “Productivity Loss in Brainstorming Groups: Toward the Solution of a Riddle,” Journal of Personality and Social Psychology 53(3) (1987): 497-509. Isolated production blocking (the turn-taking bottleneck) as the dominant cause of brainstorming productivity loss - the result that motivates silent parallel generation, the half of Crazy 8s that transfers. Measures group brainstorming, not Crazy 8s. (M, for the general mechanism)
• Brian Mullen, Craig Johnson and Eduardo Salas, “Productivity Loss in Brainstorming Groups: A Meta-Analytic Integration,” Basic and Applied Social Psychology 12(1) (1991): 3-23. Meta-analysis: interacting brainstorming groups are reliably less productive than nominal/written groups in quantity and quality; productivity loss is worse for vocalized than written contributions. Supports the silent-written half of the move. (M, for the general mechanism)
• Eric F. Rietzschel, Bernard A. Nijstad and Wolfgang Stroebe, “Productivity is not enough: A comparison of interactive and nominal brainstorming groups on idea generation and selection,” Journal of Experimental Social Psychology 42 (2006): 244-251. Found that generating more ideas does not reliably improve which ideas get selected - the counterweight to “more is better,” and the reason a high-quota drill needs a disciplined selection step it does not itself provide. (M, critical-adjacent)

Excluded under the evidence rule: there is no controlled study of Crazy 8s itself, and no traceable primary source for the popular “push to eight ideas and you get more innovative ones” threshold claim; any specific yield figure attached to it is not counted toward the grade. The supportive brainstorming statistics (Muñoz Adánez r = .893; the Diehl & Stroebe and Mullen et al. results) measure the general mechanism, not the eight-panel format, and are recorded as mechanism evidence, not as evidence for the drill.