Rosa Del Mar

Issue 24 2026-01-24

Rosa Del Mar

Daily Brief

Issue 24 2026-01-24

Process Shift Prototype First Vs Upfront Artifacts

Issue 24 Edition 2026-01-24 4 min read
General
Sources: 1 • Confidence: Low • Updated: 2026-02-06 16:59

Key takeaways

  • The traditional design process emphasizing extensive upfront research and artifacts before building anything may be outdated for today's world.
  • AI-assisted programming reduces the cost of building the wrong thing by making implementation faster.
  • When people can make almost anything, differentiation shifts from ability-to-build toward choosing and curating what to build.
  • The author reports a personal bias toward prototyping due to being a compulsive prototyper.
  • Jenny Wen (Design Lead at Anthropic and former Director of Design at Figma) delivered a keynote at Hatch Conference in Berlin described as provocative.

Sections

Process Shift Prototype First Vs Upfront Artifacts

The corpus frames the traditional research-to-artifacts-to-build sequence as potentially outdated and links that claim to reduced iteration costs from AI. The core point is presented as a critique rather than a validated conclusion, and the corpus does not specify boundary conditions where traditional process remains beneficial.

  • The traditional design process emphasizing extensive upfront research and artifacts before building anything may be outdated for today's world.
  • AI-assisted programming reduces the cost of building the wrong thing by making implementation faster.
  • AI tools make prototyping more accessible and less time-consuming than it used to be.
  • Historically, choosing the wrong design direction could waste months of development time, making such mistakes extremely expensive.

Ai Reduces Iteration Costs

The corpus asserts that AI reduces time/effort for prototyping and implementation, and contrasts this with a historical baseline where wrong directions wasted months. This cluster supports a change in the economics of iteration (lower cost of mistakes) but does not provide quantified magnitudes or corroborating measurements.

  • AI-assisted programming reduces the cost of building the wrong thing by making implementation faster.
  • AI tools make prototyping more accessible and less time-consuming than it used to be.
  • Historically, choosing the wrong design direction could waste months of development time, making such mistakes extremely expensive.

Differentiation Shifts To Curation

The corpus proposes that as making becomes easier, competitive advantage moves toward selecting and curating what to build. This is a directional claim about strategy and differentiation, but it is not grounded in examples, market structure specifics, or observed outcomes within the corpus.

  • When people can make almost anything, differentiation shifts from ability-to-build toward choosing and curating what to build.

Epistemic Qualifiers And Bias

The corpus includes an explicit note that the author is biased toward prototyping. This reduces confidence in any recommendation that favors prototyping without independent validation, but it does not by itself refute the underlying AI-driven mechanisms asserted elsewhere.

  • The author reports a personal bias toward prototyping due to being a compulsive prototyper.

Unknowns

  • In which product contexts (team size, domain complexity, safety/regulation level) does a prototype-first approach outperform a traditional artifact-heavy process, and where does it underperform?
  • What are the measurable deltas in cycle time, rework rates, and product outcomes attributable to AI-assisted prototyping and AI-assisted programming in this setting?
  • Does lowering implementation cost actually reduce the total cost of building the wrong thing, or does it merely shift cost into evaluation, integration, maintenance, and support?
  • What concrete evidence supports the claim that differentiation is increasingly driven by curation/selection rather than ability-to-build?
  • To what extent are the process recommendations influenced by individual preference (including the author's stated bias) versus validated organizational learning?

Investor overlay

Read-throughs

  • Increased demand for AI-assisted programming and prototyping tools as teams shift toward building earlier and iterating faster, reducing reliance on extensive upfront artifacts.
  • Greater strategic importance of product discovery, prioritization, and evaluation workflows if differentiation shifts from ability-to-build toward selecting what to build.
  • Pressure on traditional artifact-heavy design research workflows where value is tied to documentation volume rather than iteration speed, especially in fast-moving product contexts.

What would confirm

  • Measured improvements from AI-assisted prototyping such as shorter cycle time, fewer major reworks, or faster validated learning compared with artifact-heavy baselines.
  • Clear boundary conditions documented by organizations showing prototype-first works better in specific domains and team setups, with repeatable playbooks not tied to individual preference.
  • Concrete evidence that competitive advantage is driven by selection and curation, such as consistent outperformance from better prioritization and evaluation rather than superior implementation speed.

What would kill

  • Data showing lower implementation cost does not reduce total cost of wrong builds because evaluation, integration, maintenance, and support costs rise and dominate outcomes.
  • Prototype-first approaches underperform in complex, safety-critical, or regulated settings where upfront research and artifacts remain necessary to avoid downstream failures.
  • No measurable delta in product outcomes or rework rates despite faster AI-assisted building, implying speed gains do not translate into better decisions or differentiation.

Sources

  1. 2026-01-24 simonwillison.net